Author response: Metabolic sensing in AgRP neurons integrates homeostatic state with dopamine signalling in the striatum
Year: 2021
Type: peer-review
Abstract: Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Agouti-related peptide (AgRP) neurons increase motivation for food, however, whether metabolic sensing of homeostatic state in AgRP neurons potentiates motivation by interacting with dopamine reward systems is unexplored. As a model of impaired metabolic-sensing, we used the AgRP-specific deletion of carnitine acetyltransferase (Crat) in mice. We hypothesised that metabolic sensing in AgRP neurons is required to increase motivation for food reward by modulating accumbal or striatal dopamine release. Studies confirmed that Crat deletion in AgRP neurons (KO) impaired ex vivo glucose-sensing, as well as in vivo responses to peripheral glucose injection or repeated palatable food presentation and consumption. Impaired metabolic-sensing in AgRP neurons reduced acute dopamine release (seconds) to palatable food consumption and during operant responding, as assessed by GRAB-DA photometry in the nucleus accumbens, but not the dorsal striatum. Impaired metabolic-sensing in AgRP neurons suppressed radiolabelled 18F-fDOPA accumulation after ~30 min in the dorsal striatum but not the nucleus accumbens. Impaired metabolic sensing in AgRP neurons suppressed motivated operant responding for sucrose rewards during fasting. Thus, metabolic-sensing in AgRP neurons is required for the appropriate temporal integration and transmission of homeostatic hunger-sensing to dopamine signalling in the striatum. Editor's evaluation This manuscript will be of broad interest to behavioral neuroscientists studying energy homeostasis, hypothalamic feeding circuits, and dopamine. The paper uses a genetic mouse model to study critical connections between homeostatic circuitry and dopamine release in response to food reward. The experimental results support key claims of the paper, and tie in nicely with previously published data. https://doi.org/10.7554/eLife.72668.sa0 Decision letter Reviews on Sciety eLife's review process Introduction The motivation to approach and consume food depends not only on the palatability and caloric density of the available food source, but also on the energy state of the organism. For example, when food is abundant many prey species forage within known territories to reduce survival threats (Sih, 1980). Conversely, when food is scarce, animals are motivated to take greater risks and forage within unfamiliar territories to search for food (Whitham and Mathis, 2000). Thus, the motivation to seek palatable, energy-dense food evolved as a key mechanism for survival and maturation in an environment with limited food resources. Given that heightened motivation for palatable food in an environment of low food availability shaped an evolutionary benefit, it is not surprising that homeostatic feeding circuits can have a profound effect on motivation. With high caloric foods readily available, these circuits may contribute to the overconsumption of highly palatable and calorie dense foods, which is a leading cause for today’s obesity crisis. Indeed, human evidence shows that fasting biases reward systems to high caloric foods (Goldstone et al., 2009; Cameron et al., 2014). Agouti-related peptide (AgRP) neurons in the arcuate nucleus of the hypothalamus (ARC) are a critical population of hunger-sensitive neurons that primarily function to increase appetite or conserve energy (Aponte et al., 2011; Gropp et al., 2005; Krashes et al., 2011; Luquet et al., 2005; Ruan et al., 2014; Siemian et al., 2021). A key element of appetitive behaviour is the increased motivation for goal-directed outcomes. Intriguingly, hunger has been used for decades in behavioural neuroscience to improve performance and learning in operant tasks (Ramond, 1954). AgRP neurons function within this framework by increasing the willingness to work for food and food rewards to the same level as that seen in fasted mice (Krashes et al., 2011; Atasoy et al., 2012; Betley et al., 2015; Chen et al., 2016). Moreover, AgRP neurons influence motivation circuits (Alhadeff et al., 2019; Dietrich et al., 2012; Mazzone et al., 2020), with chemogenetic activation elevating dopamine release in response to food consumption (Alhadeff et al., 2019). Since AgRP neurons are most active during periods of energy deficit (Baskin et al., 1999; Briggs et al., 2011; Mandelblat-Cerf et al., 2015; Schwartz et al., 1998; Yang et al., 2011), we hypothesised that they transmit hunger-specific metabolic information to motivation circuits. However, whether AgRP neurons are required to gauge homeostatic state through metabolic sensing, to influence motivation neural circuits, remains unknown. Sensory cues that predict food availability and palatability rapidly suppress hunger-sensitive AgRP neuronal activity in hungry mice (Betley et al., 2015; Mandelblat-Cerf et al., 2015; Chen et al., 2015; Su et al., 2017). Although this decrease in activity occurs prior to food consumption, post-ingestive caloric feedback is required to sustainably reduce AgRP neuronal activity (Su et al., 2017; Beutler et al., 2017). Indeed, the reduction of AgRP neuronal activity correlates with the number of calories ingested and was not observed after repeated ingestion of a non-caloric sweetened gel (Su et al., 2017). These studies highlight that the maintenance of normal AgRP activity in response to post-ingestive gastrointestinal feedback requires metabolic sensing of available calories in combination with gut-brain neural communication (Su et al., 2017; Beutler et al., 2017; Goldstein et al., 2021). Previously, we demonstrated that carnitine acetyltransferase (Crat) in AgRP neurons was required as a molecular sensor for peripheral substrate utilisation during fasting and refeeding (Reichenbach et al., 2018c). Moreover, Crat activity in AgRP neurons programmed a broad metabolic response of the AgRP proteome and was required to promote normal refeeding after fasting (Reichenbach et al., 2018c). The reduced feeding response after fasting in AgRP Crat KO mice suggested that metabolic sensing of homeostatic state by AgRP neurons transmits hunger-specific metabolic information into neural circuits controlling dopamine signalling and the motivational aspects of food-directed behaviour. Here, we describe experiments demonstrating that impaired metabolic sensing in AgRP neurons, using the conditional deletion of Crat from AgRP neurons as a validated model, reduces motivated behaviour during an operant task and dopamine release in the nucleus accumbens (NAc) or dorsal striatum, over different time frames, in response to palatable food rewards. These studies highlight that metabolic-sensing in AgRP neurons is required for the appropriate temporal integration of hunger-sensing to potentiate food reward-related dopamine release in the striatum and control motivation for palatable food rewards. Results Crat deletion in AgRP neurons is a valid model of impaired metabolic sensing To demonstrate that Crat deletion in AgRP neurons is a reliable model of impaired metabolic sensing, we prepared hypothalamic brain slices from WT and KO mice for electrophysiological characterisation of glucose sensing in AgRP neurons. We detected no differences in the fundamental electrophysiological properties of AgRP neurons including resting membrane potential (Figure 1A), input resistance (Figure 1B), spontaneous firing frequency (Figure 1C), although peak action potential amplitude was significantly lower in KO mice (Figure 1D). Collectively, these studies demonstrate that Crat in AgRP neurons has little effect on the intrinsic electrophysiological properties of the cell. Similarly, no genotype-dependent changes in the frequency of spontaneous excitatory or inhibitory post-synaptic potentials (EPSPs or IPSPs, respectively) were detected (Figure 1—figure supplement 1A, C), indicating no differences in the external synaptic input onto AgRP neurons in WT compared to KO mice. To show impaired glucose-sensing in AgRP neurons, we recorded neuronal responses to an increase in extracellular glucose from a basal level of 2 mM to 5 mM. These glucose concentrations were chosen to represent brain glucose concentrations under fasting and fed conditions, based on estimates in the CSF from fasted and fed rats and mice (van den Top et al., 2017). Glucose-excited neurons were defined based upon a response characterised by membrane potential depolarisation and/or an increase in action potential firing frequency with increased extracellular glucose (Figure 1—figure supplement 1B). Glucose-inhibited cells were identified by responses to increased glucose characterised by membrane potential hyperpolarisation and/or a reduction in spontaneous action potential firing frequency (Figure 1—figure supplement 1D). In WT mice, 48% of cells (n = 10/21) were excited, 33% of cells (n = 7/21) were inhibited and 19% (n = 4/21) were insensitive to changes in extracellular glucose (Figure 1E). In KO mice, the number of glucose-excited cells was reduced compared to WT, with only 28% classified as glucose-excited (n = 7/25). Similarly, the incidence of glucose-inhibited cells was reduced to 16% (n = 4/25) of the population compared to WT mice, with the majority of cells not responding to 5 mM glucose 56% (n = 14/25) and classified as glucose insensitive (Figure 1F). In total 17/21 AgRP neurons responded to an increase in extracellular glucose in WT mice, whereas as only 11/25 AgRP neurons responded to glucose in KO mice, representing a decrease in glucose-responsive neurons (WT 81% vs KO 44%). To examine whether these ex vivo recordings translated to in vivo changes, we injected a cre-dependent GCAMP6s-expressing virus into Agrpcre/wt mice to record AgRP neural activity in freely moving WT and KO mice. An increase in activity to ghrelin was used as an index of correct viral expression and fibre optic placement. Only mice with a maximal peak z-score of >4 were included for analysis in experimental groups, using this criterion 5/19 mice, across both WT and KO mice, were excluded for experimentation (Figure 1—figure supplement 1N-O). Fed or overnight fasted WT and KO mice were injected with glucose (2.25 g/kg) and GCaMP6s fluorescence from AgRP neurons was measured for 10 min post injection (Figure 1—figure supplement 1J, K, M). Glucose injection suppressed AgRP activity in WT and KO mice; however, activity quickly returned to baseline levels in fasted WT but not KO mice (Figure 1—figure supplement 1L). Therefore, despite identical infusion of glucose, the activity of AgRP neurons in KO mice failed to rebound, which could dampen the motivation to seek food. Figure 1 with 2 supplements see all Download asset Open asset Deletion of Crat in AgRP neurons is a model of impaired metabolic sensing. Intrinsic electrophysiological properties were little affected in WT versus KO mice including resting membrane potential (A, n = 22/23 cells WT/KO), input resistance (B n = 13/16 cells WT/KO), spontaneous action potential firing frequency (C n = 14/16 cells WT/KO), although action potential amplitude was significantly reduced in KO AgRP neurons (D, n = 14/16 cells WT/KO). Glucose-responsive profiles of AgRP neurons from WT (21 cells) and KO (25 cells) mice characterised by their response to increased extracellular glucose concentration from 2 mM to 5 mM (E–F). In vivo fiber photometry analysis of AgRP neuronal activity in WT and KO mice (G–O). First exposure to a peanut butter chip (PB) (G, n = 7 WT; n = 5 KO) and time binned summary data (H). GCaMP6s expression in AgRP neurons [scale bar = 100 µm] (I). AgRP neural activity in ad libitum-fed mice in response to repeated PB exposure (J; n = 14 WT; n = 12 KO) and time binned summary data (K–L). AgRP neural activity in fasted mice in response to repeated PB exposure (M; n = 8 WT; n = 7 KO) and time binned summary data (N), n = 8 WT, n = 7 KO). Time binned summary data comparing the response to repeated PB exposure in ad libitum-fed or fasted WT and KO mice (O). Data± SEM; ANOVA with Tukey’s post hoc analysis (H, K, L, N, O) and unpaired students t-test (A–D); a, significant at p < 0.05; ** p < 0.01, *** p < 0.001, **** p < 0.0001. Dashed lines in I, J, and M indicated the moment of contact with a PB contact.1,2 Videos 1 and 2 accompany this figure. Figure 1—source data 1 Deletion of Crat in AgRP neurons is a model of impaired metabolic sensing. https://cdn.elifesciences.org/articles/72668/elife-72668-fig1-data1-v3.zip Download elife-72668-fig1-data1-v3.zip To examine these differences in AgRP response in a more physiological relevant scenario, fed WT or KO mice were presented with a novel peanut butter chip (PB) and the fall in AgRP activity was measured, as previously described (Betley et al., 2015; Mandelblat-Cerf et al., 2015; Chen et al., 2015). The fall in AgRP activity was similar in WT and KO mice in response to first PB exposure (Figure 1G–H), however, with repeated exposure across different sessions WT mice showed a greater suppression of activity relative to first exposure (Figure 1—figure supplement 1E; Figure 1L, Video 1), consistent with previous studies showing that the magnitude of activity suppression is proportional to calories consumed (Su et al., 2017; Beutler et al., 2017). Unlike WT mice, KO mice did not exhibit any further reduction in AgRP activity to repeated PB exposure (Figure 1—figure supplement 1F, Figure 1L, Video 2) and the reduction in AgRP neural activity in KO mice to repeated PB exposure was significantly attenuated compared to WT mice at all time points after PB exposure either in ad libitum-fed (Figure 1J–K) or fasted mice (Figure 1M–N). These results suggest KO mice could not integrate caloric information associated with PB consumption either in the fed or fasted state and therefore did not show the expected greater reduction in activity. To determine if this response was specific to metabolic sensing of calorie content and not a generalised response due to impaired AgRP neuronal function caused by Crat deletion, we examined ghrelin-induced food intake and AgRP neural activity (Figure 1—figure supplement 2). IP injection of ghrelin (1 mg/kg) significantly increased food intake within 4 hr (Figure 1—figure supplement 2B; main effect of ghrelin); however, no differences in genotype were observed. IP ghrelin significantly increased AgRP neural activity (Figure 1—figure supplement 2D-E; main effect of ghrelin); however, no differences in genotype were observed and the presentation of chow diet to ghrelin-injected mice produced an equal suppression in AgRP neural activity (Figure 1—figure supplement 2F-G; main effect of food access), again without differences between WT and KO mice. Finally, we observed no differences in anxiety-like behaviour in the light-dark box or elevated-plus maze (Figure 1—figure supplement 2I-P), although KO mice showed reduced locomotor activity in the elevated-plus maze but not light-dark box. The lack of genotype differences in ghrelin-induced food intake and AgRP activity, or differences in anxiety-like behaviour, which can be modulated by AgRP neurons (Dietrich et al., 2015; Padilla et al., 2016), suggests the deletion of Crat from AgRP neurons does not cause a global deficit in function. Taken together, the specific impairments in ex vivo glucose sensing and in vivo responding to IP glucose or palatable calorie consumption show that Crat deletion in AgRP neurons is a useful model to assess the impact of impaired metabolic sensing in AgRP neurons on motivated behaviour and dopamine release in response to food rewards. Video 1 Download asset This video cannot be played in place because your browser does support HTML5 video. You may still download the video for offline viewing. Download as MPEG-4 Download as WebM Download as Ogg GCaMP6s AgRP neural activity time locked to behaviour at 10 x normal speed in an ad libitum-fed WT mouse previously exposed to a peanut butter pellet. The video shows raw data collected in mV at the photoreceiver prior to df/f calculations. Video 2 Download asset This video cannot be played in place because your browser does support HTML5 video. You may still download the video for offline viewing. Download as MPEG-4 Download as WebM Download as Ogg GCaMP6s AgRP neural activity time locked to behaviour at 10 x normal speed in an ad libitum-fed KO mouse previously exposed to a peanut butter pellet. The video shows raw data collected in mV at the photoreceiver prior to df/f calculations. Metabolic sensing in AgRP neurons affects mesolimbic dopamine pathways Previous studies show that chemogenetic activation of AgRP neurons modulates dopamine release in the NAc to food (Alhadeff et al., 2019). To assess whether impaired metabolic sensing in AgRP neurons affected dopamine release in the NAc, we used GRAB-DA sensors with in vivo fiber photometry (Sun et al., 2018). Acute dopamine release in the NAc was measured in response to a non-food object (wood dowel), chow food and a peanut butter (PB) chip with each presentation separated by 2 min (presented in that order) (Figure 2A). We observed a significant increase in dopamine release in response to chow or a PB chip presentation when compared to wood dowel in both fed and fasted WT mice 0–30 s after placing the food or object into the cage (Figure 2D, E, H1; Videos 3–6). Strikingly, KO mice had significant lower NAc dopamine release between 0 and 30 s, as assessed by average z-score, after a non-food object or PB chip was placed in the cage compared to WT fed mice (Figure 2F). Average z-score was also significantly reduced in fasted KO mice in response to chow and PB chip (0–30 s) when compared to WT fasted mice (Figure 2J). At 30–60 s after food or an object was placed in the cage, the average z-score was only significantly lower in KO in response to PB in the fed mice (Figure 2G and K), suggesting most of the differential effects occur within the first 30 s after presentation. Figure 2 Download asset Open asset Impaired metabolic sensing in AgRP neurons affects dopamine release in the nucleus accumbens. Experimental design to examine dopamine signalling in nucleus accumbens (A). Female (3 WT, 4 KO) and male (4 WT, 4 KO) mice, trained to receive peanut butter chips, were tethered to a fiber optic cable in their home cage. Following acclimation, dopamine responses in the nucleus accumbens to a non-food object (NFO; wood dowel, a, black), standard chow (b, tangerine), and peanut butter chip (PB, c, blue) were measured. Schematic of GRAB-DA (AAV-hSyn-DA4.3) injection site in the nucleus accumbens (B) and fiber placement [scale bar = 500 µm] (C). Average z-scored dopamine release traces aligned (time = 0) to object dropping into cage of WT and KO fed mice (D, E) or WT and KO fasted (H, I) mice. The average z-score of WT and KO ad libitum-fed mice from 0 to 30 s (F) and 30–60 s (G) in response to a non-food object (NFO), standard chow, or peanut butter chip (PB) dropped into cage. The average z-score of WT and KO fasted mice from 0 to 30 s (J) and 30–60 s (K) in response to a non-food object (NFO), standard chow, or peanut butter chip (PB) dropped into cage. Average z-score dopamine release traces aligned to first contact with peanut butter chip (PB) from fed (L) and fasted (M) WT and KO mice, with individual traces shown in a heatmap (N). The average z-score of WT and KO ad libitum-fed mice from –20–0, 0–30, and 30–60 s (O), relative to first contact. The average z-score of WT and KO fasted mice from –20–0, 0–30, and 30–60 s (P), relative to first contact. Time spent eating during the recording session (Q) and time passed between drop and contact with peanut butter (P). Data± SEM, two-way ANOVA with Tukey’s post hoc analysis (F, G, J, K, O, P, Q – main effect of fasting, p < 0.05; R); a – significant compared to WT, p < 0.05. Dashed lines at time = 0 (D, E, H, I) represent the time at which food or an object was dropped into the cage. A continuous line at time = 0 (L, M) the time at which mice contact PB Videos accompany this figure. Figure data 1 Impaired metabolic sensing in AgRP neurons affects dopamine release in the nucleus accumbens. Download Video Download asset This video cannot be played in place because your browser does support HTML5 video. You may still download the video for offline viewing. Download as MPEG-4 Download as WebM Download as Ogg GRAB-DA activity time locked to behaviour at 10 x normal speed in an ad libitum-fed WT The mouse is first exposed to wood dowel, by chow, by a PB The video shows raw data collected in mV at the photoreceiver prior to df/f calculations. Video 4 Download asset This video cannot be played in place because your browser does support HTML5 video. You may still download the video for offline viewing. Download as MPEG-4 Download as WebM Download as Ogg GRAB-DA activity time locked to behaviour at 10 x normal speed in a fasted WT The mouse is first exposed to wood dowel, by chow, by a PB The video shows raw data collected in mV at the photoreceiver prior to df/f calculations. Video 5 Download asset This video cannot be played in place because your browser does support HTML5 video. You may still download the video for offline viewing. Download as MPEG-4 Download as WebM Download as Ogg GRAB-DA activity time locked to behaviour at 10 x normal speed in an ad libitum-fed KO The mouse is first exposed to wood dowel, by chow, by a PB The video shows raw data collected in mV at the photoreceiver prior to df/f calculations. Video Download asset This video cannot be played in place because your browser does support HTML5 video. You may still download the video for offline viewing. Download as MPEG-4 Download as WebM Download as Ogg GRAB-DA activity time locked to behaviour at 10 x normal speed in a fasted KO The mouse is first exposed to wood dowel, by chow, by a PB The video shows raw data collected in mV at the photoreceiver prior to df/f calculations. we aligned the dopamine to first contact with the PB chip (Figure and M) and showed that was a significantly reduced dopamine response 0–30 s after contact with PB in KO compared to WT mice in both fed and fasted (Figure and differences in dopamine release were observed in the s prior to first contact or from 30 to s after contact (Figure and Taken together, these results highlight that impaired metabolic sensing in AgRP neurons affects acute dopamine release in the NAc in both fed and fasted in response to chow and a peanut butter Time spent eating or time to first contact were not significantly different in WT and KO mice in fed or fasted state showing that these could not for genotype differences in dopamine release in the NAc, although was an expected main effect for fasting to increase time spent eating compared to fed mice, of genotype (Figure and Metabolic sensing in AgRP neurons affects motivated behaviour operant are in operant and a calorie to learning the operant This the that food appetitive which is a of AgRP neuronal activity (Krashes et al., 2011; Chen et al., 2016). To this potential we placed the Experimental et al., in home to learning in a low environment for periods of As operant learning can be quickly and in ad libitum-fed mice without the for caloric – an for experiments in which differential feeding responses to fasting, as previously (Reichenbach et al., et al., may task and performance (Figure With this were no differences in operant responding during sessions (Figure indicating no differences in learning the operant To sessions were overnight with and without to in fasted mice fasting at the same time were placed in home (Figure This was to the motivational response to an increasing energy which requires an of homeostatic state over in response to an energy deficit after a fed WT and KO mice a similar (Figure and total number of collected (Figure however, during a fasted KO mice showed a reduced and total consumption compared to WT mice (Figure and (Figure but not (Figure during the fasted session were significantly lower in KO mice compared to WT, as well as during the session (Figure Intriguingly, active and collected only hr after the of fasting and the similar to when food intake after fasting as previously described (Reichenbach et al., et al., This the at which impaired metabolic sensing in AgRP neurons can no homeostatic state and energy Collectively, these results suggest that the appropriate of homeostatic state metabolic sensing in AgRP neurons affects motivation for a palatable food reward. Figure Download asset Open asset Impaired metabolic sensing in AgRP neurons affects motivation for sucrose rewards during fasting. WT, KO) were trained to for sucrose rewards using to on average correct for with or without chow For under fasting conditions, mice were not fasted prior to but fasted from the the of the session This the motivation to the increasing energy deficit in response to an energy from fed and fasted mice and at the of sessions The number of active (G) and collected during fasted sessions over Data± SEM, two-way repeated ANOVA with Tukey’s post hoc analysis (F, – main effect of Data± SEM, two-way ANOVA with Tukey’s post hoc analysis (D, a – significant compared to WT, p < ** p < 0.01, *** p < Figure data 1 Impaired metabolic sensing in AgRP neurons affects motivation for sucrose rewards during fasting. Download To that in motivation were associated with impaired dopamine release in the NAc, we used as it for to and with of dopamine release by GRAB-DA photometry (Figure release was measured during a of dopamine release to or In fed and fasted WT mice, a significantly increased NAc dopamine as assessed by average z-score s after a compared to 5 s prior to (Figure H, and However, a in fed or fasted KO mice did not significantly increase NAc dopamine release within s of the compared to 5 s prior to (Figure and In NAc dopamine release was significantly reduced during the after a in KO compared to WT mice in both fed and fasted (Figure and The time between was lower in fasted mice compared to fed mice Figure but no genotype differences were observed. Figure 4 Download asset Open asset Impaired metabolic sensing in AgRP neurons reduces dopamine release in the nucleus accumbens during a Experimental design mice with in the nucleus accumbens were trained to for a sucrose prior to experimental mice were tethered to a cable in their home cage and after 5 min acclimation, mice to on Schematic of GRAB-DA (AAV-hSyn-DA4.3) injection site in the nucleus accumbens (B) and fiber placement [scale bar = 500 µm] (C). average dopamine release response aligned (time = 0) to and correct during a session from WT n = 10 and KO n = 12 ad libitum-fed mice and WT n = 12 or KO n = 12 after an overnight The average z-score of from WT and KO ad
Source: eLife
Authors: Alex Reichenbach, Rachel E Clarke, Romana Stark, Sarah Haas Lockie, Mathieu Mequinion, Harry Dempsey, Sasha Rawlinson, Felicia Reed, Tara Sepehrizadeh, Michael DeVeer, Astrid C Munder, Juan Nunez-Iglesias, David C Spanswick, Randall Mynatt, Alexxai V Kravitz, Christopher V Dayas, Robyn Brown, Zane B Andrews
Institutions: Monash University, Florey Institute of Neuroscience and Mental Health, University of Warwick, Pennington Biomedical Research Center, Louisiana State University System, Washington University in St. Louis, University of Newcastle Australia, Newcastle University, The University of Melbourne
Language: English
Subfield: Endocrine and Autonomic Systems
Field: Neuroscience
Domain: Life Sciences
Open Access status: gold
Estimated APC paid: $2,500
OpenAlex ID: W4210788614