Metabolic sensing in AgRP regulates sucrose preference and dopamine release in the nucleus accumbens

Hunger increases the motivation for calorie consumption, often at the expense of low‐taste appeal. However, the neural mechanisms integrating calorie‐sensing with increased motivation for calorie consumption remain unknown. Agouti‐related peptide (AgRP) neurons in the arcuate nucleus of the hypothalamus sense hunger, and the ingestion of caloric solutions promotes dopamine release in the absence of sweet taste perception. Therefore, we hypothesised that metabolic‐sensing of hunger by AgRP neurons would be essential to promote dopamine release in the nucleus accumbens in response to caloric, but not non‐caloric solutions. Moreover, we examined whether metabolic sensing in AgRP neurons affected taste preference for bitter solutions under conditions of energy need. Here we show that impaired metabolic sensing in AgRP neurons attenuated nucleus accumbens dopamine release in response to sucrose, but not saccharin, consumption. Furthermore, metabolic sensing in AgRP neurons was essential to distinguish nucleus accumbens dopamine response to sucrose consumption when compared with saccharin. Under conditions of hunger, metabolic sensing in AgRP neurons increased the preference for sucrose solutions laced with the bitter tastant, quinine, to ensure calorie consumption, whereas mice with impaired metabolic sensing in AgRP neurons maintained a strong aversion to sucrose/quinine solutions despite ongoing hunger. In conclusion, we demonstrate normal metabolic sensing in AgRP neurons drives the preference for calorie consumption, primarily when needed, by engaging dopamine release in the nucleus accumbens.


| INTRODUCTION
Hunger has a profound effect on enacting behaviours associated with food seeking and consumption.A core feature of this hunger-driven behavioural repertoire is the motivation to continue food-seeking until energy requirements are met.Hypothalamic Agouti-related peptide (AgRP) neurons, located in the arcuate nucleus (ARC), are key hunger-sensing neurons in the brain, [1][2][3] whereas dopamine neurons in the ventral tegmental area (VTA) play a critical role in motivation.
Indeed, there are many parallels between AgRP and dopamine neurons with respect to food seeking and motivation.For example, dopamine-deficient and AgRP-ablated mice are both hypophagic and prone to starvation 1,4 and the activation of AgRP or dopamine neurons increases motivated food seeking. 2,3,5nger increases dopamine release in the nucleus accumbens in response to food or cues predicting food [6][7][8] highlighting the interaction between homeostatic and motivational systems.Chemogenetic activation of AgRP neurons increases nucleus accumbens dopamine release and VTA dopamine neural activity to food, 8,9 demonstrating that AgRP neurons relay information pertaining to homeostatic state to motivation pathways.Indeed, impaired metabolic sensing in AgRP neurons weakens the ability to transmit homeostatic information between AgRP neurons and dopamine pathways. 10This impaired metabolic sensing in AgRP neurons not only reduced dopamine release to sucrose pellets, it is also suppressed motivated food seeking during fasting.
Hunger not only increases motivation for food, but it also modifies liquid taste preference to maintain calorie intake.For example, fasted mice will consume more sucrose solution at a lower concentration or consume more sucrose solution adulterated with a bitter tastant, 11 and this effect was replicated by the chemogenetic or optogenetic activation of AgRP neurons. 11Moreover, cues predicting sucrose consumption produce a greater phasic release of dopamine compared to cues predicting saccharin, 12 and mice lacking sweet taste receptors prefer sucrose over sucralose. 13These results imply that AgRP neurons promote the consumption of caloric solutions, which are inherently rewarding and elicit a dopamine response independent from taste.In our current studies, we examined whether metabolic sensing in AgRP neurons distinguished caloric from non-caloric solutions by regulating dopamine release in the nucleus accumbens.Moreover, we examined whether metabolic sensing in AgRP neurons affected taste preference for bitter solutions under conditions of energy need.To model impaired metabolic sensing in AgRP neurons we deleted Crat from AgRP neurons as our previous studies demonstrated aberrant glucose sensing using ex vivo electrophysiological recordings or in vivo AgRP GCaMP6 recordings. 10[16] 2 | RESULTS

| Metabolic sensing in AgRP neurons facilitates dopamine release to sucrose intake
To test if metabolic sensing in AgRP neurons helps to distinguish caloric sucrose solutions from sweet but non-caloric saccharin solutions by regulating dopamine release, we exposed AgRP Crat wild-type (WT) or knockout (KO) mice to 4% sucrose and 0.1% saccharin for 30 min while recording nucleus accumbens dopamine release using GRAB-DA 17 (Figure 1A-C).All mice were naïve to caloric and noncaloric sweetened solutions to ensure an untrained dopamine response.In response to sucrose consumption, AgRP Crat KO mice had significantly lower dopamine release in the nucleus accumbens compared to WT mice (Figure 1D,E).In response to saccharin consumption, however, the increase in dopamine release was not different between genotypes (Figure 1F,G).Furthermore, when comparing the effect of sucrose and saccharin on dopamine release in the nucleus accumbens (Figure 1H-K), we observed greater persistence of the dopamine signal after sucrose consumption compared to saccharin in WT mice (Figure 1H,I), but not KO mice (Figure 1J,K).Thus, metabolic sensing in AgRP neurons was necessary for the appropriate dopamine response to sweet caloric solutions but not sweet noncaloric solutions, indicating that AgRP neurons transmit caloric information into a persistent dopamine signal within the nucleus accumbens.
Based on persistent dopamine release in response to a caloric sucrose solution in WT mice, we hypothesised that WT and KO mice would show significant differences in sucrose preference under fed and fasted conditions.To test this idea, mice were placed in BioDaq cages to allow for precise measurement of fluid intake over time (Figure 2A).Initial two-bottle comparison between water and saccharin (Figure 2B,C) revealed a clear preference for the sweet taste of saccharin for both genotypes, indicating impaired metabolic sensing in AgRP neurons does not affect taste preference for sweet solutions, an observation that is supported by similar dopamine responses to saccharin in WT and KO mice (Figure 1F).Given the choice between 4% sucrose and 0.1% saccharin, mice consume more sucrose than saccharin irrespective of genotype (Figure 2D,E) and the sweet taste of solutions increases the fluid intake compared to the average water intake (Figure 2I).However, the increased need for calories during fasting elevates the preference for sucrose more in WT than in KO (Figure 2F-H).Thus, metabolic sensing in AgRP neurons potentiates the shift towards sucrose preference during fasting to restore energy balance.
Bitter tastes such as quinine or denatonium are often used as adulterants to create taste aversions and suppress intake.However, both fasting and AgRP neuronal stimulation drive greater consumption of sucrose despite the presence of bitter tastes, 11 implying AgRP neurons monitor energy state and prioritise caloric intake over taste.To test whether metabolic sensing in AgRP neurons is important in this process, we performed sucrose/quinine-saccharin preference tests (Figure 3A).Pairing the caloric solution with an unpalatable taste shifts the preference to saccharin in KO, with WT mice showing no preference (Figure 3B,C).Strikingly, an overnight fast increases the preference for the quinine-laced sucrose solution in WT, but not in KO mice (Figure 3D,E), resulting in a significant increase in preference from fed to fasted in WT mice but no change in preference in KO mice (Figure 3F).Thus, AgRP neurons require intact metabolic-sensing capabilities during fasting to increase caloric consumption by increasing sucrose preference despite the bitter quinine taste.
The dramatic increase in saccharin preference over sucrose/ quinine solutions in KO mice highlights potentially greater sensitivity to bitter tastes or stronger taste avoidance.To examine this further, we examined 4% sucrose consumption laced with 0, 0.1, 0.2, 0.5 mM quinine in fed and fasted mice.Increasing the quinine concentrations in the sucrose solution gradually reduced the volume consumed with 0.5 mM quinine being too unpalatable for mice to drink (Figure 3G-I).
In fed mice, lacing sucrose with 0.1 mM quinine did not reduce the volume consumed in WT compared to 4% sucrose but it significantly reduced the consumption in KO mice so that they consumed significantly less compared to WT. Importantly, inducing a metabolic need state by removing food during the 16 h access period increased consumption in WT but not KO mice, resulting in a significant main effect of genotype for overall consumption (Figure 3H).This suggests that metabolic sensing in AgRP neurons balances the caloric value of sucrose with the tolerance for aversive taste.Collectively, these results show that metabolic sensing in AgRP neurons during fasting increases caloric sucrose intake by increasing the tolerance for quinine, which is used to devalue the taste of sucrose.
F I G U R E 1 Impaired metabolic sensing in AgRP neurons affects NAc dopamine response to sucrose but not saccharin solutions.Naïve WT and AgRP Crat KO mice expressing GRAB DA sensor in the nucleus accumbens were exposed to 4% sucrose and 0.1% saccharin solutions and the dopamine response during lick bouts were recorded (A, B).A 30-min representative recording highlighting NAc dopamine release in response to sucrose (dark blue) and saccharin (light blue) licking events (C).Dopamine response to sucrose (D) and saccharin bouts (F) and average z-score during response (E, G).Comparison of dopamine release during sucrose and saccharin bouts and average z-score of early (0-5 s) and late (5-10 s) response in WT (H, I) and KO (J, K).Data ± SEM, two-way ANOVA (see Table S1 for all statistical analysis), significant at p < .05.

| DISCUSSION
In these studies, we assessed how metabolic sensing in AgRP neurons affected sucrose preference and NAc dopamine release.5][16] With this model, we demonstrated that impaired metabolic sensing in AgRP neurons attenuated NAc dopamine release in response to sucrose, but not saccharin, consumption.Furthermore, WT mice displayed a greater NAc dopamine response to sucrose when compared with saccharin, in line with previous reports showing F I G U R E 2 Impaired metabolic sensing in AgRP neurons impedes sucrose preference.In a two-bottle choice paradigm, male and female mice had ad libitum access to chow for 4 days, followed by an overnight fast.The position of bottles was swapped daily, and the volume consumed overnight was recorded.Bottles were washed after each choice pair and mice had 2 days of water in both bottles (A).Pie charts compare preference between genotypes for saccharin (light blue) and comparator solution as indicated (dark blue).After acclimation to the BioDaq cages, mice received a choice of water and 0.1% saccharin (B, C), followed by 4% sucrose and 0.1% saccharin with ad libitum access to chow (D, E) or without access to chow (F, G).Fasting induced change in sucrose preference in WT and KO (H) and average total (saccharin plus comparator solution) fluid consumed per day (I).Data ± SEM, two-way ANOVA (see Table S1 for all statistical analysis), significant at p < .05.
that cues predicting sucrose produce larger increases in dopamine release compared to those predicting saccharin. 12As there were no genotype differences in NAc dopamine release to saccharin, our results suggest metabolic sensing in AgRP neurons is essential to transmit neural information about caloric content, but not sweet taste, to dopamine terminals in the NAc.
Previous studies showed mice lacking sweet taste receptors could still increase striatal dopamine release and form a preference for caloric (sucrose) solutions compared to non-caloric sucralose solutions. 13 suggest metabolic sensing in AgRP neurons may be the essential step to form sucrose preference in the absence of sweet taste perception, as reported. 13Based on this logic, the absence of caloric processing may shift preferences towards sweet solutions independent of calorie content or need.In line with this, the preference for saccharin over sucrose laced with quinine did not change in fed or fasted KO mice.Thus, the inability to discern caloric and non-caloric sweetened solutions meant that choice behaviour was guided mainly by taste.A similar observation was reported after ablating AgRP neurons in neonatal mice, as feeding relied more on palatability and dopamine tone. 18 AgRP neurons are critical for sensing hunger, our results suggest the effects of hunger on NAc dopamine release 7 in response to food 8 or cues predicting food 6 are driven by AgRP neurons.This assumption is further supported by studies showing the activation of AgRP neurons increases NAc dopamine release, 8,9 a response that is required for AgRP neurons to drive appropriate food motivation during fasting. 10Whether or not metabolic sensing in AgRP neurons affects VTA DA cell firing in a similar manner remains to be determined.In support of this view, chemogenetic activation of ARC NPY (AgRP) neurons elevates VTA dopamine activity to food presentation. 8However, NAc dopamine release can occur independent from changes from VTA neural firing, a process thought to dissociate roles for dopamine in learning and motivation. 19Intriguingly, chemogenetic activation of AgRP neurons drives NAc dopamine release, 8,9 but reduces the number of active VTA dopamine neurons following intragastric Ensure infusion. 20Thus, it is possible that these opposing AgRP-driven changes at the NAc nerve terminal and VTA cell bodies mediate different behaviours, although this requires further experimental evidence.
Our results show that impaired metabolic sensing in AgRP neurons diminished sucrose preference, although we cannot completely exclude the possibility that a part of the KO phenotype is a consequence of Crat deletion in AgRP neurons without affecting metabolic sensing.However, arguing against this possibility is the recent observation that pharmacological manipulation of glucose availability affects dopamine neuronal activity to sucrose or sucrose-predicting cues. 21In addition, our results are consistent with chemogenetic studies showing that AgRP activation and inhibition increased and decreased sucrose-licking activity, respectively. 11The inability of KO mice to increase their preference for sucrose in the fasted state, as occurs in WT mice, highlights the key role of AgRP neurons to guide behaviour based on appropriate metabolic sensing of energy needs.3][24] Moreover, our previous studies demonstrate that appropriate metabolic sensing in AgRP neurons is required for the accurate encoding of caloric information in response to glucose and peanut butter. 10RP neurons, however, also affect taste preference since the activation of AgRP neurons increases the licking rate, whereas AgRP inhibition decreases the licking rate for sucrose laced with a bitter taste.11 Our results showed a similar response in which fasting increased consumption of sucrose laced with quinine in WT but not KO mice.Indeed, KO mice showed a strong preference for saccharin, which was unaffected by fasting.These results suggest that taste preference is mediated in part by the ability of hunger-sensing AgRP neurons to sense energy needs, as observed in WT mice, where caloric need caused by fasting increased sucrose/quinine consumption compared to fed mice.Although the influence of hunger on taste sensitivity has been well documented, 25,26 our results establish a key role for metabolic sensing in AgRP neurons.Furthermore, saccharin-induced NAc dopamine release is similar in WT and KO mice, suggesting KO mice have a stronger taste aversion for sucrose/quinine consumption.Thus, impaired metabolic sensing in AgRP neurons may affect taste aversion, although future studies are required to explore this possibility.Although a dopamine signal of aversive taste occurs separate from gut-brain processing of caloric value, 27 both require appropriate subsequent and potentially overlapping neural integration. Infact, gut-brain feedback carries both post-ingestive caloric information 22,23 and malaise signals 28 and is one potential way that AgRP neurons could integrate caloric information with taste aversion.These post-ingestive malaise signals target regions with AgRP neural terminals, including the parabrachial nucleus and amygdala.29 Moreover, AgRP neurons dynamically encode the caloric value of nutritive signals in learned flavour nutrient pairings 22,23,30 making them perfect candidates to mediate post-ingestive taste-nutrient associations.
In summary, metabolic sensing in AgRP is required to correctly transmit and transform AgRP hunger signalling into NAc dopamine release after calorie consumption.Importantly, metabolic sensing in F I G U R E 3 Metabolic sensing in AgRP neurons enables the choice of calories over taste during fasting.Same two-bottle choice paradigm as in Figure 2 but with 0.1% saccharin and 4% sucrose laced with 0.1 mM quinine (A).Fluids consumed (B) and preference (C) during ad libitum access to chow and during fasting (D, E).Fasting-induced changes in sucrose/quinine preference in WT and KO (F).Dose-response curve for quinine concentrations in 4% sucrose solution from different cohorts of mice (n = 5-14) ad libitum fed (G) and without access to chow (H) and combined saccharin and sucrose/quinine average volume consumed (I).Data ± SEM, two-way ANOVA (see Table S1 for all statistical analysis), significant at p < .05.
AgRP neurons does not affect NAc dopamine release after sweet non-caloric saccharin consumption, highlighting different mechanisms used to distinguish calorie need from sweet taste.Furthermore, metabolic sensing in AgRP neurons adjusts the preference to caloriecontaining solutions under conditions of energy need, even at the expense of aversive bitter tastants.Overall, these studies demonstrate that hunger-sensing AgRP neurons drive the preference for calorie consumption, primarily when needed, by engaging dopamine release in NAc.The abnormal function of this pathway may contribute to impaired feeding behaviour and lead to either under or over-eating, which characterise human conditions such as eating disorders and obesity.

| Animals
Male and female mice were kept under standard laboratory conditions with free access to food (chow diet, catalogue no.8720610; Barastoc Stockfeeds, Victoria, Australia) and water at 23 C in a 12 h light/ dark cycle and were group-housed to prevent isolation stress unless otherwise stated.All mice were aged 8 weeks or older for experiments unless otherwise stated.Agrp-ires-cre mice were obtained from Jackson Laboratory AgRP tm1(cre)Low/J (stock no.012899; The Jackson Laboratory, Maine, USA).To delete Crat from AgRP neurons, Agrp-irescre mice were crossed with Crat fl/fl mice (Randall Mynatt, Pennington Biomedical Research Center, LA, USA) (AgRP cre/wt ::Crat fl/fl mice; designated as KO).AgRP wt/wt ::Crat fl/fl littermate mice were designated as WT and used as controls.All experiments were conducted in compliance with the Monash University Animal Ethics Committee guidelines.

| Two-bottle choice tests
Male and female mice were single-housed in BioDaq feeding cages with ad libitum access to chow and two water bottles during 3 days of acclimatization.For water versus saccharin consumption, one bottle was filled with 0.1% saccharin and the position of bottles was swapped daily for 2 h before the onset of the dark phase.After 4 days ad libitum feeding, mice were overnight fasted (5 PM-9 AM) with only access to drink bottles, followed by 2 days wash out period with ad lib access to water and chow.In the same manner as above, mice were presented with a choice of 4% sucrose and 0.1% saccharin solutions or 4% sucrose laced with quinine HCl (concentrations for different cohorts ranging from 0.1 to 0.5 mM) and 0.1% saccharin to measure consumption in fed and fasted states.

| Statistical analysis
Statistical analyses were performed using GraphPad Prism for MacOS X.Data are represented as mean ± SEM.Two-way ANOVAs with post hoc tests were used to determine statistical significance (all statistical information is supplied in a detailed Table S1).p < .05 was considered statistically significant.

Fibre
photometry processor RZ5P.Mice for fibre photometry experiments Reichenbach: Conceptualization; data curation; formal analysis; investigation; methodology; writingoriginal draft.Harry Dempsey: writingoriginal draft; writingreview and editing.ACKNOWLEDGMENTOpen access publishing facilitated by Monash University, as part of the Wiley -Monash University agreement via the Council of Australian University Librarians.