Sustained NPY signaling enables AgRP neurons to drive feeding

1AgRP neurons are a small population of cells in the arcuate nucleus of the hypothalamus that are critical for regulating food intake. These cells are progressively activated during food deprivation, in part due to fluctuations in nutritionally-regulated hormones such as leptin and ghrelin. Optogenetic or chemogenetic stimulation of AgRP neurons promotes intense food seeking and consumption, whereas silencing of these cells leads to aphagia or starvation. Thus AgRP neurons play a key role in the homeostatic negative feedback loop that links the need for energy to the stimulation of feeding behavior.

2Although AgRP neurons are gradually activated during fasting, they are inhibited within seconds when a hungry animal sees or smells food. Paradoxically, this rapid inhibition of AgRP neurons often occurs before a single bite of food has been consumed and then persists for the duration of the ensuing meal. These observations raise the question of how AgRP neuron activity is able to drive food consumption, given that these neurons are inhibited during the act of feeding itself.

3To reconcile these observations, we hypothesized that AgRP neurons may transmit a long-lasting signal that continues to drive food consumption even after AgRP neurons have been inhibited by food cues. Consistent with this, we showed that brief optogenetic stimulation of AgRP neurons in the absence of food can drive voracious food intake tens of minutes later, long after AgRP neuron stimulation has been terminated. This indicates that AgRP neurons can generate an unusual ‘sustained hunger drive’ that outlasts their acute firing (Figure 1A). The mechanisms underlying this long-lasting potentiation of feeding behavior are unknown.

4AgRP neurons release three neurotransmitters that can influence feeding: AgRP, NPY and GABA. We hypothesized that one of these neurotransmitters may play a specific role in the sustained hunger drive generated by AgRP neurons. To test this hypothesis, we blocked signaling through each of these pathways and then tested their role in AgRP neuron driven food intake (Figure 1B). To inactivate GABA signaling, we crossed AgRP-IRES-Cre mice to animals with a floxed allele of the GABA transporter Slc32a1 (referred to as GABA-); to block NPY signaling we used NPY knockout mice (NPY-); and for AgRP we used Agouti mice (AgRP-), in which the melanocortin pathway is inactivated by constitutive expression of a melanocortin 4-receptor antagonist. We then measured the effect of optogenetic pre-stimulation of AgRP neurons in each of these strains, or littermate controls.

5We first stimulated AgRP neurons for 15 min in the absence of food, and then shut off the laser and allowed animals to eat for one hour. This pre-stimulation paradigm produced robust food intake in mice lacking GABA or AgRP signaling, and the magnitude of food consumption was similar between mutants and littermate controls (Figure 1C). In striking contrast, pre-stimulation of mice lacking NPY caused no increase in food intake (Figure 1C, bottom row). To measure the quantitative requirement for these neurotransmitters, we repeated this experiment with different durations of pre-stimulation (5, 15, 30 and 60 min; Figure 1D). We found that control mice, as well as mice lacking GABA or AgRP, showed a clear dose-response relationship between pre-stimulation duration and total food intake (Figure 1E,F and Figure 1—figure supplement 1). This increase in food intake was due to an increase in bout size, rather than bout number (Figure 1—figure supplement 2). In contrast NPY- mice showed no dose-dependent increase in food intake following longer durations of AgRP neuron pre-stimulation (Figure 1G). This behavioral defect was not due to a failure to activate AgRP neurons, because we observed robust Fos expression in AgRP neurons after optogenetic stimulation of NPY- mice (Figure 1L). Instead this indicates that NPY is specifically required for the sustained hunger produced by AgRP neuron activation.

6To examine the kinetics of AgRP neuron driven hunger, we pre-stimulated AgRP neurons for five minutes and then inserted a delay of varying lengths between the termination of stimulation and the availability of food (15, 30, or 60 min; Figure 1H). We found that all genotypes except for NPY- mice exhibited robust food consumption that gradually declined with increasing delay duration up to 60 min (Figure 1I–K). By pooling data from all control animals, we estimated that this hunger signal has a rise time of approximately four minutes (logarithmic growth; N = 23; R-Square = 0.72; Figure 1M) and a decay-time of approximately 47 min (exponential decay; N = 19; R-Square = 0.87; Figure 1N). This indicates that AgRP neuron activity rapidly generates a durable, NPY-dependent signal that drives feeding behavior.

7AgRP neurons promote not only food consumption but also behaviors that lead to food discovery. We therefore examined the requirement for each of these neurotransmitters in AgRP driven appetitive behavior. To do this we used a progressive ratio three lever-pressing task, in which delivery of each subsequent pellet of food requires three additional lerel presses (Figure 2A). We observed vigorous lever pressing following pre-stimulation of AgRP neurons in AgRP- and GABA- mice, as well as controls (Figure 2B–F). In contrast pre-stimulation had no effect on operant responding in animals lacking NPY (Figure 2B,C,F). This indicates that NPY is critical for the ability of AgRP neurons to drive food seeking behavior as well as food consumption.

8To investigate with greater temporal resolution the requirement for NPY in AgRP driven feeding behavior, we developed a paradigm that alternates between concurrent and pre-stimulation. In this paradigm, animals are given constant access to food, and then subjected to two minutes of stimulation (20 Hz; Laser ON) followed by two minutes of no stimulation (Laser OFF) in a pattern that repeats for two hours (Figure 3A). In control mice, this protocol stimulated robust food consumption (Figure 3B,C). Importantly, food intake was equally distributed between the Laser ON and OFF epochs (Figure 3C–I), indicating that wild-type animals do not distinguish the boundary between these two-minute intervals, presumbly due to the sustained hunger produced by AgRP neuron stimulation.

9We then repeated this experiment in each of the genotypes described above. Remarkably, loss of NPY caused food intake to become strictly time-locked to the laser stimulus: that is, NPY- animals ate during concurrent stimulation, but then stopped eating within seconds of the laser shutting off (Figure 3C,D,F,I). In contrast, mice lacking GABA or AgRP signaling, as well as their littermate controls, showed no difference in feeding between the Laser ON and OFF periods (Figure 3E–H). These results were robust to variations in the paradigm, such as changing the duration of the stimulation epoch (Figure 3G–I). This indicates that, in the absence of NPY, AgRP neuron driven hunger dissipates within seconds of the offset of AgRP neuron firing.

10NPY is expressed in cell types other than AgRP neurons. This raises the possibility that global NPY knockout could cause changes in other circuits that indirectly lead to deficits in AgRP neuron driven feeding. To exclude this possibility, we used a viral approach to rescue NPY expression selectively in AgRP neurons of NPY- mice. We generated two Cre-dependent AAVs, one expressing NPY and the other expressing GFP as a control, and then injected these AAVs into the ARC of NPY- mice (Figure 3J). Mice injected with AAV expressing NPY, but not GFP, showed restoration of NPY expression in the ARC (Figure 3K) and complete rescue of the feeding phenotypes associated with NPY- knockout (Figure 3L–P). For example, NPY re-expression in AgRP neurons caused feeding behavior in NPY- mice to become uncoupled from the onset and offset of laser stimulation, in a way that was indistinguishable from wild-type animals (Figure 3N–P). This indicates that the phenotypes observed in NPY- mice are due to loss of NPY expression in AgRP neurons, rather than effects of this mutation on other circuits or development.

11We have focused here on the behavioral effects of AgRP neuron pre-stimulation. While this paradigm cannot replicate the precise details of asynchronous natural activity, it does mimic the broad features of the natural regulation of these cells (i.e. that AgRP neurons are more active before food discovery than during food consumption). In contrast, many earlier studies used a protocol in which AgRP neurons are stimulated concurrently with food consumption. This protocol induces an activity pattern that is the opposite of what occurs naturally (i.e. greater AgRP neuron activity during feeding than before). To place our results in the context of this earlier work, we concurrently stimulated AgRP neurons using either chemogenetics (Figure 4A–C) or optogenetics (Figure 4E–H) and then measured food intake in NPY+ and NPY- mice. We found that concurrent chemogenetic stimulation increased food consumption in mice lacking NPY (Figure 4C,D and Figure 4—figure supplement 1), consistent with a previous report. In contrast, during concurrent optogenetic stimulation NPY- mice consumed only a small fraction of the food consumed by NPY+ controls (Figure 4E,F). We could partially rescue food intake in NPY- mice by using a higher intensity optogenetic stimulation protocol (Figure 4F), whereas NPY+ controls showed no difference in food consumption between high and moderate intensity stimulation (high intensity refers to continuous 20 Hz stimulation, whereas moderate intensity refers to 20 Hz stimulation in a 2 s ON, 3 s OFF pattern; Figure 4G). This indicates that intense concurrent optogenetic stimulation of AgRP neurons can partially overcome the requirement for NPY, likely through increased release of GABA. It remains unclear why concurrent chemogenetic stimulation of AgRP neurons does not also require NPY to induce feeding (Figure 4C,D and ). In contrast to these variable responses to concurrent stimulation, we found that increased stimulation intensity could not compensate to any degree for loss of NPY in a pre-stimulation paradigm (Figure 4H). This is presumably because endogenous GABA and AgRP are unable to drive food intake effectively on the timescale of sustained hunger (tens of minutes immediately after the offset of stimulation).

12NPY was implicated in the regulation of feeding decades ago, but its precise contribution has been difficult to establish. Early studies showed that infusion of picomoles of NPY into the PVH can elicit voracious food intake, identifying NPY as an unusually potent orexigen. Consistent with this, NPY is highly expressed in hunger-promoting AgRP neurons, and its expression in these cells is increased by energy deficit. However genetic deletion of NPY, alone or in combination with AgRP, had little effect on food intake or body weight. Similarly deletion of NPY receptors, alone or in combination, failed to produce the expected lean phenotype. More recently, the advent of cell-type-specific tools for neural manipulation made it possible to acutely stimulate AgRP neurons and then measure the effect of blocking NPY signaling. One study found that food intake following chemogenetic stimulation of AgRP neurons requires either NPY or GABA signaling, but not both. Another study showed that antagonists of NPY or GABA receptors in PVH could partially block feeding induced by optogenetic stimulation of AgRP neurons. Taken together, these observations have led to a general view that NPY and GABA have partially redundant functions in mediating food intake by AgRP neurons.

13The discovery that AgRP neurons are rapidly silenced by the sensory detection of food, and consequently can drive feeding via a long-lasting hunger signal, led us to wonder whether a specific neurotransmitter might mediate this unusual mechanism for controlling behavior. We therefore reinvestigated the role of these neurotransmitters using an optogenetic pre-stimulation paradigm that mimics the broad features of the natural regulation of AgRP neurons. This revealed a remarkably specific role for NPY in extending the duration of AgRP neurons’ behavioral effects for tens of minutes beyond the window of their immediate firing. This in turn provides a molecular correlate for the long-lasting hunger signal that bridges AgRP neuron dynamics with feeding behavior.

14While our findings have emphasized the role of NPY in sustained hunger, previous work has shown that the neuropeptide AgRP can also drive food intake over long timescales. For example, infusion of AgRP into the brain or stimulation of AgRP release can cause modulation of feeding behavior that persists for days. However, this long-lasting potentiation of feeding behavior requires several hours to emerge when endogenous AgRP release is stimulated. Consistent with this, we find no evidence that the AgRP peptide is required for the intense food consumption that occurs immediately following AgRP neuron pre-stimulation (Figure 1). Thus our data suggest that NPY and AgRP play complementary roles in the regulation of feeding, with NPY sustaining hunger selectively on the timescale of a meal (tens of minutes after food discovery), whereas AgRP modulates the tone of the feeding circuit over longer timescales. This role for AgRP is consistent with an extensive literature that has implicated the melanocortin system in the long-term regulation of energy balance.

15A number of questions remain unanswered about how NPY mediates the sustained hunger response described here. Slice recordings have shown that NPY can have persistent effects on membrane excitability and neurotransmitter release in vitro, indicating that NPY may act directly on post-synaptic targets of AgRP neurons to cause durable changes in their activity. We have shown that pre-stimulation of AgRP neuron axon terminals in PVH, BNST, and LH is sufficient to drive sustained feeding, suggesting that these are the downstream sites where NPY has its long-lasting effects on hunger. However it remains unknown which NPY receptors are involved, in which cell types these receptors are expressed, which molecular mechanisms they utilize, and how this sustained modulation is represented in the dynamics of the next circuit node. Likewise, while we have focused on NPY’s role in appetitive and consummatory aspects of feeding, it remains unknown whether NPY has similar long-lasting effects on metabolism or other adaptive responses to energy state.

16A central challenge of neuroscience is to explain how the neurochemical organization of the brain gives rise to behavior. Neural circuits contain an extraordinary diversity of neuropeptides, yet in most cases we have little understanding of how those peptides relate to the function of the underlying cells. We have shown here how a neuropeptide that drives feeding can function on a timescale that is inaccessible to classical neurotransmitters. This in turn allows a neuron that controls hunger to direct ongoing behavior while simultaneously anticipating and responding to impending physiologic changes. It will be important to investigate to what extent other neuropeptides influence behavior in this way.

17Mice were group housed on a 12:12 light:dark cycle with ad libitum access to water and mouse chow (PicoLab Rodent Diet 20, 5053 tablet, TestDiet). Adult mice (8–20 weeks old) were used for experiments. For channelrhodopsin-2 expression in AgRP neurons, Agrp-IRES-Cre mice (Jackson Labs stock 012899, B6; Agrptm1(cre)Lowl/J; RRID:IMSR_JAX:012899) were crossed with Ai32: ROSA26-loxStoplox-ChR2-eYFP mice (Jackson Labs stock 012569, B6; 129S-Gt(ROSA)26Sortm32(CAG-COP4*H134R/EYFP)Hze/J; RRID:IMSR_JAX:012569) to generate double mutant animals. For hM3Dq expression in AgRP neurons, Agrp-IRES-Cre mice were crossed with CAG-LSL-HA-hM3Dq-pta-mCitrine transgenic mice (Jackson Labs Stock 026220, B6; Tg(CAG-CHRM3,-mCitrine)1Ute/J.; RRID:IMSR_JAX: 026220) to generate double mutant animals. To block melanocortin signaling, mice were bred into heterozygotic Ay background (Jackson Labs stock 000021, B6.Cg-Ay/J; RRID:IMSR_JAX:000021); wild type cohorts were used as controls. To ablate GABA signaling specifically in AgRP neurons, mice were crossed into Agrp-IRES-Cre;Vgatflox/flox (Jackson Labs stock 012897, Slc32a1tm1Lowl/J; RRID:IMSR_JAX:012897) background; Agrp-IRES-Cre;Vgatflox/+ animals were used as controls. To block NPY signaling, mice were crossed into NPY-/- background (Jackson Labs stock 004545, 129S-Npytm1Rpa/J; RRID:IMSR_JAX:004545); NPY-/+ cohorts were used as controls.

18Experimental protocols were approved by the University of California, San Francisco IACUC (Protocol AN133011) following the NIH guidelines for the Care and Use of Laboratory Animals.

19We performed intracranial surgery as previously described. For re-expression of NPY, pAAV-EF1a-DIO-NPY were generated from pAAV-EF1a-DIO-EYFP. Both vectors were made into AAV8 in Stanford Vector Core. 200 nL of AAV8-EF1a-DIO-NPY or AAV8-EF1a-DIO-EYFP was bilaterally injected into the ARC (bregma: AP: −1.75 mm, ML: ±0.3 mm, DV: −5.9 mm). For optogenetic implants, custom-made fiberoptic implants (0.39 NA Ø200 µm core Thorlabs FT200UMT and CFLC230-10) were installed above the ARC (bregma: AP: −1.75 mm, DV: dorsal surface −5.6 mm, ML: −0.25 mm).

20To generate light pulses, a 473 nm laser was modulated by Coulbourn Graphic State software through a TTL signal generator (Coulbourn H03-14) and synchronized with behavioral experiments for each subject. The laser was passed through a single patch cable (Doric Lenses) to a custom-made fiber optic patch cable (Thorlabs FT200UMT, CFLC230-10; Fiber Instrument Sales F12774) through a rotary joint (Doric Lens FRJ 1 × 1). Patch cables were connected to the implant on each mouse through a zirconia mating sleeve (Thorlabs ADAL1). Laser power was set to 12–18 mW at the terminal of each patch cable unless otherwise specified.

21Mice were allowed to recover for seven days after implant surgery before experiments. In addition to regular chow, mice were supplied ad libitum with the food pellets used during testing (20 mg Bio-Serv F0163) in their home cage. Mice were habituated to the behavioral chambers (Coulbourn H10-11M-TC with H10-11M-TC-NSF) and pellet dispensing systems (Coulbourn H14-01M-SP04 and H14-23M) for three days prior to the first experiment. Mice were provided ad libitum access to food and water unless otherwise specified and were tested during the early phase of the light cycle. All pre-stimulation food intake experiments followed this general structure: 70 min habituation/pre-stim period with no food access followed by 60 min of food access. Pellet removal was detected using a built-in photo-sensor (Coulbourn H20-93).

22The following protocols were performed back to back to test the dose dependence of pre-stimulation on food intake. The sequence of testing was counter-balanced based on pre-stimulation duration.

23*Laser was modulated at 20 Hz on a 2 s ON and 3 s OFF cycle with 10 ms pulse width.

24The following protocols were performed back to back to test the decay kinetics of pre-stimulation on food intake. The sequence of testing was counter-balanced based on delay duration.

25*Laser was modulated at 20 Hz on a 2 s ON and 3 s OFF cycle with 10 ms pulse width.

26The following protocols were performed back to back to test the kinetics of the sustained hunger signal on the timescale of tens of seconds.

27Repeated 30 times. Food always available.

28Repeated 60 times. Food always available.

29*Laser was modulated at 20 Hz with 10 ms pulse width.

30The following protocols were performed back to back to test the effect of concurrent stimulation and its interaction with stimulation intensity.

31Food always available. Laser was modulated at 20 Hz with 10 ms pulse width.

32Food always available. Laser was modulated at 20 Hz on a 2 s ON and 3 s OFF cycle with 10 ms pulse width.

33Housing and habituation of animals was identical to the optogenetic food intake assays. Mice were additionally habituated to handling for I.P injection twice before actual experiments. After habituation, mice were tested at most once every day during light phase for food intake (2 hr). Mice were injected with CNO (0.3 mg/kg) or vehicle (0.01 mg/kg DMSO) at the beginning of each food intake assay (2 hr).

34For training, mice were acutely food deprived 5 hr before the start of dark cycle and trained with fixed-ratio 1 (FR1), FR3, and FR7 protocols overnight until active lever presses exceeded learning thresholds, which were 50, 150, and 300 respectively. Of note, NPY- mice were trained for longer periods due to their slower learning rate that was partially caused by their lack of exploration in the operant chamber. Mice were then acutely food deprived overnight and tested with a progressive ratio 3 (PR3) task for 1.5 hr at the start of the dark phase.

35For optogenetic experiments, during the first 70 min of the testing protocol (habituation/pre-stim), access to the levers and pellet trough was blocked using a custom-cut acrylic board. After the initial 70 min, the acrylic board was removed, and a single pellet was delivered following active lever presses according to the PR3 schedule (3,6,9,12…). Each experiment was subject to multiple repetitions to test robustness.

36Immunofluorescence was performed as previously described using the following pairs of primary and secondary antibodies: chicken anti-GFP (1:1000; Abcam Cat# ab13970, RRID:AB_300798) and polyclonal donkey anti-chicken FITC (1:1000; Abcam Cat# ab63507, RRID:AB_1139472); goat anti-Fos (1:500; Santa Cruz Biotechnology Cat# sc-52-G, RRID:AB_2629503) and donkey anti-goat Alexa-fluorophore 568 (1:1000; Thermo Fisher Scientific Cat# A-11057, RRID:AB_2534104); rabbit anti-NPY (1:1000; Cell Signaling Technology Cat# 11976, RRID:AB_2716286) and donkey anti-rabbit Alexa-fluorophore 647 (1:1000; Thermo Fisher Scientific Cat# A-31573, RRID:AB_2536183); goat anti-HA (1:1000; for validation of HA-tagged hM3Dq expression; Abcam Cat# ab9134, RRID:AB_307035) and donkey anti-goat Alexa-fluorophore 568 (1:1000; Thermo Fisher Scientific Cat# A-11057, RRID:AB_2534104).

37Raw behavioral data were analyzed with custom MATLAB scripts to obtain time-stamped behavioral events. Multiple measurements from the same mouse in the same experiment (e.g. on different days) were considered technical repeats and were averaged before statistical analyses. Control experiments were repeated at the beginning and end of each set of assays. Experiments with technical failures (e.g. dysfunction of the behavioral system, pellets stuck in feeder, broken fiberoptic patch cables) were excluded from further analysis and repeated. The average of these technical repeats for each mouse in each experiment was considered a single biological repeat. All sample sizes are numbers of biological repeats. Data was analyzed by two-way ANOVA using Graphpad Prism seven to test for an effect of genotype and stimulation protocol or one-way ANOVA. Individual p-values were corrected using Holm-Sidak’s multiple comparisons test. Regression analysis for experiments investigating feeding kinetics was performed using Graphpad.