Identifying Parabrachial Neurons Selectively Regulating Satiety for Highly Palatable Food in Mice

Visual Abstract


Introduction
Proper regulation of nutrient intake and food consumption is essential for survival and homeostasis in chemotrophic organisms. Feeding is heavily regulated by a multitude of signals coming from exteroceptive processes, such as the availability and taste of food, and interoceptive processes, such as feedback from the gut and adipose tissues (Chambers et al., 2013;Andermann and Lowell, 2017). The signals from these exteroceptive and interoceptive processes are relayed by multiple neural circuits that regulate the initiation and termination of feeding (Sternson and Eiselt, 2017). These neural circuits may undergo maladaptive plasticity in humans suffering from conditions such as anorexia nervosa and obesity, resulting in dysfunctional appetite (Farooqi and O'Rahilly, 2008). Additionally, within these neural circuits, there are a variety of subcircuits which regulate various aspects of feeding, such as hunger, palatable appetite, satiety, and aversive meal termination (Sternson et al., 2016;Sternson and Eiselt, 2017). This complexity makes it challenging to dissect which circuits regulate which aspects of appetite and satiety.
The termination of feeding behavior can result from two separate processes: either due to caloric sufficiency (feeling satiated) or due to receiving aversive anorexigenic signals such as malaise. Significant progress has been made toward the pathways underlying aversive feeding suppression. Specifically, several studies revealed a group of calcitonin gene-related peptide CGRP ϩ expressing neurons in the lateral parabrachial nucleus (PB L CGRP ) as the key mediator of the aversive signal (Campos et al., 2016(Campos et al., , 2018Carter et al., 2013Carter et al., , 2015. Activation of PB L CGRP neurons induces meal termination, anorexia, and affective nociceptive behavior (Carter et al., 2013;Cai et al., 2014;Han et al., 2015;Campos et al., 2016), whereas silencing these cells prevents anorexia in a cancer model . These PB L CGRP cells receive both direct excitatory input from Th ϩ and Cck ϩ cells in the nucleus of the solitary tract (NTS; Kreisler et al., 2014;D'Agostino et al., 2016;Roman et al., 2016), as well as direct inhibitory input from agouti-related peptide (AGRP ϩ ) expressing neurons in the hypothalamic arcuate nucleus (ARC), which are hunger-activated and promote feeding (Aponte et al., 2011;Krashes et al., 2011;Atasoy et al., 2012;Betley et al., 2013;Burnett et al., 2016;Livneh et al., 2017). The PB L CGRP neurons project to the central amygdala (CeA) and synapse with a subset of protein kinase c delta (PKC-␦ ϩ ) expressing CeA GABAergic neurons to suppress feeding (Cai et al., 2014).
By contrast, our understanding of the circuitry underlying physiologic/homeostatic satiety-mediated meal termination remains incomplete. After food consumption, a cocktail of hormones and neurotransmitters is secreted from enteroendocrine cells residing in the stomach and small intestinal epithelium which signal caloric value and satiety. Vagal sensory afferents detect these signals and transmit the information to NTS (Grill and Hayes, 2009;Chambers et al., 2013;Gribble and Reimann, 2016;Williams et al., 2016;Roman et al., 2017;Kaelberer et al., 2018). Distinct groups of NTS neurons then send this information to multiple brain centers, such as the dorsomotor nucleus of the vagus, the PB L , ARC, and the paraventricular hypothalamic nucleus (PVH; Rinaman, 2010;Trapp and Cork, 2015;D'Agostino et al., 2016;Travagli and Anselmi, 2016). Interestingly, melanocortin-4 receptor neurons in PVH (PVH MC4R ) send direct projections onto an unidentified population of neurons in the dorsal portion of PB L , which appears to be spatially distinct from where CGRP ϩ neurons are located in the ventral portion of PB L (Shah et al., 2014;Garfield et al., 2015;Li et al., 2019). Activation of PVH MC4R terminals in PB L induced satiety, while silencing the projections increased food consumption in mice (Garfield et al., 2015;Li et al., 2019). Thus, there appears to be two separate populations in PB L : one with an unknown identity that mediates caloric-sufficient satiety versus the CGRP ϩ population that mediates aversive appetite suppression.
In this study, we show that consumption of various highly palatable liquids and food activates a subset of neurons located in the caudal lateral subnucleus of the PB L (PB cl ) which are distinct from PB L CGRP neurons. We used the activity-dependent capturing method called CANE (Sakurai et al., 2016;Rodriguez et al., 2017), to identify and characterize PB cl -palatable-food activated neurons (PANs). We further discovered that activation of these PB cl PANs induces reward-like place preference behaviors yet decreases condensed milk consumption, while inhibition of these neurons increases milk consumption.

Animal statement
All experiments were conducted according to protocols approved by the Duke University Institutional Animal Care and Use Committee.

Animals
Adult (P30 -P60) male and female C57/BL6 mice (The Jackson Laboratory, stock 000664) were used for immunohistochemistry and in situ hybridization. Male and female Fos TVA mice (Sakurai et al., 2016; The Jackson Laboratory, stock 027831) were used for capturing PB cl PANs with the CANE technology, immunohistochemistry, and behavioral experiments for both ChR2 or TelC experimental and GFP control groups. All mice were housed in a vivarium with normal light/dark cycles in cages with one to five mice. A day before experiments, we singly housed mice. We used two exclusion criteria for our subjects: (1) poor recovery or other health concerns following surgical intervention or (2) missed injection or implantation target, as determined by histologic analysis. Animals were randomly selected from each litter. Random group allocation was maintained throughout the study, within constraints set by availability of in-house, purpose bred lines. Experimenter blinding was sufficient to control for selection bias. Furthermore, behavioral analysis relied on objective, automated measurements.

Surgery
Animals were anesthetized with isoflurane in a stereotaxic frame (David Kopf Instruments) and small craniotomies were made over the target area. To target the caudal-lateral region of PB L , mice were mounted in the stereotaxic frame at an angle such that lambda was ϳ180 m ventral to bregma (in practice, 140 -240 m). The stereotaxic coordinates of virus injection and custommade optic fiber (200-m core diameter, Thorlabs) were AP -4.25 Ϯ 0.15 mm, ML 1.35 Ϯ 0.15 mm, and DV -3.1 Ϯ 0.1 mm. The thin glass capillary was slowly lowered to the target site to minimize the brain injury. Virus was delivered into the target site at a flow rate of 100 nl/min using a pulled thin glass capillary (Warner Instruments) connected to an UltraMicroPump controlled by a SYS-Micro4 Controller 15 (World Precision Instruments).

Floating section in situ hybridization
Floating section in situ hybridization was performed following a previously described protocol (Stanek et al., 2016;Bellavance et al., 2017). All mice were deeply anesthetized with isoflurane, and then transcardially perfused  Figure 2. Optogenetic activation of PB cl PANs induces place preference and decreases condensed milk consumption. A, Schematic illustration of strategy to selectively express ChR2 or GFP in PB cl PANs in Fos TVA mice using CANE. B, Schematic illustration of RTPP test. C, Quantification of time the PAN-ChR2 group spent in preferred chamber before, during, and after optogenetic stimulation (n ϭ 6 mice; one-way repeated measures ANOVA; ‫ء‬p ϭ 0.01; ‫ء‬p ϭ 0.04; p Ͼ 0.99; F (1.46,7.29) ϭ 9.22). Data are mean Ϯ SEM. D, Quantification of time the PAN-GFP group spent in preferred chamber before, during, and after optogenetic stimulation (n ϭ 5 mice; one-way repeated measures ANOVA; p Ͼ 0.99; p Ͼ 0.99; p Ͼ 0.99; F (1.42,5.69) ϭ 0.05). Data are mean Ϯ SEM. E, Schematic illustration of CPP test. F, Quantification of time the PAN-ChR2 group spent in preferred chamber before and after 2 d of optogenetic stimulation (n ϭ 8 mice; two-tailed paired Student's t test; ‫ء‬p ϭ 0.02; t (7) ϭ 3.05) Data are mean Ϯ SEM. G, Quantification of time the PAN-GFP group spent in preferred chamber before and after 2 d of optogenetic stimulation (n ϭ 5 mice; two-tailed paired Student's t test; p ϭ 0.19; t (4) ϭ 1.57) Data are mean Ϯ SEM. H, Schematic illustration of liquid/food intake assay. I, Measured amount of condensed milk consumed at 1 h (n ϭ 6 mice), 2 h (n ϭ 3 mice), and 3 h (n ϭ 3 mice) after condensed milk presentation with and without optogenetic stimulation (p Ͼ 0.99; p ϭ 0.37; ‫ء‬p ϭ 0.01; two-way repeated measures ANOVA; F (3,14) ϭ 4.38). Data are mean Ϯ SEM. J, Measured amount of regular chow consumed at 1 h (n ϭ 6 mice), 2 h (n ϭ 3 mice), and 3 h (n ϭ 3 mice) after chow presentation with and without optogenetic stimulation (p Ͼ 0.99; p ϭ 0.98; p ϭ 0.92; two-way repeated measures ANOVA; F (3,14) ϭ 0.93). Data are mean Ϯ SEM.
PFA. Samples were rinsed with three times with PBS in DEPC water (DEPC PBS), washed with 0.3% Triton X-100 in DEPC PBS, treated with 5 g/ml protinease K in DEPC PBS, acetylated using a solution of 0.465-g TEA, 28-l 10 N NaOH, 25-ml DEPC water, and 62.6-l acetic anhydride. Samples were incubated for 1 h at 60°C in a prehybridization buffer solution containing 0.2ϫ SSC, 10% blocking reagent (Sigma-Aldrich), 50% formamide, 2% N-lauroylsarcosine (NLS), 10% SDS for 60 min. This was followed by hybridization of fluorescein isothiocyanate-labeled Fos and digoxigenin-labeled prodynorphin probes which were prepared by amplifying cDNA fragments of Fos and prodynorphin by PCR with the antisense primers containing the T7 promoter sequence. In vitro transcription was then performed from the PCR-amplified template using T7 RNA polymerase (Roche, 10881767001) for the synthesis of the antisense probes (Sakurai et al., 2016). Samples were incubated with probes overnight in hybridization buffer at 60°C. The following day, samples were washed in prehybridization buffer at room temperature. Samples were incubated at room temperature in a 10% blocking solution in TBST, followed by application of 1st primary antibody (anti-DIG-AP 1:3500, Roche, 11093274910) overnight at 4°C. The following day, samples were washed with TBST, followed by 100 mM Tris-HCl, pH 8.0 and hybridization signals for prodynorphin were visualized using FastRed (Sigma-Aldrich, F4648) for 90 min. This was followed by application of 2nd primary antibody (POD anti-FITC 1:500, Roche 1426346910) at RT for 1 h, and hybridization signals for Fos were visualized using FITC-TSA (Sigma-Aldrich, F4648) for 10 min. After one wash in PBS-DAPI (1:5000) for 5 min and two washes in buffer solution, sections were mounted on glass slides.

Image acquisition and quantification
Samples were imaged using a Zeiss 700 laser scanning confocal microscope. In situ samples were imaged at 20ϫ resolution at three z-positions. All z-positions for each slice were merged into a single image in Adobe Photoshop CS6 for quantification. All other samples were imaged at 10ϫ resolution. The captured neurons and Fos expressing neurons in all immunohistochemistry and in situ hybridization experiments were manually counted, and percentages were calculated within each animal before averaging percentages across animals.

Behavioral experiments for capturing PB cl PANs with CANE virus
The CANE capturing method is as follows (Sakurai et al., 2016). In Fos TVA mice, activated neurons transiently express Fos which induces expression of a destabilized TVA (dsTVA) receptor. Lentivirus pseudotyped with an engineered mutated envelope protein (CANE envelope) specifically binds cells expressing high-level TVA receptor, which are strongly Fos ϩ neurons. In this way, CANE viruses selectively infect Fos ϩ neurons and deliver desired transgenes to be expressed in Fos ϩ neurons.
Here, adult male and female Fos TVA mice at ages more than six weeks were singly housed for at least 1 d and water restricted overnight and were given diluted condensed milk in their home cage for 30 min; 60 -90 min later, mice were anesthetized and underwent stereotaxic surgery for CANE-virus injection.

Optogenetic activation of CANE-captured PB cl PANs in a real-time place preference (RTPP) test and liquid/food intake assay
Channelrhodopsin (ChR2 ϩ ) or control (GFP ϩ ) protein was expressed in CANE-Cre ϩ PB cl PANs by injection of either AAV-CBA-Flex-ChR2(H134R)-mCherry, AAV-EF1␣-DIO-hChR2-eYFP or AAV-CAG-Flex-GFP in adult Fos TVA mice (as described above). Three weeks later, virus injected mice were implanted with custom made optic fibers which were placed above PB cl on both sides and fixed on the skull with dental cement (Parkell). One week later, the animals were subjected to a two-chamber RTPP test in light cycle, using a modified method described in previous studies (Stamatakis and Stuber, 2012;Jennings et al., 2013). The custom-made behavior chamber is 50.1 ϫ 27.7 ϫ 31.2 cm (W ϫ L ϫ H), made with clear acrylic Plexiglas that had distinct stripe patterns in each chamber. For optogenetic stimulation, laser is delivered through patch cables attached to the implanted optic fiber. The RTPP test is as follows. The mouse is placed in the center of the box and allowed to explore both chambers without light stimulation (prestimulation) for 10 min. Generally, after exploration, the mouse shows a small preference for one of the two chambers. Subsequently, blue light stimulation (20 Hz, 20-ms pulse width, ϳ3.5 mW) is delivered whenever the mouse enters or stays in the non-preferred chamber, and light is turned off when the mouse moves to the other chamber (stimulation phase, total 10 min). Finally, the mouse can freely explore both chambers without blue light stimulation (poststimulation) for 10 min. We recorded behavioral data via a webcam (Logitech webcamera, PN 960-000764) interfaced with Bonsai software (Lopes et al., 2015). Real-time laser stimulation was controlled by Bonsai software through Arduino with a custom-made Arduino sketch (Arduino UNO, A00073).
After one week, the same group of mice were subjected to another behavioral test, where animals were subjected to a two-chamber classic conditioned place preference (CPP) test in same behavior chamber used for RTPP. The mouse is first habituated to the chamber on day 1. On day 2, the mouse is placed in the center of the box and allowed to explore both chambers without light stimulation (prestimulation) for 10 min. Generally, after exploration, the mouse shows a small preference for one of the two chambers. In the following 2 d (day 3 and day 4), the mouse is closed off in the preferred chamber with no stimulation for 30 min in the morning, and then closed off in the non-preferred chamber with blue light stimulation (20 Hz, 20-ms pulse width, ϳ3.5 mW) for 30 min in the afternoon. On the final day (day 5), the mouse can explore both chambers without blue light stimulation (poststimulation) for 10 min, and their behaviors are recorded and analyzed.
After one week, the ChR2 ϩ group underwent a liquid/ food intake assay in their home cage within 6 -9 P.M. (normal feeding cycle), using a modified method as previously described (Garfield et al., 2015;Fenselau et al., 2017;Li et al., 2019). The liquid/food intake assay is as follows: Mice were given a measured amount of diluted condensed milk or regular mouse chow for 3 h with optogenetic blue light stimulation (5 s on and 1 s off, 20-ms pulses, 20 Hz, ϳ3.5 mW). The order of the liquid or food was randomized among the groups. The amount of liquid or food ingested was measured as the difference between the final measurement and the initial measurement. After completion of the photostimulation experiments, mice were perfused, and the locations of optic fiber tips were identified.

Genetic silencing of CANE-captured PB cl PANs in a liquid/food intake assay
Tetanus toxin (TeLC) or control GFP was expressed in CANE-Cre ϩ PB L -satiety neurons by injection of either AAV-hSyn-Flex-TeLC-P2A-eGFP or AAV-CAG-Flex-GFP in adult Fos TVA mice (as described above). Three days before virus injection (baseline), the experimental group was subjected to the liquid/food intake assay in their home cage from 6 -9 P.M. (normal feeding cycle), where they were given a measured amount of diluted condensed milk or regular mouse chow. The order of the liquid or food was randomized among the groups. The amount of diluted condensed milk or water was measured every hour for 3 h. This assay was repeated for both TelC and control groups 8 d after virus injection (TeLC, control). The amount of liquid or food ingested was measured as the difference between the measurements taken at each hour and the initial measurement. After completion of the silencing experiments, mice were perfused, and viral expression was assessed.

Statistical analysis
All data are presented as mean Ϯ SEM. Statistical analysis of data were conducted and graphed using GraphPad Prism 8 (Prism software). Significance was set at p Ͻ 0.05. All data were normally distributed as determined by a Shapiro-Wilk normality test. For Figure 1D, unpaired Student's t test was used to compare milk and water groups. For Figure 1E, two-way ANOVA of treatment over brain region was used with a Bonferroni post hoc test to assess differences in Fos expression. For Figure 2F,G, paired Student's t tests were used to compare before and after optogenetic stimulation. For Figure  2C,D, one-way repeated measures ANOVA of before, during, and after optogenetic-stimulation was used followed by a Bonferroni post hoc test to assess changes in time spent in preferred side. For Figures 2I,J, 3B , twoway repeated measures ANOVA of mouse group and time was used followed by a Bonferroni post hoc test to assess changes in liquid or food intake over time. For Figure  3C,D, one-way ANOVA of mouse group over liquid or food intake was used followed by a Bonferroni post hoc test. Relevant values used for statistical analysis are included in the figure legends as follows: (t test) t subscript degrees of freedom ϭ t statistic; (ANOVA) F subscript betweengroups degrees of freedom, within groups degrees of freedom ϭ F statistic.

Data availability
The data collected in this study are available from the corresponding author on request.

Consumption of highly palatable foods activates PB cl neurons
Mice were given access to drink from either a bottle containing water, a non-caloric liquid, or a bottle containing 50% condensed milk, a high-caloric palatable liquid food, alone, for 30 min after overnight water restriction, euthanized, and brain sections were immunostained for expression of immediately-early gene Fos as a marker for activated neurons in PB L (Fig. 1A). Drinking condensed milk activated neurons in PB L , resulting in significantly more Fos ϩ neurons than drinking water (condensed milk, 116.4 Ϯ 9.83, and water, 69 Ϯ 11.52 total Fos ϩ neurons; p ϭ 0.007; n ϭ 10, 7; Fig. 1B,C). Furthermore, drinking condensed milk induced robust Fos ϩ expression specifically in PB cl , whereas drinking water preferentially activated neurons in the superior lateral subnucleus of the PB L (PB sl ; PB cl : condensed milk 105.3 Ϯ 8.93, water 6 Ϯ 2.61 Fos ϩ neurons; p Ͻ 0.0001; n ϭ 10, 7; and PB sl : condensed milk 11.1 Ϯ 3.37, water 63 Ϯ 9.28 Fos ϩ neurons; p Ͻ 0.0001; n ϭ 10, 7; Fig. 1B-D). To determine whether PB cl neurons were activated by intake of other palatable liquid or food, separate groups of mice were given access to either chocolate Ensure, peanut butter, 146 mM sucrose dissolved in water, or 1 mM sucralose dissolved in water (Extended Data Fig. 1-1A). Drinking chocolate Ensure, another high-caloric palatable liquid, also induced robust Fos ϩ expression specifically in PB cl (Extended Data Fig. 1-1B). Interestingly, consuming peanut butter, a high-caloric solid food, induced robust Fos ϩ expression in the same PB cl region (Extended Data Fig.  1-1C). In contrast, drinking either sucrose or sucralose, which are low-or non-caloric sweet tastants, preferentially activated neurons in PB sl , but not in PB cl (Extended Data Fig. 1-1D,E). These results are similar to that observed for drinking water. In other words, non-caloric and caloric liquid intakes are regulated by distinct circuits in PB L , a result consistent with a recent finding Oxtr ϩ PB L neurons selectively regulating non-caloric fluid drinking (Ryan et al., 2017). Therefore, we called these PB cl neurons palatable-food activated neurons (PANs). Notably, a previous study showed that the satiety-signaling PVH MC4R afferents terminate in the PB cl region where the PANs are located (Shah et al., 2014;Garfield et al., 2015;Li et al., 2019).

PB cl PANs are distinct from CGRP ؉ PB el neurons
Forkhead box protein P2 (FoxP2) transcription factor was previously shown to be primarily expressed in dorsal subregions of PB L (Geerling et al., 2011;Gaub et al., 2016). Using two-color immunofluorescence, we found that a majority of Fos ϩ PB cl PANs expressed FoxP2, but they represented a small fraction of the total FoxP2 ϩ PB L cells (83.19 Ϯ 3.29 of total Fos ϩ PB cl PANs were FoxP2 ϩ ; n ϭ 4; Fig. 1F,J). We next determined whether Fos ϩ PB cl PANs express CGRP and found that there was no overlap (0 of total Fos ϩ PB cl PANs were CGRP ϩ ; n ϭ 4; Fig. 1G). This finding suggests that these Fos ϩ PB cl PANs are likely functionally distinct from anorexigenic PB L CGRP neurons (located in the external lateral part of PB L ).
We next consulted the Allen Brain Atlas to find other potential genetic markers enriched in PB cl subregion. Two candidate genes were identified: tyrosine hydroxylase (Th), which encodes the enzyme regulating the synthesis of the neurotransmitter dopamine, and prodynorphin (Pdyn), which encodes the neuropeptide dynorphin and is previously shown to be expressed in cells implicated in thermo-sensation (Geerling et al., 2011). Using immunohistochemistry, we found that Fos ϩ PB cl PANs do not express Th (0 of total Fos ϩ PB cl PANs were Th ϩ ; n ϭ 4; Fig. 1H,J). Interestingly, using two-color in situ hybridization, we found that ϳ44% of Fos ϩ PANs activated by condensed milk consumption and 38% of Fos ϩ PANs activated by chocolate Ensure consumption were Pdyn ϩ [43.62 Ϯ 8.16; n ϭ 3 (Fig. 1I,J); 38 Ϯ 5.55; n ϭ 4 (Extended Data Fig. 1-1F-H)], suggesting the possibility that dynorphin is released on consumption of highly caloric palatable liquids.

CANE is efficient and selective in activity-dependent capture of PB cl PANs
To determine the functional role of Fos ϩ PB cl PANs in liquid and food consummatory behaviors, we needed to label and manipulate these cells specifically. Since we did not identify any molecular markers that are expressed by all PANs and are only expressed in PANs, we therefore used an activity-dependent method called CANE, to virally tag transiently activated neurons in the genetically engineered Fos TVA mice (Sakurai et al., 2016;Rodriguez et al., 2017). We first determined whether CANE could indeed selectively and efficiently label PB cl PANs. In a two-bout experimental paradigm, CANE was used to capture PB cl neurons activated by drinking condensed milk through co-injection of CANE-LV-Cre and AAV-flex-GFP into the PB cl . Three weeks later, the same animal drank milk again to induce Fos ϩ expression and was then anesthetized and killed to obtain samples for immunostaining (Fig. 1K). In the milk-milk condition, 64.66 Ϯ 4.3% (n ϭ 5) of CANE-captured cells were Fos ϩ (Fig. 1L,M). This indicated that the second condensed milk consumption reactivated many (ϳ65%) of the same cells excited by the first consumption. Thus, CANE is sufficiently specific and efficient to tag PANs.

Optogenetic activation of PB cl PANs neurons elicits place preference
Given that PB cl PANs are activated by consuming highcaloric palatable liquids and solid food, we asked whether artificially activating these cells will produce a positive feeling. To test this, we used CANE to express either channelrhodopsin or GFP (control) in PB cl PANs, and subjected mice to a modified RTPP assay, which has been used in recent studies to assay affective behavior (Stamatakis and Stuber, 2012;Jennings et al., 2013). Optic fibers were implanted bilaterally above PB cl in either PAN-ChR2 mice (n ϭ 6) or PAN-GFP mice (n ϭ 5; Fig. 2A). Mice were habituated and placed in a two-chamber arena. Their behaviors were recorded under three conditions: (1) freely exploring with no stimulation for 10 min (baseline), followed by (2) 10 min of conditioned photoactivation when the mouse is in its non-preferred chamber (stimulation), and followed again by (3) 10 min without stimulation (poststimulation; Fig. 2B). On photo-stimulation PAN-ChR2 mice spent significantly more time on the stimulated side (pre: 37.31 Ϯ 7.61%, stim: 70.63 Ϯ 8.38%, p ϭ 0.01, n ϭ 6; Fig. 2C). In the poststimulation period, all stimulated mice still showed preference of the chamber in which they received photostimulation (pre: 37.31 Ϯ 7.61%, post: 75.58 Ϯ 5.24%, p ϭ 0.04, n ϭ 6; Fig. 2C). Light illumination had no effect on movement and behavior of the PAN-GFP mice (pre: 52.08 Ϯ 1.23%; stim: 50.11 Ϯ 7.19%, post: 52.92 Ϯ 12.17%, p Ͼ 0.99, n ϭ 5; Fig. 2D). These results suggest that the optogenetic stimulation of the PB cl PANs caused an appetitive effect, indicating that the neurons likely encode a positive emotional valence.
We further wanted to determine whether optogenetic activation would be sufficient to induce a longer lasting positive memory using the conventional CPP assay. Mice were habituated first by placing them in the two-chamber arena and allowing free exploration. Subsequently, they were subjected to 2 d of conditioning: mice were paired with photostimulation in the non-preferred chamber for 30 min, and 8 h later, they were placed in the preferred chamber without light stimulation for 30 min. On the fourth day, they explored the arena freely in the absence of light stimulation for 10 min (poststimulation; Fig. 2E). PAN-ChR2 mice (n ϭ 8) spent significantly more time in the chamber where they were photostimulated previously (pre: 21.12 Ϯ 4.912, post: 47.27 Ϯ 6.541, P ϭ 0.02; Fig.  2F). Light illumination had no effect on the movement and behavior of the control PAN-GFP mice (n ϭ 5; pre: 51.46 Ϯ 5.86, post: 36.01 Ϯ 7.753, p ϭ 0.19; Fig. 2G). These results suggest that optogenetic activation of PB cl PANs produced a preference memory.

Optogenetic Activation of PB cl PANs neurons decreases condensed milk consumption
Previous studies have shown that activation of PVH MC4R afferents in PB cl region induced homeostatic regulated satiety (Garfield et al., 2015;Li et al., 2019). Therefore, we wanted to determine whether activating CANE-captured PANs would also affect liquid and food intake. PAN-ChR2 mice (n ϭ 6) underwent a caloric liquid/ food intake assay in their home cage during their normal feeding cycle, where they were given a measured amount of condensed milk or regular mouse chow, which is less palatable, for 3 h either without (baseline) or with optogenetic stimulation ( Fig. 2H; Garfield et al., 2015;Fenselau et al., 2017;Li et al., 2019). At 1 h after the start of feeding, the amount of condensed milk ingested was not different between the two conditions (with or without light stimulation; n ϭ 6; no light, 0.43 Ϯ 0.07; light, 0.5 Ϯ 0.2; p Ͼ 0.99; Fig. 2I). However, when measured after 3 h, the total amount of condensed milk ingested in the PAN-activated condition was significantly less (n ϭ 3; no light, 1.43 Ϯ 0.15; light, 0.83 Ϯ 0.3; p ϭ 0.01; Fig. 2I). Light illumination had no effect on the consumption of regular chow either measured after 1 h (n ϭ 6; no light, 0.13 Ϯ 0.04; light, 0.17 Ϯ 0.09; p Ͼ 0.99; Fig. 2J) or after 3 h (n ϭ 3; no light, 0.43 Ϯ 0.2; light, 0.43 Ϯ 0.12; p Ͼ 0.99; Fig. 2J). These results suggest that activation of PB cl PANs is sufficient to specifically reduce the consumption of highly caloric and appetitive condensed milk, but not the regular chow, over a long duration. These results complement the previous findings revealing the homeostatic role of activating PVH MC4R -PB cl pathway in reduction of liquid/food intake (Shah et al., 2014;Garfield et al., 2015;Li et al., 2019).

TeLC-mediated silencing of PB cl PANs increases condensed milk consumption
We next asked whether silencing the PANs would increase caloric liquid or food consumption. We used CANE to express either TeLC or GFP in PB cl PANs. TeLC abolishes synaptic transmission from PANs. PAN-TeLC (n ϭ 6) and PAN-GFP mice (n ϭ 6) underwent the liquid/food intake assay in their home cage during their normal feeding cycle before (baseline) and after CANE-mediated expression of TeLC or GFP (Fig. 3A). We observed a steady increase of condensed milk intake occurring 5-7 d after TeLC expression (data not shown). This increase in condensed milk intake plateaued after 7 d. On day 7, we observed a significant increase in condensed milk intake compared to Baseline, and compared to PAN-GFP mice at 7 d after GFP expression (hour 1-3: p Ͻ 0.0001; total milk intake: baseline vs TeLC; p ϭ 004; GFP vs TeLC; p ϭ 0.02; Fig. 3B,C). Although we saw momentary body weight gain during the 3-h access to condensed milk, we did not observe an overall increase in body weight 7 d after TeLC expression (data not shown), and we did not test the animals over long-term on a high-caloric diet. There was no difference in condensed milk intake between PAN-GFP mice and baseline of PAN-TeLC mice before TeLC was expressed (total milk intake: baseline vs GFP; p Ͼ 0.99; Fig. 3B,C). Interestingly, there was no significant difference across all groups in regular chow intake (baseline vs TeLC, p Ͼ 0.99; control vs TeLC, p Ͼ 0.99; baseline vs control, p Ͼ 0.99; Fig. 3D). These results suggested that PB cl PANs is specifically required for preventing over consumption of a high-caloric palatable diet, while other pathways (see Discussion) are sufficient for signaling satiety from regular food.

Discussion
Here, we molecularly and functionally characterized an ensemble of neurons located within PB cl that likely relay appetitive and satiating signals after consuming highly caloric liquid and food diets. In previous studies, it was discovered that PVH MC4R neurons regulate satiety and prevent over-feeding and consequent obesity (Balthasar et al., 2005). PVH MC4R axons innervate PB cl (Shah et al., 2014;Garfield et al., 2015). Optogenetic activation of PVH MC4R -PB cl axons reduced food intake and induced positive affective behavior (Garfield et al., 2015). These and other studies have implied that neurons in PB cl play a role in homeostatic caloric-dependent satiety in sated states, but the identities of the caloric-dependent satietysignaling neurons in PB cl were unknown (Balthasar et al., 2005;Shah et al., 2014;Garfield et al., 2015;Li et al., 2019). Using a combination of activity-dependent tagging, in vivo optogenetic activation, and TeLC silencing experiments, we uncovered palatable food activated neurons in PB cl that are molecularly and functionally distinct from CGRP ϩ anorexigenic neurons located in PB el and spatially and functionally distinct from Oxtr ϩ non-caloric fluid regulating neurons located in PB sl/dl (Campos et al., 2016;Ryan et al., 2017).
A total of 44% of PB cl PANs expressed Pdyn suggesting that a heterogeneous population of PANs work together to regulate satiety of highly caloric food in sated states. Optogenetic activation and TeLC silencing of PANs bidirectionally regulated consumption of condensed milk, a highly caloric liquid diet, but not regular chow. Therefore, it is possible that PANs are a subpopulation of a caloric-dependent satiety regulating ensemble in PB cl . Further molecular characterization, potentially using RNA sequencing in combination with the CANE strategy, may be required to reveal the full ensemble of caloric-dependent satiety-signaling neurons in PB cl . This result also demonstrates the selectivity of the CANE method in capturing PB cl neurons activated by consumption of condensed milk. Nevertheless, it has not been determined whether these CANE-captured PB cl PANs neurons are the same PB cl population implicated in receiving satiety-related interoceptive information directly from PVH MC4R neurons. Based on the anatomic location of PANs and their functional role in caloric-dependent satiety, activity-dependent transsynaptic tracing may reveal that majority of these neurons do receive direct input from caloric-dependent satiety regulating PVH MC4R neurons. Thus, future activity-dependent capturing studies are necessary for a detailed analysis and characterization of the PVH MC4R -PB cl satiety-regulating pathway.
Our finding that consumption of high caloric liquid and foods preferentially activates PB cl neurons that signal satiety maintenance sheds some light on a complex circuit of PB L neurons signaling different facets of liquid and food intake regulation: caloric versus non-caloric, liquid versus food, nutritional composition, sated versus hunger states, and appetitive versus aversive. Recently, multiple parallel PB L circuits have been discovered to be critical regulators of satiety. Although the PVH MC4R -PB cl circuit plays a role in regulating satiety in a sated state, these neurons comprise of ϳ50% of PVH neurons that mediate satiety. Recently, it was shown that the other ϳ50% of PVH neurons, which expressed Pdyn and project to the prelocus coeruleus (preLC), similarly regulate satiety in sated states (Li et al., 2019). In parallel, a group of Oxtr ϩ neurons located in PB sl/dl specifically regulate non-caloric fluid intake, but not caloric-dependent satiety (Ryan et al., 2017). Together, some of these distinct PB L and preLC circuits (i.e., Pdyn ϩ caloric-signaling neurons and Oxtr ϩ neurons fluid intake-signaling neurons) may act together for integrated food and fluid intake regulation. Other PB L circuits, however, may act antagonistically with one another in regards to the valence of food intake (i.e., Pdyn ϩ appetitive-signaling neurons and CGRP ϩ malaise-signaling neurons). Further work investigating how much these circuits contribute in working synergistically or antagonistically is critical to understanding PB L 's role within the complex fluid/food regulating circuit.
The PB cl PANs were identified through Fos expression after condensed milk consumption. Although Fos is a surrogate for elevated neural activity, it lacks temporal resolution and does not give any information about second-to-second neural dynamics. Therefore, it is unclear when PB cl neural activity increases during consumption. Future studies can take advantage of the CANE method to express a genetically encoded calcium indicator (such as GCaMP), and use in vivo calcium imaging (Ghosh et al., 2011;Chen et al., 2013) to track the activity of these cells in real-time while the mouse is consuming various liquids or foods (e.g., water, sucrose, condensed milk, chow) in a multitude of contexts and states (e.g., hungry vs sated). These techniques will shed light on the population dynamics of PB cl cells in appetite/satiety regulation, whether and when the magnitude or firing rate of neural response to appetitive liquid and food consumption changes, and if there are subsets of neurons firing at distinct time points throughout feeding. Overall, the findings in this study further advanced our understanding of the neural circuit regulating satiety mediated termination of feeding. Future studies should take advantage of these PB cl PANs to identify their downstream targets and elucidate how signals from PANs are used to control feeding behaviors.