Appetite controlled by a cholecystokinin nucleus of the solitary tract to hypothalamus neurocircuit

The nucleus of the solitary tract (NTS) is a key gateway for meal-related signals entering the brain from the periphery. However, the chemical mediators crucial to this process have not been fully elucidated. We reveal that a subset of NTS neurons containing cholecystokinin (CCKNTS) is responsive to nutritional state and that their activation reduces appetite and body weight in mice. Cell-specific anterograde tracing revealed that CCKNTS neurons provide a distinctive innervation of the paraventricular nucleus of the hypothalamus (PVH), with fibers and varicosities in close apposition to a subset of melanocortin-4 receptor (MC4RPVH) cells, which are also responsive to CCK. Optogenetic activation of CCKNTS axon terminals within the PVH reveal the satiating function of CCKNTS neurons to be mediated by a CCKNTS→PVH pathway that also encodes positive valence. These data identify the functional significance of CCKNTS neurons and reveal a sufficient and discrete NTS to hypothalamus circuit controlling appetite. DOI: http://dx.doi.org/10.7554/eLife.12225.001


Introduction
Obesity has emerged as one of the global healthcare challenges of the 21st century. Common obesity is primarily a consequence of food intake beyond the body's energetic requirements, with the excess energy consequently stored as fat. Mechanistically, the brain integrates and responds to multiple homeostatic hormones, neurotransmitters, nutrients and peripherally generated neural signals to maintain energetic balance (Dietrich and Horvath, 2013;Morton et al., 2014). One of the primary integration nodes within the brain for meal-related and metabolic signals from the periphery is the nucleus of the solitary tract (NTS) (Grill and Hayes, 2012;Rinaman, 2010;Schwartz, 2010;Wu et al., 2012). The NTS hosts a variety of factors to control homeostatic functions, including a small subset of neurons that express cholecystokinin (CCK NTS ) (Garfield et al., 2012;Herbert and Saper, 1990;Vitale et al., 1991).
CCK consists of a family of peptides, the best characterized of which is a 33 amino acid peptide secreted from endocrine cells in the jejunum in response to nutrients in the intestinal lumen. This gut-derived peptide has a number of gastrointestinal (GI) functions including the promotion of satiation/satiety (Gibbs et al., 1973;Saito et al., 1981;Smith and Gibbs, 1994). Peripherally derived CCK does not readily penetrate the brain (Fan et al., 1997;Passaro et al., 1982). However, its short-term effect on appetite has been in part attributed to stimulation of vagal sensory neurons influencing the brainstem (Fan et al., 1997;Gibbs et al., 1973;Saito et al., 1981;Smith and Gibbs, 1994).
CCK is also synthesized within the brain, where it is post-translationally processed into an 8 amino acid peptide (CCK-8) that reduces food intake when centrally infused (Blevins et al., 2000;Hirosue et al., 1993). However, the source of brain CCK that controls appetite has not been established. Recently developed chemogenetic and optogenetic approaches now provide a means to decipher the real-time contribution of discrete neuronal populations and networks to behavior (Sternson and Roth, 2014;Tye and Deisseroth, 2012), although, no direct functional assessment of CCK NTS neurons has yet been undertaken.
Results and discussion CCK NTS neurons are responsive to nutritional state Although CCK is produced in the NTS (Garfield et al., 2012;Herbert and Saper, 1990), little is known about the function of this discrete source of brain CCK. To facilitate visualization and characterization of CCK NTS , we employed a knock-in mouse line expressing Cre recombinase at the Cck locus (Cck-iCre) crossed with a Cre-dependent enhanced yellow fluorescent protein reporter (Ai3) line (Cck YFP ; Figure 1A). CCK-eYFP cells were most abundantly expressed within the caudal aspect of the NTS ( Figure 1B,C), a brain region innervated by gastrointestinal vagal afferents and involved in nutrient sensing (Appleyard et al., 2005;Blouet and Schwartz, 2012;Ritter, 2011;Schwarz et al., 2010).
To determine whether CCK NTS cells are responsive to food intake, Cck YFP mice were exposed to either ad libitum fed, dark cycle fasted or dark cycle fasted followed by 2 hr re-feeding conditions, and a surrogate marker of neuronal activation, c-Fos immunoreactivity (IR), was assessed. In contrast to light cycle ad libitum fed and the dark cycle fasted conditions, ingestion of food on an empty stomach significantly increased c-Fos-IR within CCK NTS cells, indicating responsiveness to food consumption ( Figure 1D,E). To clarify whether this response is related to the nutrients, as opposed to stomach stretching or orosensorial aspects of feeding, we next investigated whether CCK NTS cells are responsive to nutrients if directly delivered to the stomach. Dark cycle fasted Cck YFP mice were eLife digest Obesity primarily results from eating more food than the body requires, the energy from which is then stored as fat. In recent years obesity has become increasingly common, with the resulting health problems presenting one of the major healthcare challenges of the twenty-first century. New ways to tackle the obesity epidemic are therefore required to improve human health on a global scale.
To regulate how much food is eaten, the gut sends chemical messengers to the brain about how much food has been consumed. These messengers activate particular cells in the brain that signal to other brain regions to trigger a decision about whether we've had enough food to eat. This raises a question: if we can artificially activate these cells, can we 'trick' the brain into thinking that food has been consumed? A brain region called the nucleus of the solitary tract (NTS) is known to play a key role in receiving signals from the gut about meals. By studying mice, D'Agostino et al. found that cells in the NTS that make a brain hormone called cholecystokinin (CCK) are particularly activated by food.
Further experiments then used a technique called optogenetics to activate these cells in mice that had free access to different types of food. This activation significantly reduced how hungry the mice were, causing them to eat less food and lose weight. D'Agostino et al. also showed that CCK cells relay the signal about food intake to a brain region called the hypothalamus.
Overall, D'Agostino et al. have found a way to trick the brain into thinking that food has been eaten when it actually hasn't, and for this reason mice eat less without feeling hungry and lose weight. The next step is to try and find a way to activate the CCK cells in obese humans who have health complications associated with excess body weight.
intragrastrically delivered isovolumetric (0.5 ml) non-nutritive water or isocaloric (1 kcal) sucrose or amino acids. As observed with chow intake, gavage of sucrose or amino acids significantly increased c-Fos-IR within CCK NTS cells compared with water ( Figure 1F). These results suggest that CCK NTS cells are activated by nutrient intake.

CCK NTS neuron activation reduces food intake and body weight
We next considered whether activation of CCK NTS neurons could promote satiety by communicating a nutrient consumption signal. Cck-iCre mice were bilaterally injected into the NTS with AAVs that mediate the Cre-dependent expression of designer receptors exclusively activated by designer drugs (DREADDs; expressed as DREADD-mCherry fusion proteins, hM3Dq; Cck-iCre-hM3D q -mCherry NTS ) ( Figure 2A). DREADDs are designer muscarinic receptor variants that can only be activated by an otherwise biologically inert designer drug, clozapine-N-oxide (CNO) (Alexander et al., 2009;Krashes et al., 2011). Using this approach, we achieved expression of the DREADD-fused mCherry reporter protein in the caudal aspect of the NTS ( Figure 2B). CNO activated Cck-iCre-hM3D q -mCherry NTS cells in vivo as shown by increased c-Fos-IR ( Figure 2C) and ex vivo using electrophysiology in NTS slices ( Figure 2D). Activation of CCK NTS neurons in ad libitum fed Cck-iCre-hM3D q -  ; tick marks on x-axis represent 3 hr, measurements collected with 30-min intervals) and (F) a reduction of post-fast re-feeding following CNO compared to saline administration (n = 6; paired t test, t (5) = 4.769, p = 0.005). (G) CNO-induced Figure 2 continued on next page mCherry NTS mice injected with CNO (0.3 mg kg -1 , IP) suppressed food intake for 24 hr ( Figure 2E). Likewise, CCK NTS neuron activation suppressed re-feeding in food deprived mice ( Figure 2F).
We next tested whether CCK itself mediates these food supressing effects of CCK NTS neuron activation. Cck-iCre-hM3D q -mCherry NTS mice were pre-treated with the CCK-A receptor antagonist devazepide which blunted CNO-induced appetite suppression both 4 hr and 24 hr after treatment ( Figure 2G), with no effects in control mice. These results demonstrate CCK-A receptors as downstream effectors of CCK NTS neuron-mediated reduction in feeding.
Given our recent findings that parabrachial CCKergic (CCK PBN ) transmission regulates hepatic glucose production Garfield et al., 2014), we next investigated whether CCK NTS neurons impact glucose homeostasis. Unlike CCK PBN neurons, chemogenetic activation of CCK NTS neurons in Cck-iCre-hM3D q -mCherry NTS mice altered neither circulating blood glucose levels ( Figure 2H) nor glucose disposal following a systemic glucose load ( Figure 2I). These results reveal a neuroanatomical functional divergence in the energy-intake-versus energy-metabolic features of CCKergic neurotransmission within the brain, with CCK NTS neurons modulating food intake, but not glucose homeostasis.
We next tested whether long-term stimulation of CCK NTS neurons promotes a sustained reduction in food intake and body weight loss. We treated Cck-iCre-hM3D q -mCherry NTS and control Cck-iCre-mCherry NTS mice with CNO for 4 days (two injections per day). Repeated activation of CCK NTS neurons resulted in a pronounced reduction in food intake with a concomitant reduction in body weight ( Figure 2J). Following withdrawal of CNO treatment, mice returned to the original body weight within approximately 48 hr and to baseline food intake after a transient phase of rebound feeding ( Figure 2J). These data demonstrate that sustained activation of CCK NTS neurons is sufficient to promote pronounced and reversible anorexia and body weight loss in mice.
Together, these data reveal that activation of CCK NTS neurons promotes anorexia and weight loss, attenuating the homeostatic drive to eat even in face of marked negative energy balance (as occurs in food deprivation).

A CCK NTS fiPVH circuit controls appetite
As a first step to delineate the circuits through which CCK NTS neurons modulate food intake, we surveyed the neuronal activation-like profile of Cck-iCre-hM3D q -mCherry NTS mice treated with CNO or saline. Chemogenetic activation of CCK NTS neurons elicited striking c-Fos-IR in the paraventricular nucleus of the hypothalamus (PVH) in food-deprived mice compared with saline-treated mice (Figure 3-figure supplement 1). The PVH represents a crucial hub for the brain's regulation of energy balance and integrates a diverse range of nutritionally related hormonal and synaptic inputs, including inputs from the NTS (Grill and Hayes, 2012;Riche et al., 1990). To evaluate whether CCK NTS neurons directly innervate the PVH, a Cre-inducible AAV vector expressing Channelrhodopsin-2-eYFP (ChR2-eYFP) was stereotaxically injected into the NTS of Cck-iCre mice ( Figure 3A,B). Cremediated recombination of the vector led to expression of eYFP in CCK NTS cells and their projection fields (due to axonal localization of the ChR2-eYFP fusion protein), revealing robust CCK NTS fi PVH innervation ( Figure 3C To probe for the physiological significance of this CCK NTS fi PVH circuit, Cck-iCre mice injected into the NTS with AAVs encoding for ChR2-eYFP (a blue-light-sensitive cation channel) were implanted with an optic fiber above the PVH to stimulate these CCK NTS efferents ( Figure 3A-D). Ex Sidak's post hoc comparisons, *p<0.05, **p<0.01, ***p<0.001) in Cck-iCre-hM3D q -mCherry NTS as compared to Cck-iCre-mCherry NTS mice. DOI: 10.7554/eLife.12225.004 vivo application of blue light (473 nm) to CCK NTS neurons expressing ChR2 using slice electrophysiology resulted in faithful action potential discharge with stimulation/discharge fidelity preserved at both 10 Hz and 30 Hz (but not 50 Hz, data not shown) in Cck-iCre-ChR2-eYFP NTS cells ( Figure 3E). In vivo photostimulation of CCK NTS axon terminals in the PVH elicited c-Fos-IR in PVH neurons in close proximity of transduced CCK NTS axons ( Figure 3F).
We investigated whether specifically activating CCK NTS axon terminals within the PVH (10 ms, 30 Hz, 1 s on, every 4 s) prior the initiation of mouse nocturnal feeding could communicate a postprandial-like signal, and thereby reduce subsequent spontaneous feeding. Monitoring home-cage food intake revealed that optogenetic activation of the CCK NTS fiPVH circuit produced an almost complete, and reversible, suppression of spontaneous feeding ( Figure 3G). Importantly, optogenetic activation of the CCK NTS fiPVH pathway neither suppresses locomotor activity ( Figure 3H) nor water consumption ( Figure 3I). Furthermore, real-time optogenetic stimulation of this circuit (10 ms, 30 Hz, 1 s on, every 4 s) significantly suppressed re-feeding in food deprived Cck-iCre-ChR2-eYFP NTS mice ( Figure 3J), but not control Cck-iCre-eYFP NTS mice ( Figure 3K). The reduction in food intake in Cck-iCre-ChR2-eYFP NTS stimulated mice was blunted by pre-treatment with the CCK-A receptor antagonist devazepide ( Figure 3J), which did not affect food intake in control mice ( Figure 3K). Next, we tested the role of the CCK NTS fiPVH circuitry to supress feeding in response to palatable food (chocolate pellets). Food deprived Cck-iCre-ChR2-eYFP NTS mice were offered chocolate pellets via a computer-controlled delivery system using a within-subjects design (optogenetic activation or no activation). Optogenetic activation of CCK NTS fiPVH efferent reduced the total number of chocolate pellets consumed ( Figure 3L), indicating that the appetite-suppressing properties of the CCK NTS fiPVH circuit maintains salience even when the homeostatic drive to eat following food deprivation is boosted with a hedonic component.
To increase the stringency of genetic identification of MC4R-expressing neurons, we also used a second and recently described knock-in mouse line (Mc4r-t2a-Cre) that express Cre recombinase from the endogenous and transcriptionally active Mc4r locus (Garfield et al., 2015) and in which MC4R-expressing neurons are fluorescently labeled by means of Cre-enabled tdTomato expression (Mc4r-t2a-Cre::tdTomato; Figure 3-figure supplement 3A). Single-cell qPCR of manually sorted

A CCK NTS fiPVH circuit encodes positive valence
In addition to the homeostatic regulation of energy balance, the NTS is also associated with negative valance and anorexia related to nausea (Lachey et al., 2005;Rinaman, 2004;Swank and Bernstein, 1994). We next considered whether the CCK NTS fiPVH circuit is associated with the induction of an aversive physiological state. The motivational valence of this circuit was assessed using a real-time place preference test (Stamatakis and Stuber, 2012). We observed no place preference when Cck-iCre-ChR2-eYFP NTS or Cck-iCre-eYFP NTS controls were tested under normal energy balance conditions, an important indication that activation NTS CCK fiPVH efferents is not aversive (Figure 4A-C). However, while Cck-iCre-eYFP NTS controls maintained no chamber preference when food deprived, calorie depleted Cck-iCre-ChR2-eYFP NTS mice exhibited a significant preference for the photostimulation-paired chamber ( Figure 4B,C). This behavioral response is reminiscent of that observed following direct optogenetic activation of PVH MC4R-expressing neurons (Garfield et al., 2015). We further examined this observation using a T-shaped maze. In this test, exploration of one of the three maze's arms was paired with the photostimulation of the NTS CCK fiPVH efferent ( Figure 4D). Over the three testing trials Cck-iCre-ChR2-eYFP NTS mice, but not Cck-iCre-eYFP NTS controls, developed a clear preference for the photostimulation-paired arm ( Figure 4E,F). Thus, when hungry and in the absence of food, mice sought out activation of the NTS CCK efferent to PVH, revealing the CCK NTS fiPVH circuit encodes positive valence. These data suggest that the activation of this circuit provides relief from the unpleasantness of energy deficit by mimicking a post-prandial phase.
While further studies attempting site-specific Cck loss of function are needed to fully clarify the physiological necessity of NTS CCKergic transmission in eating behavior and body weight regulation, here we reveal that activation of CCK NTS produces a prolonged effect on appetite and a rapid reduction in body weight. We did not find evidence that this reduction in energy intake is due to an aversive state, but rather, reveal that the CCK NTS fiPVH circuit transmits positive valence in an energy-state-dependent manner; a particularly attractive prospect considering that in patient populations the negative valence of energy deficit is a major factor contributing to low compliance on weight-loss diets. Cre mice were previously described (Garfield et al., 2015). Mice were provided with standard mouse chow and water ad libitum, unless otherwise noted and housed at 22-24˚C with a 12-hr light/ 12-hr dark cycle. All experimental procedures were performed in accordance with the UK Animals (Scientific Procedures) Act 1986 or the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee.

Stereotaxic surgery
NTS delivery of AAVs was achieved through a modified stereotaxic procedure. Briefly, 5-8 week-old male mice were anesthetized with a mixture of ketamine and xylazine dissolved in saline (100 and 10 mg/kg, respectively; 10 ml/kg). Mice were placed in a stereotaxic frame (Kopf Instruments) using ear bar with head angled at about 45˚. Under magnification, an incision was made at the level of the cisterna magna and neck muscles carefully retracted. A 33G needle was used for dura incision. The obex served as reference point for injection. Injections were performed using a glass micropipette (diameter 20-40 mm). NTS coordinates were A/P, -0.2; M/L, ±0.2; D/V, -0.2 from the obex. Virus was delivered under air pressure using a PLI-100A Pico-Injector (Harvard Apparatus, Cambridge, UK). About 150 nl of virus/side were delivered with multiple microinjections over 4-5 min. The pipette remained in place for a minimum of 3 min after injection. Animals were administered an analgesic (5 mg/kg Carprofen, subcoutaneously) for 2 days post-operatively and given a minimum of 14 days recovery before being used in any DREADD experiments. During this time, mice were acclimated to handling and IP injections. For optogenetic experiments, optic fibers were placed above the PVH at least 4 weeks after the NTS viral delivery to ensure labelling of distal projection sites. Based on the Franklin and Paxinos Mouse Brain Atlas (Franklin and Paxinos, 2008), the following coordinates were used for targeting optic fibers (200 mm diameter core; CFMLC22U-20; Thorlabs, Newton, NJ) at the PVH (mm from Bregma): A/P, À0.65; M/L, ± 0; D/V, À4.20 mm. After placement of the optic fiber, mice were allowed additional 14 days post-surgical recovery, as described above. Mice were excluded from behavioral analysis if post hoc histological validation showed either no viral transduction or misplaced optical fibers.

Energy balance and body weight studies
Dark-cycle food intake. Mice were injected with saline or CNO (0.3 mg/kg; IP Tocris Bioscience) 30 min prior to the onset of the dark. At onset of dark, food was returned and intake recorded automatically via the TSE Phenomaster system (TSE, Bad Homburg, Germany) or manually (Figure 2g). Post-fast re-feed. Mice were fasted overnight. The following morning, mice were treated IP with saline or CNO and food returned 30 min later and intake recorded manually. Daily treatment. Mice were treated IP twice daily for 4 days with 0.3 mg/kg of CNO or saline near the onset of the dark and light cycles. Food intake and body weight were recorded daily prior to the dark cycle. Nutrient gavage. Mice were fasted overnight and the following morning, received intragastric 0.5 ml volume gavage of water, amino acids (Peptone enzymatic digest from casein, Sigma-Aldrich, Dorset, UK; Cat. No. 70172) or sucrose Cat. No. 84100). Amino acids and sucrose were diluted in drinking water and delivered in an amount of 1kcal. Blood glucose studies. Mice were food deprived overnight and blood glucose concentration was determined from tail bleeds using OneTouch Ultra glucometer and test strips (LifeScan, Johnson and Johnson Medical Limited, Livingstone, UK). Basal blood glucose concentration was determined prior to injection of any substances.
Photostimulation for feeding experiments: light pulse trains 10-ms pulse width, 30 Hz, 1 s on, 4 s off. Photostimulation for place preference and T-maze tests: light pulse trains 10-ms, 30 Hz, 1 s on, 0.5 s off. Light power was adjusted such that the light power exiting the fiber optic cable was at least 10 mW using a digital optical power meter (PM100D, Thorlabs) and an online light transmission calculator for brain tissue (http://web.stanford.edu/group/dlab/cgi-bin/graph/chart.php). After the completion of photostimulation experiments, mice were perfused and the approximate locations of fiber tips were identified according to the atlas coordinates (Franklin and Paxinos, 2008). After removal, optic fiber were connected to optic fiber cable and tested for light transmission.

Behavioral testing
Mice were tested for real-time place preference in a model in which one chamber was paired with 30-Hz photostimulation and the other, identical chamber resulted in no photostimulation. Total test duration was 20 min. A similar protocol was used for the T-maze experiments. For the T-maze test mice were tested in three 10-min trials with an inter-trail interval of 15-20 min, during which mice were returned to the home cage. Time spent in the photostimulation versus non-photostimulation zones was recorded via a CCD camera interfaced with Any-maze software (Stoelting, Wood Dale, IL).

Brain tissue preparation and immunohistochemistry
Following deep terminal anesthesia with pentobarbital, mice were transcardially perfused with phosphate-buffered saline (PBS) followed by 10% neutral buffered formalin (Fisher Scientific, Loughborough, UK). Brains were extracted, cryoprotected in 30% sucrose in PBS, sectioned coronally on a freezing sliding microtome (Bright solid state freezer series 8000, Bright Instruments, Luton, UK) at 30 mm and collected in four equal series. IHC was performed using standard methods and as previously described (Garfield et al., 2012). Briefly, sections were washed in PBS before blocking in 0.5% BSA/0.25% Triton X-100 in PBS for 1 hr at room temperature. Tissue was incubated overnight at room temperature in blocking buffer containing the primary antibodies: rabbit anti-c-FOS (EMD Millipore, Livingston, UK; Cat. No. PC05; diluted 1/5000), or chicken anti-GFP (Abcam, Cambridge, UK; Cat. No. ab13970; diluted 1/1000). The next day, sections were washed in PBS then incubated in blocking buffer containing appropriate secondary antibody (Alexa Fluor, Life Technologies; diluted 1/500) for 1 hr. 3,3 0 -Diaminobenzidine (DAB) staining was used for used for detection of c-Fos-IR using: a rabbit anti-c-FOS (EMD Millipore Cat. No. PC05; diluted 1/ 5000), secondary biotin-SP donkey anti-rabbit (Jackson Immunoresearch, West Grove, PA; Cat. No. 711-065-152; diluted 1/250), an avidin/biotin-based peroxidase system (Vectastain Elite ABC Kit, Vector Laboratories, Burlingame, CA; Cat. No. PK-6100) and a developing kit (DAB Peroxidase Substrate Kit, Vector Laboratories, Cat. No. SK-4100). c-Fos-IR was pseudocolored using Photoshop software to appear colored in images. Sections were mounted onto microscope slides and coverslipped in an aqueous mounting medium (Vectashield Antifade Mounting Medium, Vector Laboratories, Cat. No. H-1000). Slides were imaged on a VS120 slide scanner (Olympus, Essex, UK) or AXIOSKOP2 (Zeiss, Oberkochen Germany). For counting of c-FOS-IR nuclei, the boundaries of the nucleus were defined using neuroanatomical landmarks and the Franklin and Paxinos Mouse Brain Atlas (Franklin and Paxinos, 2008).

Statistics
Statistical analyses were performed using Prism 6 (Graphpad Software, La Jolla, CA). Data were analyzed using t-test, one-way ANOVA, two-way or repeated measures (RM) ANOVA with post hoc comparisons, where appropriate. N represents independent biological replicates. No statistical methods were used to predetermine sample sizes. Sample size was computed based on pilot data and published literature. Data are presented as mean ± SEM and statistical significance was set at p<0.05.