Abstract
The central nucleus of the amygdala plays a significant role in alcohol use and other affective disorders; however, the genetically-defined neuronal subtypes and their projections that govern these behaviors are not well known. Here we show that ablation of neurotensin-expressing neurons in the central nucleus of the amygdala of mice decreases their ethanol consumption and preference for ethanol. Furthermore, optogenetically stimulating projections from these neurons to the parabrachial nucleus is reinforcing, and increases ethanol consumption while reducing food consumption. These data suggest that this central amygdala to parabrachial nucleus projection influences the expression of reward-related phenotypes and is a novel circuit promoting alcohol consumption and regulating state-dependent food consumption.
Introduction
The central nucleus of the amygdala (CeA) is a heterogeneous structure that plays an important role in the regulation of appetitive, aversive, and ethanol-mediated behaviors (Mahler and Berridge, 2009; Tye et al, 2011; Robinson et al, 2014; McCall et al, 2015; Warlow et al, 2017; Kim et al, 2017; Douglass et al, 2017). While some data have shed light on the neuronal subpopulations influencing fear- and feeding-related behaviors (Haubensak et al, 2010; Cai et al, 2014; Douglass et al, 2017), it remains unclear which neuronal subpopulations within the CeA and which CeA efferents influence ethanol consumption (Gilpin et al, 2015). One CeA neuron subpopulation that may regulate ethanol phenotypes are the neurons that produce the 13 amino-acid neuropeptide neurotensin (NTS). Indeed, considerable evidence suggests that NTS systems are critical for reward and anxiety processes (Cáceda et al, 2006; Leinninger et al, 2011; Fitzpatrick et al, 2012; Prus et al, 2014, 2014; McHenry et al, 2017), and global manipulations of the NTS system disrupt ethanol-related phenotypes (Lee et al, 2010, 2011). The majority of studies investigating NTS-expressing neurons, however, focus on those that project from the lateral hypothalamus (LH) to the ventral tegmental area (VTA) and on NTS interactions with dopamine (Binder et al, 2001; Kempadoo et al, 2013; Leinninger et al, 2011; McHenry et al, 2017).
Little is known about the role that CeA projections to the hindbrain play in the context of ethanol consumption. Early studies found that the CeA projection to the hindbrain’s parabrachial nucleus (PBN) contains NTS+ fibers (Moga and Gray, 1985). The PBN plays a role in the development of conditioned taste aversion (Carter et al, 2015; Grigson et al, 1998), as well as in fluid satiation (Ryan et al, 2017). Additionally, the PBN has been shown to be activated by intraperitoneal injections of ethanol (Chang et al, 1995; Thiele et al, 1996). The PBN, in turn, sends projections back to the CeA that suppress food consumption (Carter et al, 2013). In order to investigate the complex relationship between the CeA and PBN and better understand the role of the CeA-NTS neuronal subpopulation in ethanol consumption and appetitive behaviors, we utilized NTS-ires-cre mice (Leinninger et al, 2011) in conjunction with region-directed genetic lesioning, optogenetic stimulation, and behavioral assays.
Methods
Subjects, stereotaxic surgery, virus injection and fiber implantation
Mice All procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals, as adopted by the NIH, and with approval of an Institutional Animal Care and Use Committee at UNC-Chapel Hill. Adult male (>22g) NTS-ires-cre mice (Leinninger et al, 2011) (Jackson Laboratories, Bar Harbor, ME) were used for all experiments. Animals were maintained on a reverse 12 hour light cycle with ad libitum access to food and water (unless noted). Surgery Mice were anesthetized with inhaled isoflurane (1-3%) and placed in a stereotaxic frame (David Kopf, Germany). For all experiments coordinates for the CeA were as follows (from Bregma, in mm): ML: ± 2.95, AP: − 1.1, DV: − 4.8, for the PBN: ML + 1.4, AP: −5.4, DV: −4.0 (optical fibers). 300 nL of AAV5-Ef1α-FLEX-taCasp3-TEVp (denoted as: CeANTS::casp), AAV5-Ef1α-ChR2-eYFP (denoted as: NTS::ChR2 or NTSCeA→PBN::ChR2), or AAV5-Ef1a-eYFP (denoted as: NTS::eYFP or NTSCeA→PBN::eYFP) was infused into the CeA at a rate of 100 nL/min. Optical fibers were constructed as previously described (Sparta et al, 2011). Mice were allowed to recover for at least 4 weeks prior to experimentation (8 weeks for optogenetic experiments) to ensure adequate expression of virally encoded genes, and lesioning of target neurons or protein incorporation into the membrane. All viruses were made by the UNC Viral Vector Core (Chapel Hill, NC) or the Stanford Viral Vector (Palo Alto, NC). Following behavioral studies, animals with ChR2-eYFP construct were perfused, and brains were sliced to verify expression of virus. Animals with no viral expression in either CeA were removed (n=1), while animals with either bilateral or unilateral viral expression were included in the analysis as our pilot data indicated that unilateral expression of the virus was sufficient to drive real-time place preference (RTPP) behavior (data not shown). Animals expressing the Caspase construct were euthanized and brains were flash frozen for validation using fluorescent in situ hybridization (FISH, see below) and compared to their eYFP controls.
Fluorescent in situ hybridization
Mice were anesthetized (isoflurane), decapitated, and brains were flash frozen on dry ice. 12 µm slices were made using a Leica cryostat (CM 3050S, Germany). FISH was performed using probes constructed against PKCδ, CRF, CRFR1 (type-6, fast blue) and NTS (type 1, fast red) and reagents in the View RNA kit (Affymetrix, Santa Clara, CA). Slides were counterstained with DAPI.
Immunohistochemistry
As previously described (Pleil et al, 2015), mice were perfused with 4% paraformaldehyde (in 0.01 M PBS), brains were removed and allowed to fix for 24 hours followed by cryoprotection in 30% sucrose/PBS. Subsequently brains were sliced at 40 µm on either the CM 3050S or the VT1000 (Leica, Germany). Sections were incubated overnight at 4°C in blocking solution containing primary antibody – anti-sheep tyrosine hydroxylase 1:500 (Pel Freeze), rabbit anti-neurotensin 1:500 (ab43833, Abcam). The following day, sections were incubated in fluorescence-conjugated donkey anti-rabbit IgG Alexa Fluor 647 secondary antibody (1:800, Jackson Immuno) for 2 hr in darkness. 405 neurotrace or DAPI was used as a counterstain.
Microscopy
Images were collected and processed on a Zeiss 710, 780 or 800 a using 20X/0.8 objective and the Zen software (Carl Zeiss, Germany). Image J/Fiji was used for cell counting and data analysis.
Slice preparation and whole-cell electrophysiology
As previously described (Pleil et al, 2015), animals were anesthetized (isoflurane or euthosol) and decapitated. Brains were removed and sliced at a thickness of 200 µm (hindbrain) or 300 µm (CeA) using a Leica VT1200 or VT1000 (Germany) in ice-cold high-sucrose low Na+ artificial cerebral spinal fluid (ACSF in mM: 194 sucrose, 20 NaCl, 4.4 KCl, 2 CaCl2, 1 MgCl2, 1.2 NaH2PO4, 10 glucose, 26 NaHCO3) that had been oxygenated (95% O2, 5% CO2) for a minimum of 15 min. Following slicing, brains were allowed to equilibrate in normal ACSF (in mM: 124 NaCl, 4.4 KCl, 2 CaCl2, 1.2 MgSO4, 1 NaH2PO4, 10 glucose, 26 NaHCO3, 34° C) for at least 30 minutes. Next, slices were transferred to the electrophysiology rig and allowed to equilibrate in oxygenated ACSF (28-30 °C) perfused at 2 mL/min for an additional 30 minutes. Recordings examining cell excitability were performed in current clamp using K-gluconate intracellular recording solution (K-gluconate 135, NaCl 5, MgCl2 2, HEPES 10, EGTA 0.6, Na2ATP 4, Na2GTP 0.4). Recordings examining synaptic currents were made either in CsCl intracellular solution (130 CsCl, 1 EGTA, 10 HEPES, 2 ATP, 0.2 GTP) or CsMethanosulfonate (in mM: 117 Cs methanesulfonic acid, 20 HEPES, 0.4 EGTA, 2.8 NaCl, 5 TEA, 2 ATP, 0.2 GTP). CsCl recordings were conducted in kynurinic acid (3mM) to block glutamatergic currents.
Blood Ethanol Content
Blood ethanol content (BEC) was measured by administering a dose of 2.0 g/kg (20% ethanol w/v, i.p.). Mice were restrained (<2 min) in plexiglass tubes (Braintree Scientific, Braintree, MA) and a scalpel was used to make a small nick in the mouse tail. Blood was collected in a heparinized capillary tube at 30 and 60 minutes following the injection and transferred to a heparinized collection tube, which was then spun down. The plasma was then analyzed for BEC using an Analox-G-5 analyzer (Analox Instruments, Lunenbug, MA)
Ethanol Drinking Paradigms
2-bottle choice In their homecage, mice were given 24 hour access to a bottle of ethanol (3%, 6%, 10% ethanol, unsweetened) and a bottle of water at 3 days/dose. Measurements of ethanol consumption and water consumption were taken every 24 hours. Intermittent Access (IA) was performed as described by Hwa et al (Hwa et al, 2011). Briefly, mice were allowed access to both a bottle of 20% (w/v) ethanol (unsweetened) and water in their homecage on Monday, Wednesday, and Friday. On other days, they only had access to 2 bottles of water. Bottles were rotated daily to ensure that animals did not associate ethanol or water with a particular side of the cage.
Locomotor and Anxiety Assays
All locomotor and anxiety assays were performed using the Ethovision XT tracking software (Noldus Information Technology, Netherlands) to measure location, distance moved, and velocity.
RTPP
Mice were placed in an apparatus (50 × 50 × 25 cm) that was divided down the middle with a door for exploration on both sides, and no distinguishing features on either side. For 20 minutes, mice were allowed to explore the apparatus and received optical stimulation (20 Hz, 473 nm, 10 mW, Aurdino UNO, or Master 8, AMP Instruments, Israel) on one side (counterbalanced) and no stimulation on the other side. oICSS Mice were food-restricted to 80% of their normal food intake for 2 days before optical intracranial self-stimulation (oICSS). They were tethered to the laser and placed in the chamber (15.9 cm × 14.0 cm × 12.7 cm; MedAssociates, VT, USA) for 1 hour. Both nose ports (active and inactive) were baited with a very small amount portion of their normal feed to encourage exploration. A dim house light flashed when the animal had poked the active port along with 5 seconds of stimulation during which time further pokes had no effect (40 Hz, 473 nm, 10 mW).
Open field
Mice were allowed to explore the open field (50 × 50 cm) for 30 minutes where distance traveled, and velocity were measured (Ethovision, Noldus, Amsterdam).
Light-dark box
Mice were placed into the dark enclosed side of the apparatus (Med Associates) and time spent in the light side and entries to the light were monitored for 15 minutes (Ethovision, Noldus, Amsterdam). Elevated Plus Maze. Mice were placed in the center of the apparatus at the beginning of the test. CeANTS::casp and control mice were given 5 minutes to explore the open arm, closed arm, and center portion of the maze, and time spent in arms versus center and number of entries were monitored. NTSCeA→PBN::ChR2 and control mice were similarly monitored but given 5 minutes to explore the maze without stimulation, 5 minutes with stimulation (20 Hz, 473 nm, 10 mW) and an additional 5 minutes without stimulation (Ethovision, Noldus, Amsterdam).
Marble burying
12 marbles were placed on a 5 cm deep layer of corn cob bedding in a standard size mouse cage (39×20×16 cm) in a grid-like fashion. Mice were then placed in the cage for 30 minutes and the degree of marble burying was hand-scored. If a marble was more than ½ way buried it was considered buried. The experimenter was blinded to the mouse treatment prior to the experiment.
Novelty-suppressed feeding
Mice were singly-housed a week prior to testing. 48 hours prior to testing, animals were allowed to consume a fruit-loop in their homecage. Food was then removed from the homecage for 24 hours. Mice were then placed in a corner of an open field (26.7×48.3 cm) at the center of which was a fruit loop on filter paper. Latency to feed was measured as the time required for the mouse to begin to consume the fruit loop. If the mouse had not approached the fruit loop after 10 min, it was removed from the open field and scored as 10 min. Immediately following, the mouse was returned to its homecage and allowed to freely consume fruit loops for 10 min. If the mouse did not consume any fruit loops in the cage, it was not included for this measurement.
Statistical Analysis
Data are presented as mean ± SEM. Significance is presented as *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. All statistical analyses were performed using GraphPad Prism version 6.02 for Windows, GraphPad Software, La Jolla California USA, www.graphpad.com.
Data was first tested for normality using the D’Agostino-Pearson test. Where data was not normal, a Wilcoxon matched-pairs test was performed. If data was normal, a Student’s t-test, paired t-test or matched 2-way ANOVA was performed where appropriate. If a significant interaction was detected in the 2-way ANOVA, a post hoc Bonferroni test was performed.
One NTSCeA::eYFP (control) animal was removed from the caspase drinking studies due to extremely low alcohol consumption, never reaching higher than a 2.1 g/kg average per week, and its preference for alcohol was greater than 2 standard deviations from the mean for control animals. One NTSCeA→PBN::ChR2 was removed from the water-drinking phenotyper experiment. Stimulation-day drinking for this mouse was a ROUT outlier from all other drinking days (stim and non-stim, NTSCeA→PBN::ChR2 and NTSCeA→PBN::eYFP).
Results
Neurotensin neurons in the central amygdala express a variety of markers
We first explored how Nts-expressing neurons overlap with other previously described genetically-defined populations in the central amygdala (CeA). Using dual fluorescent in situ hybridization (FISH) across the entire CeA (approx. bregma −0.8 to −1.9 mm), we found that CeA-NTS neurons are almost exclusively a subpopulation of protein kinase c delta (PKCδ) and corticotropin-releasing factor (CRF) expressing neurons (Supplementary Figure 1A-B), two populations that have been reported to have limited overlap (Cai et al, 2014). CeA-NTS neurons also express CRF-receptor 1 mRNA (Crfr1; Supplementary Figure 1C). Additionally, approximately two-thirds of CeA-NTS labeled neurons also express preprodynorphin mRNA (Ppdyn; Supplementary Figure 1D), the precursor of the endogenous ligand of the kappa opioid receptor (Chavkin et al, 1982). As CRF and dynorphin systems in the CeA both play a role in ethanol consumption (Lowery-Gionta et al, 2012; Anderson and Becker, 2017), these data suggested that manipulating the CeA-NTS subpopulation may also influence ethanol-related behaviors.
Ablation of neurotensin neurons in the central amygdala decreases ethanol consumption
To determine if CeA-NTS neurons play a role in ethanol-related behavior, we utilized NTS-ires-cre-recombinase (NTS-cre) mice (Leinninger et al, 2011) in conjunction with viral manipulations in the CeA. First, we validated the fidelity and penetrance of Cre in the CeA of this line. Using FISH (Supplementary Figure 1F), we double-labeled Nts and Cre mRNA in CeA slices from 5 separate NTS-Cre mice. We found that 61.4% of Nts mRNA-expressing cells (241.2 ± 29.7 Nts+ cells per slice) also expressed Cre (145.4 ± 23.7 Nts+Cre+ cells per slice,) and we found that 82.2% of Cre mRNA-expressing cells (173.2 ± 22.8 Cre+ cells per slice) also expressed Nts mRNA.
Next, we injected a cre-dependent virus encoding a modified Caspase 3 and TEV protease (AAV5-Ef1a-Flex-taCasp-TEVp; (Morgan et al, 2014; Yang et al, 2013)) into the CeA of NTS-ires-cre mice to selectively lesion CeA-NTS neurons (NTSCeA::casp, Figure 1A). This strategy resulted in a 51.7% reduction in NTS-positive cells in the CeA (measured by FISH; wild-type: 171.9 ± 3.4 cells, caspase: 83.0 ± 10.0 cells) and a 40.9% reduction in CeA-NTS immunoreactivity (ir; wild-type: 43.3 ± 3.2 a.u., caspase: 25.6 ± 1.3 a.u.), without altering NTS-ir in the neighboring lateral hypothalamus (Figure 1B-E). This lesion of CeA-NTS neurons did not alter locomotor behavior, anxiety-like behavior, body weight, or ethanol metabolism, when compared to controls that were injected with cre-dependent eYFP (NTSCeA::eYFP, Supplementary Figures 2-3).
Due to the importance of the CeA in ethanol consumption and reward (Gilpin et al, 2015), we next investigated if the loss of CeA-NTS neurons would alter voluntary ethanol consumption in a continuous 2-bottle choice paradigm. NTSCeA::casp mice showed significant decreases in ethanol consumed in 24-hour 2-bottle choice drinking when compared to NTSCeA::eYFP (Figure 1F; Two-way ANOVA: interaction, F(2,42)= 6.340, p=0.0039; ethanol concentration, F(2,42)=98.23, p<0.0001; ablation, F(1,21)=16.52, p=0.0006), with no effect of preference for the ethanol bottle (Figure 1G; Two-way ANOVA: interaction, F(2,42)=1.793, p=0.1790; ethanol concentration, F(2,42)=7.727, p=0.0014; ablation, F(1,21)=3.283, p=0.0843). NTSCeA::casp animals also showed decreased liquid consumption at lower ethanol concentrations, which was driven by increased total drinking by the NTSCeA::casp mice at lower ethanol concentrations (Figure 1H; Two-way ANOVA: interaction, F(2,42)=6.551, p=0.0033; ethanol concentration, F(2,42)=47.02, p<0.0001; ablation, F(1,21)=9.208, p=0.0063). To determine whether NTSCeA::casp mice showed differences in liquid consumption compared to controls, we measured water drinking over 5 days. NTSCeA::casp mice drank the same levels of liquid as NTSCeA::eYFP mice (Figure 1I; Two-way ANOVA: interaction, F(4,44)=2.459, p=0.0593; day, F(4,44)=2.714, p=0.0418; ablation, F(1,11)=1.005, p=0.3377), confirming that this manipulation affects ethanol consumption as opposed to general liquid consumption.
We next examined whether ablation of NTSCeA neurons would still impact alcohol consumption, and perhaps preference, in a drinking paradigm with higher levels of intake. We used an intermittent access (IA) drinking paradigm in an attempt to engender higher levels of drinking. In this paradigm, animals have access to both a 20% ethanol bottle and a water bottle on Monday, Wednesday, and Friday, with water only on other days over the course of 7 weeks. NTSCeA::casp mice again showed significant decreases in ethanol consumed across all weeks as compared to NTSCeA::eYFP (Figure 2A; Two-way ANOVA: interaction, F(6,126)=0.4321, p=0.8564; week, F(6,126)=2.539, p=0.0235; ablation, F(1,21)=11.19, p=0.0031) as well as cumulative ethanol consumption (Figure 2B; Two-way ANOVA: interaction, F(20,380)=13.53, p<0.0001; day, F(20,380)= 194.5, p<0.0001; ablation, F(1,19)= 11.69, p=0.0029. Bonferroni-corrected post-hoc test shows significant difference between NTSCeA::casp and NTSCeA::eYFP at days 26 through 47). At the same time, total liquid consumed was unaffected whether measured by week (Figure 2C; Two-way ANOVA: interaction, F(6,126)=1.525, p=0.1752; week, F(6,126)=8.358, p<0.0001; ablation, F(1,21)=0.00005215, p=0.9943) or cumulatively (Figure 2D; Two-way ANOVA: interaction, F(20,420)=0.1298, p>0.9999; day, F(20,420)=861.7, p<0.0001; ablation, F(1,21)=0.01703, p=0.8976). NTSCeA::casp mice also showed significantly decreased preference for the ethanol bottle (Figure 2E; Two-way ANOVA: interaction, F(6,126)=0.7778, p=0.588; week, F(6,126)=3.992, p=0.0011; ablation, F(1,21)=15.88, p=0.0007). Lastly, we compared total amount consumed at the end of the 7 weeks of IA. NTSCeA::casp mice consumed significantly less total ethanol than NTSCeA::eYFP mice (Figure 2F; Unpaired t-test t(21)=3.413, p=0.0026), with no detectable difference in total liquid consumed (Figure 2G; Unpaired t-test: t(21)=0.04085, p=0.9678). These experiments suggest that CeA-NTS neurons have a role in models of both lower and higher levels of alcohol consumption.
CeA-NTS neurons send a dense projection to the parabrachial nucleus (PBN)
To begin to examine the targets of CeA-NTS neurons, we injected a cre-dependent virus expressing channelrhodopsin-2 tagged with eYFP (ChR2-eYFP) into the CeA of NTS-ires-cre mice (Figure 3A-B). Using whole-cell ex vivo slice electrophysiology, we found that 473 nm light stimulation (20 Hz, 5 ms pulse, LED, Thor Labs) evoked action potentials in these cells (Supplementary figure 4A). We observed a robust projection from CeA-NTS neurons to the hindbrain near the 4th ventricle with fluorescence expression restricted to the PBN and the lateral edge of the locus coeruleus (LC, Figure 3C). We found significantly greater fluorescence expression in the PBN versus the LC (Figure 3D; Unpaired t-test: t(6)=14.59, p<0.0001), but since LC neurons extend long dendritic processes (Swanson, 1976), we next determined where CeA-NTS neurons make functional synaptic connections within the hindbrain. Monosynaptic input was isolated in whole-cell patch clamp recordings with TTX (500 µM) and 4-AP (100 mM). 473 nm light stimulation of CeA-NTS terminals induced an optically-evoked inhibitory post-synaptic current (IPSC) in both the medial and lateral PBN which was blocked by the GABAA receptor antagonist gabazine (10uM), while no inhibitory or excitatory synaptic currents were observed in the LC (Figure 3E-F, Supplementary figure 4B), similar to what was observed with the CeA-CRF projections (McCall et al, 2015). These data suggest that the CeA-NTS neurons make functional inhibitory synaptic connections in the PBN (11 of 13 cells) but not the LC (0 of 10, n=6 mice). We also observed a weaker fluorescent projection to the dorsal and ventral bed nucleus of the stria terminalis (BNST) where 5 of 9 cells (2 dorsal, 3 ventral) exhibited an optically evoked IPSC (Supplementary figure 5A-B), as well as local connections in the CeA (Supplementary figure 4C representative trace). In order to determine whether CeA-NTS projections to the BNST and PBN originated at the same cell bodies, we injected the retrograde tracer Alexa-488 CTXb into the PBN, and Alexa-555 CTXb into the BNST of the same animal. We found minimal overlap between BNST- and PBN-projecting neurons (1.6%, Supplementary figure 5C-D) suggesting that these are distinct cell populations.
CeA-NTS projection to the parabrachial nucleus (PBN) is reinforcing
The PBN is engaged by ethanol (Thiele et al, 1996), regulates consummatory behavior (Douglass et al, 2017), and sends reciprocal connections to the CeA (Carter et al, 2013). Because of the robust projection from the CeA-NTS neurons to the PBN, we wanted to directly examine how stimulation of this pathway impacts behavior. Additionally, the manipulations of CeA-NTS neurons described above suggested that they may influence motivated behaviors. To investigate the impact of activation of this projection on behavior, we expressed cre-dependent ChR2-eYFP in CeA-NTS neurons and implanted fibers just dorsal to the PBN (NTSCeA→PBN::ChR2-eYFP, Figure 3A, 3G).
To probe if stimulation of the NTSCeA→PBN pathway altered affective valence, we examined response to photostimulation in the real-time place preference (RTPP) assay (Jennings et al, 2013b). Photo-stimulation of these fibers at 20 Hz induced a significant RTPP in NTSCeA→PBN::ChR2-eYFP mice, but not in NTSCeA→PBN::eYFP controls, suggesting that these neurons convey positive valence (Figure 3H-I; Unpaired t-test: t(25)=6.128, p<0.0001). Optical stimulation significantly increased the distance traveled on the stimulation-associated side of the apparatus (Supplementary figure 6A; Wilcoxon matched-pairs signed rank test: control W= −28, p= 0.2402; ChR2 W= −105, p=0.0001), without altering velocity (Supplementary figure 6B; Wilcoxon matched-pairs signed rank test: control W= −34, p= 0.1475; ChR2 W= 85, p=0.0052). NTSCeA→PBN::ChR2 mice also performed optical intracranial self-stimulation (oICSS) for 40 Hz stimulation of this projection (Figure 3J; Wilcoxon matched-pairs signed rank test: control W= −20, p= 0.3379; ChR2 W= −89, p= 0.0005). Similar to the lack of effect on anxiety-like behavior noted with NTSCeA::casp mice, optical activation of the NTSCeA→PBN::ChR2 pathway did not alter behavior in the elevated plus maze either in open arm entries (Supplementary figure 6C; Two-way ANOVA: interaction F(2,27)=0.01082, p=0.9892; stimulation, F(2,27)=0.1085, p=0.8976; virus type, F(1,27)= 0.4477, p=0.5091) or in time spent in the open arm (Supplementary figure 6D; Two-way ANOVA: interaction F(2,27)= 0.6265, p=0.5421; stimulation, F(2,27)= 3.034, p=0.0648; virus type, F(1,27)= 0.6867, p=0.4146), indicating that this pathway modulates reward-associated behavior but does not alter anxiety-like phenotypes.
Stimulation of the CeA-NTS projection promotes alcohol consumption
We next examined the impact of photostimulation on ethanol consumption in NTSCeA→PBN::ChR2 mice. Mice were habituated to Ethovision Phenotyper boxes (Noldus) over the course of 4 days for 3 hours each, with 6% ethanol (in bottle with attached Lick-O-Meter) and food available. We tested 6% ethanol because it was the dose where our mice exhibited the highest preference, and the first dose where there was a significant difference in the NTSCeA::casp mice in the 2-bottle choice drinking (Figure 1F-G). Over the following 4 days, mice were placed in the same boxes, again with 6% ethanol and their standard mouse chow, and either received optical stimulation across 3 hours (20 Hz, 5 min on-off cycles, see Figure 4A), or received no stimulation (counterbalanced). Prior to and after each session, mice had ad libitum water and chow. NTSCeA→PBN::ChR2 and NTSCeA→PBN::eYFP mice showed similar levels of drinking during habituation days (Supplementary figure 7). We found that 473 nm, 20 Hz, 10 mW optical stimulation of the NTSCeA→PBN pathway increased consumption of 6% ethanol (Figure 4B; Wilcoxon matched-pairs signed rank test: control W=-20, p=0.4131; ChR2 W=61, p=0.0327) as compared to non-stimulation days, whereas stimulation of NTSCeA→PBN::eYFP mice did not alter ethanol consumption. On the days that the mice received stimulation, we found that significantly more ethanol consumption (number of licks) occurred during the 5-min laser on versus laser off phases (Figure 4C; Wilcoxon matched-pairs signed rank test: control W= −20, p= 0.4131; ChR2 W= 61, p= 0.0327). Surprisingly, in this experimental paradigm, NTSCeA→PBN::ChR2 mice decreased chow consumption on days when they received stimulation (Figure 4C; Two-way ANOVA: interaction F(1,22)=4.313, p=0.0497; virus type, F(1,22)=0.5391, p=0.4705; stimulation, F(1,22)= 7.387, p=0.0126). In summary, these data show that stimulation of the NTSCeA→PBN projection can promote ethanol consumption in freely-moving mice. We next sought to determine whether this increase in ethanol consumption and concurrent decrease in chow consumption was due to a generalized increase in liquid consumption, a generalized decrease in chow consumption, or an ethanol-specific phenotype. In mice given ad libitum food and water, we performed the same experimental paradigm as above (Figure 4A), but with water instead of ethanol. Stimulation of NTSCeA→PBN::ChR2 mice did not significantly alter water consumption (Figure 5A; Two-way ANOVA: interaction F(1,14)=0.9252, p=0.3524; virus type, F(1,14)=9.541e-005, p=0.9923; stimulation, F(1,14)=1.203, p=0.2913), licks to the bottle (Figure 5B; Wilcoxon matched-pairs signed rank test: control W=13, p=0.2188; ChR2 W=26, p 0.0781), or chow consumption (Figure 5C; Two-way ANOVA: interaction F(1,14)=0.5311, p=0.4782; virus type, F(1,14)=4.257, p=0.0581; stimulation, F(1,14)=1.218, p=0.2883. Together, these data demonstrate an alcohol-specific role for the NTSCeA→PBN projection in consumptive behaviors, disassociated from the CeA and PBN’s reported roles in general food and liquid consumption.
Discussion
The central nucleus of the amygdala (CeA) is well known to regulate several behaviors associated with alcohol use disorders. The particular genetically defined cell types and circuits that mediate these behaviors, however, are currently unknown. Here we have shown that NTS-expressing neurons in the CeA contribute to voluntary alcohol-consumption in mice. Additionally, our data demonstrate that a subset of these neurons project to the PBN, and that stimulation of this projection is positively reinforcing and leads to increased alcohol consumption. Surprisingly, this stimulation decreases food consumption, but only under conditions where the animal has the option to drink alcohol. Taken together, these data suggest a role for CeA NTS-containing neurons in the positively-reinforcing aspects of alcohol consumption at the expense of food consumption.
CeA neurotensin neurons do not engage anxiety-like behaviors
Our data confirm other reports that CeA-NTS neurons are a subpopulation of neurons that also express CRF, but differ in that we also saw that NTS neurons are a subpopulation of PKCδ mRNA-containing neurons (Kim et al, 2017). It is noteworthy that others have observed a small portion of CeA neurons that contain both CRF and PKCδ, which may comprise the NTS expressing cells (Haubensak et al, 2010; Cai et al, 2014). Here, we also report substantial overlap with cells containing mRNA for ppDyn. Intriguingly, neither ablation of these neurons nor their excitation altered any tests of anxiety-like behavior (Supplementary figures 3, 6) similar to what has been observed with cell body activation of PKCδ-containing neurons (Cai et al, 2014). Data examining the CRF projection to the LC (which is located just medial of the PBN), however, demonstrated that stimulation of this projection engaged anxiety-like behavior via CRF stimulation of CRF-receptorl (CRFR1), which then induced an increase in the firing rate of LC neurons. These discrepancies could be due to multiple factors: 1. The NTS cells are a subpopulation of CRF mRNA-containing neurons, and it is possible that expression of CRF mRNA does not mean that these cells release CRF protein; 2. Only CRF cells that lack the expression of NTS release CRF in proximity to LC neurons/processes that express CRFR1; or 3. That expression of Cre in these driver lines may target distinct neurons projecting to the hindbrain, as we observe a penetrance of 64% of NTS containing neurons. As CeA neurons expressing PKCδ also did not engage anxiety-like behavior and our data suggest that CeA-NTS neurons comprise the overlap between these two populations, it stands to reason that a subpopulation of CeA-CRF containing neurons do not engage anxiety-like behaviors.
CeA neurotensin neurons promote positive valence and reinforcement
In contrast to what was observed with the CRF-cre projection to the LC discussed above, stimulation of the NTSCeA→PBN resulted in behaviors associated with reward and positive valence. Other groups have shown that stimulation of the CeA can increase responding for a laser-paired positive reinforcer and even shift preference towards a non-preferred outcome when paired with stimulation (Robinson et al, 2014; Warlow et al, 2017). These experiments target the entire CeA, however, and animals do not show intracranial self-stimulation behavior. This has led to a hypothesis that the CeA has a role in amplifying motivation for reward-seeking, but does not have a direct role in reward in and of itself. In comparison, recent work demonstrated that mice will perform olCSS of CeA cell bodies in genetically-defined subpopulations of CeA neurons (Kim et al, 2017) and that stimulation of different genetically defined CeA→PBN projection can reinforce behavior (Douglass et al, 2017). These data are in line with our finding that animals will perform olCSS and demonstrate real-time place preference for stimulation of the NTSCeA→PBN projection (Figure 3J). We posit that stimulation of heterogeneous groups in the CeA may obscure the role of specific projections or genetically-defined subtypes, particularly when they may have reciprocal inhibitory connections.
The role of CeA neurotensin neurons in alcohol and food consumption
The CeA is well known to be engaged by alcohol consumption and is implicated in mediating both the negative and positive reinforcing properties of alcohol (Koob et al, 1998; Koob, 2015). In keeping with this, an earlier study from Hyytiä and Koob found that pharmacological inhibition of the CeA reduces ethanol consumption (Hyytiä and Koob, 1995) and amygdalar lesions can cause reductions in alcohol consumption (Möller et al, 1997). Our data show that selective lesioning of NTSCeA neurons can decrease alcohol intake and reduce preference for alcohol without altering other fluid consumption. Concordant with this finding, we found that optogenetic stimulation of the NTSCeA→PBN projection increased alcohol consumption during stimulation epochs, but again did not alter consumption of water. This data suggests that stimulation of this projection is related to engagement with the salient stimulus, as opposed to inhibiting the previously described PBN fluid satiation neurons, which would lead to increased overall fluid intake (Ryan et al, 2017).
Select populations of the CeA are known to alter the consumption of food. Activation of PKCδ populations reduces food intake (Cai et al, 2014), and in contrast, activation of serotonin 5-HT2A receptor-containing populations increase consumption of food (Douglass et al, 2017). In order to exclude that changes in alcohol drinking were due to alterations in general consummatory behavior, and to better mimic the drinking experiments performed in the NTSCeA::caspase mice, we measured food consumption during the optogenetic alcohol/water experiments. Interestingly, we observed that when alcohol was present, but not when water was present, stimulation of the NTSCeA→PBN projection decreased the consumption of food. These data suggest that the effect that we are observing on alcohol is not due to the mice seeking a caloric source or increased fluid, but may be selective for alcohol. Indeed, we hypothesized that the presence of alcohol altered the motivational state of the mice and shifted their desire to consume less food.
Cai et al. show that the PKCδ neurons send a projection to the PBN, but they only observed a small IPSC in 1 out of the 6 neurons recorded, while we observed IPSCs in 13 of 15 cells recorded. We hypothesize that this discrepancy may have been due to recording location. The dorsal-most aspect of the PBN does not appear to receive GABAergic input from the CeA-NTS neurons, as indeed we did not observe current in the 2 dorsal neurons we patched there, nor does it appear that the more ventrolateral aspect of the PBN (where the pro-anorectic calcitonin gene-related peptide-containing neurons reside) receives a projection from the CeA-NTS neurons. We observed current in 13 cells from the median aspect of both the medial and lateral PBN.
CeA neurotensin neurons, a potential role in salience
There is considerable evidence that stimulating non-specific CeA cell bodies contributes to an increase in the incentive salience associated with a reward (Mahler and Berridge, 2009; Robinson et al, 2014; Warlow et al, 2017). The data presented here suggests a more nuanced role for the NTSCeA→PBN pathway than simply driving or suppressing general consumptive states. Particularly, we find that the effects of stimulating the NTSCeA→PBN projection on food consumption are context- and choice-dependent. When mice are presented with the choice of drinking alcohol or eating food, stimulation of the NTSCeA→PBN projection increases alcohol consumption at the expense of food consumption. When mice are offered water and food, however, stimulation of this pathway fails to increase or decrease consumption of either substance. In addition, we did not observe differences in growth curves of the NTSCeA::casp mice, suggesting that ablation of this pathway does not result in enhanced chow consumption. Our results are additionally interesting because many of the pathways that drive decreases in food consumption are inherently aversive (Jennings et al, 2013a; Betley et al, 2015), but our results suggest the stimulation of the NTSCeA→PBN projection exhibits a positive valence and is reinforcing.
CeA to PBN pathway as a pathway for alcohol consumption phenotypes
Mounting evidence points to the amygdala as an important substrate for the actions of ethanol, particularly synaptic modulation of GABAergic signaling (Roberto et al, 2004; Herman et al, 2013). Much of this work, however, has focused on the consumption of alcohol as a means to alleviate an increase in anxiety-like behavior via negative reinforcement (Koob et al, 2014; Anderson and Becker, 2017). Interestingly, using the 20% ethanol IA paradigm, a study from the George lab has shown that discrete neuronal ensembles in the CeA are recruited during abstinence from ethanol, and lesioning these neurons transiently reduces ethanol intake in “non-dependent” rats, with alcohol consumption returning to control levels within a week (de Guglielmo et al, 2016). In contrast, our data demonstrate that lesioning the CeA-NTS population decreases drinking with no recovery to control levels. Furthermore, we identify a specific CeA output population, NTSCeA→PBN, and show that it promotes alcohol consumption and reward, without modulating anxiety-like behavior. In our experiments, we investigated the NTSCeA→PBN in animals with relatively brief alcohol experience (habituation days). Future studies will characterize the role of alcohol experience/dependence in modulating NTSCeA neurons, and their projections to the PBN.
Our data have demonstrated that NTS-expressing neurons in the CeA bi-directionally control ethanol consumption in mice. Lesioning these neurons decreases ethanol drinking and preference, while stimulating their projections to the PBN increases ethanol drinking in non-dependent mice. Interestingly, neither of these manipulations alters anxiety-like behavior, while stimulation of the NTSCeA→PBN projection promotes positive valence and reinforcement. Surprisingly, stimulation of this pathway reduces food consumption, but this is dependent on the motivational state of the animal. These data suggest that NTS-expressing CeA neurons, and especially their projection to the PBN, play a significant role in modulation of specific consumptive behaviors and may alter incentive salience.
Acknowledgements
The authors thank Drs. Thomas Kash and Elyse Dankoski for on previous versions of the manuscript. This work was supported by: K01 AA023555 (Z.A.M.), P60 AA011605 (C.W.H., G.D.S)