Chemogenetic inhibition of the bed nucleus of the stria terminalis suppresses the intake of a preferable and learned aversive sweet taste solution in male mice

Conditioned taste aversion (CTA) is established by pairing a taste solution as a conditioned stimulus (CS) with visceral malaise as an unconditioned stimulus (US). CTA decreases the taste palatability of a CS. The bed nucleus of the stria terminalis (BNST) receives taste inputs from the brainstem. However, the involvement of the BNST in CTA remains unclear. Thus, this study examined the effects of chemogenetic inhibition of the BNST neurons on CS intake after CTA acquisition. An adeno-associated virus was microinjected into the BNST of male C57/BL6 mice to induce the inhibitory designer receptor hM4Di. The mice received a pairing of 0.2% saccharin solution (CS) with 0.3 M lithium chloride (2% BW, intraperitoneal). After conditioning, the administration of clozapi- ne – N – oxide (CNO, 1 mg/kg) significantly enhanced the suppression of CS intake on the retrieval of CTA compared with its intake following saline administration (p < 0.01). We further assessed the effect of BNST neuron inhibition on the intake of water and taste solutions (saccharin, sucralose, sodium chloride, monosodium glutamate, quinine hydrochloride, and citric acid) using naïve (not learned CTA) mice. CNO administration significantly decreased the intake of saccharin and sucralose (p < 0.05). Our results indicate that BNST neurons mediate sweet taste and regulate sweet intake, regardless of whether sweets should be ingested or rejected. BNST neurons may be inhibited in the retrieval of CTA, thereby suppressing CS intake.


Introduction
Experiencing malaise, such as abdominal pain or nausea, after consuming a novel tasting food or drink induces conditioned taste aversion (CTA). CTA is acquired through association learning between a food's taste as a conditioned stimulus (CS) and malaise as an unconditioned response (UR). The neural mechanisms underlying the acquisition and generalization of CTA have been revealed by numerous previous studies, which showed the involvement of brain regions receiving both taste and visceral sensations, such as the nucleus of the stria tractus, parabrachial nucleus, amygdala, and insular cortex [1]. The critical findings for the neural mechanisms were caused by the irreversible inactivation of brain regions using electrolytic or excitotoxic lesions of neurons. Animals with irreversible inactivation of a pivotal brain region before the association between a CS and an UR (conditioning procedure) showed an increase in the amount of CS consumption compared to animals with the sham operation. Because measuring CS consumption is convenient, irreversible inactivation is a very useful technique for investigating whether a brain region is involved in CTA acquisition. In contrast, the findings about the neural mechanisms underlying the process in which a CS is avoided after CTA acquisition, referred to as the process of CTA retrieval, are relatively small. Irreversible inactivation before conditioning cannot be used to assess the neural mechanisms of CTA retrieval, because it may disrupt the acquisition process rather than the retrieval process. Thus, temporary inactivation of a brain region using drugs or genetic tools is appropriate for CTA retrieval. CTA is essential for survival because it enables humans and animals to avoid foods and drinks that are potentially contaminated with toxins, suggesting the existence of common neural mechanisms underlying CTA retrieval in humans and animals. Therefore, we attempted to identify the brain regions and neural networks involved in CTA retrieval using temporary inactivation techniques.
We previously found that the temporary inactivation of the basolateral amygdala (BLA) elevates the intake of CS during CTA retrieval [2]. We had also revealed that intraoral infusion of a CS augments the activity of the efferents projecting from the BLA to the nucleus accumbens (NAc), the bed nucleus of the stria terminalis (BNST), and the central amygdala (CeA) [3]. While previous studies have shown that CTA retrieval requires the involvement of the NAc [4][5][6][7] and CeA [8][9][10][11][12][13], whether the BNST is part of the neural mechanisms underlying CTA remains unclear.
The neurons in the BNST receive projections from neurons in the pontine parabrachial nucleus (PBN) [14,15]. The PBN receives taste and visceral sensations from the nucleus of tractus solitarius (NTS) [16][17][18][19]. The PBN neurons seem to transmit taste stimuli to the BNST because the bilateral lesion of the medial part of the PBN decreases the expression of c-Fos in BNST neurons induced by the sham drinking of sucrose solution (orally consumed but drained through gastric cannulas) [20]. Conversely, the BNST neurons project axons to the NTS and PBN [21][22][23][24][25][26], which express somatostatin and corticotrophin-releasing factor (CRF) [27,28]. Electrophysiological studies have shown that BNST neurons modify the gustatory responses of PBN neurons [29,30]. GABAergic neurons in the BNST project axons to the pre-locus coeruleus and alter the responses to NaCl in animals under sodium deficiency [31]. These findings suggest that BNST neurons mediate the gustatory information.
Additionally, BNST neurons may be involved in visceral sensation because the intraperitoneal administration of a malaise-inducing substance, lithium chloride, induces c-Fos expression in the BNST [32]. The neurons in the caudal NTS, which receive visceral sensation, densely project to the BNST [33]. Neurons projecting from the NTS or PBN to the BNST express Fos after administering the anxiogenic drug yohimbine [34]. The BNST receives projections of neurons positive for calcitonin gene-related peptides in the PBN [35]. The stimulation of CGRP-positive PBN neurons projecting to the BNST after drinking sucrose solution induces the establishment of CTA [36]. Taken together, the BNST receives both taste and visceral sensations, which are critical associative components of CTA. However, several previous studies have provided negative evidence regarding the involvement of BNST in CTA acquisition [37][38][39][40]. Additionally, no report has demonstrated the involvement of the BNST in CTA retrieval. Therefore, to assess the possibility of involvement of the BNST in CTA, this study aimed to investigate the effect of the inhibition of BNST neurons on CS intake during CTA retrieval.
CTA is thought to cause a decrease in palatability because a conditioned animal exhibits disgusted reactions to a CS [41]. CTA also likely produces a conditioned fear of a CS because it signals danger to animals, especially species that cannot vomit (see Parker's review [42]). This hypothesis is supported by the findings of our previous study, which showed that conditioned rats developed not only disgust but also fear [2]. The BNST plays an important role in fear and anxiety states [43]. Thus, the manipulation of BNST neurons purportedly affects disgust and fear exhibited toward a CS. A novel environment induces anxiety or fear responses when animals are artificially forced to challenge novel stimuli [44]. Taken together, we hypothesized that BNST neurons may participate in CTA when animals experience a conditioning procedure (CS-US pairing) in an unfamiliar environment (not the home cage environment). Therefore, in this study, mice were presented with fluid in a test chamber other than their home cage.

Subjects
Twenty male C57BL/6 mice (CLEA Japan, Inc., Osaka, Japan), [8][9] weeks old at the time of arrival, served as subjects. They were individually housed in transparent cages in a colony room under standard environmental conditions (23 ± 2 • C; 12-hour light-dark cycle; lights on at 7:00 am). Water and standard chow (CE-2, CLEA Japan, Inc., Osaka, Japan) were available ad libitum, except for when a restricted schedule was specified. All experiments were approved by the Hokkaido University Animal Care and Use Committee (#17-0045) and Hokkaido University Safety Committee on Genetic Recombination Experiments (#2019-008) and were performed in accordance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health Guide).

CTA test
After at least two weeks of recovery from the surgery (Table 1), the mice were habituated to a test chamber for 15 min under non-waterdeprived conditions (day 1). The test chamber consisted of a transparent plastic cage and a wire lid identical to that used for the home cage. Each mouse was placed in a test chamber. The home cages were equipped with bedding material (Japan SLC, Inc., Shizuoka, Japan) and The mice in each test chamber were placed at the same place on a rack in the colony room. Mice were allowed to drink deionized water (DW) using a syringe for 15 min. After leaving the mice in the test chamber for 3 h and 45 min, they were returned to their home cage and given supplemental water for 2 h. The mice were allowed free access to chow in their home cages. The water training session was conducted for six days. In the following conditioning session, the mice were presented with 0.2% saccharin solution in a syringe as the CS instead of DW. Immediately after the presentation, the mice were administered 0.3 M lithium chloride (2% BW i.p.) as the US. After 3 h and 45 min, the mice were returned to their home cage and given water and chow ad libitum. We presented the mice with water using a syringe in the test chamber two days after conditioning to confirm normal consumption behavior. Subsequently, the mice were presented with the CS in three tests (acquisition and retrieval) every alternate day. The acquisition test aimed to examine the establishment of the CTA. We found that all mice had a decreased intake of CS in the acquisition test compared to that in the conditioning test ( Fig. 1). Half of the mice were assigned to the experimental group, and the remainder to the control group. CS intake in the acquisition test was equivalent between the groups as closely as possible. To inhibit BNST neurons, the experimental group was intraperitoneally injected with clozapine-N-oxide (CNO; Enzo Life Sciences, Inc., NY, USA; 1 mg/kg; 0.5% BW). Conversely, the control group was intraperitoneally injected with a vehicle (0.9% sterile saline; 0.5% BW). Thirty minutes after the injection, the rats were presented with CS for 15 min (first retrieval test). To examine whether the CNO injection altered the animal's behavior temporarily, CS intake was measured again 2 days later with no drug injection (second retrieval test). All fluid presentations on training, conditioning, and tests were conducted at noon after 18 h of water deprivation.
One week after the second retrieval test, we assessed the effect of CNO injection on baseline consumption behavior. Half of the mice received an injection of CNO in the first session and the vehicle in the second session two days later. The remaining mice first received an injection of the vehicle and then CNO. Thirty minutes after the injection, they were presented with DW. Mice were deprived of water for 18 h, identical to the CTA test.

Intake test in naïve animals
The CTA test and subsequent water intake test revealed that the injection of CNO decreased the intake of CS and water, suggesting that the inhibition of BNST neurons suppresses fluid consumption behavior.
To assess this, we investigated the effect of chemogenetic inhibition of BNST neurons on fluid intake in naïve (non-conditioned) mice. Twelve new male mice received microinjections of a virus identical to that used in the CTA test. However, we did not use the test chamber to eliminate the effect of environmental changes on fluid consumption behavior. The mice were trained to drink water from a syringe in their home cage for six days. Two days after the final training, the mice were injected with CNO or vehicle 30 min before the presentation of water. The order of the drug injections was counterbalanced as well as the water intake test mentioned above: half of the mice (n = 6) firstly received the CNO and then the vehicle. Conversely, the remaining the mice (n = 6) first received the vehicle and then CNO. We administered two doses of CNO:0.2 mg/kg and 1.0 mg/kg. As each dose of CNO was paired with the vehicle (0.9% saline), four water tests were performed every two days. Two days after the last session, we examined the effect of CNO injection on the intake of the taste solutions: 0.2% saccharin sodium, 10 mM sucralose, 0.075 M sodium chloride (NaCl), 0.3 M monosodium glutamate (MSG), 0.5 mM quinine hydrochloride (QHCl), and 20 mM citric acid. These concentrations were based on previous studies (saccharin [46], sucralose [47], NaCl [48], MSG [49], QHCl [50], and citric acid [51]). They revealed that mice show preferences for the sweeteners, NaCl and MSG, and in contrast, aversions to QHCl and citric acid. The concentrations of QHCl and citric acid were not high because we aimed to avoid a ceiling effect, which could mask the effect of the CNO injection.
Similar to the water intake test, the mice were presented with each taste solution 30 min after the injection of CNO (vehicle) in the first session and vehicle (CNO) in the last session. There were two days of no test between the first and last session. To minimize potential neophobia, the mice were allowed to experience each taste solution two days before the first session.

Histological analysis
After the completion of the behavioral experiments, the mice were deeply anesthetized with an overdose of pentobarbital sodium. They were transcardially perfused with 0.02 M phosphate-buffered saline (PBS; pH 7.4) followed by 4% paraformaldehyde in 0.1 M phosphate buffer. The brains were removed from the skull, post-fixed in paraformaldehyde solution overnight, and then immersed in 30% sucrose in 0.1 M PB for cryoprotection. Sections (30 µm) were prepared using a sliding microtome. The tissue sections were collected and stored at − 20 • C in cryoprotectant solution until immunohistochemical staining. Fig. 1. Images of a brain slice of an animal. Thirty μm-thick slices were immunohistochemically stained for mCherry (a marker of hM4Di receptors) and NeuN (neuronal marker). We observed the co-localization of mCherry and NeuN (blue arrowheads) in the dorsal and ventral bed nucleus of the stria terminalis (BNST), suggesting that the BNST neurons express hM4Di.
After incubation in a blocking solution (5% normal goat serum, 0.05% Triton-X, and 1x PBS) for 2 h at room temperature, free-floating tissue sections were incubated with a rabbit polyclonal antibody against red fluorescent proteins (1:5000; Medical & Biological Laboratories Co. Ltd., Tokyo, Japan; PM005) overnight at 4 • C. The sections were incubated in a solution of Alexa 594-conjugated anti-rabbit IgG (1:200; Jackson ImmunoResearch Laboratories Inc., PA, USA; 111-585-003) for 3 h at room temperature. Subsequently, the sections were incubated in the blocking solution for 1 h at room temperature and then reacted with a mouse monoclonal antibody against NeuN (1:5000; Sigma-Aldrich Japan K.K., Tokyo, Japan; MAB377) for 72 h at 4 • C. Finally, the sections were incubated with Alexa 488-conjugated anti-mouse IgG (1:200; Jackson ImmunoResearch Laboratories Inc., PA, USA; 115-545-003) for 3 h at 23-27 • C. Each antibody was diluted in blocking solution. Before and after the reactions with the antibodies, the sections were washed for 10 min at least thrice in PBS. The stained sections were mounted on coated glass slides. The glass slides were coverslipped with mounting medium (Mounting Medium with DAPI; Aqueous, Fluoroshield; Abcam, Cambridge, UK; ab104139).
Tissue sections were viewed under a microscope (ECLIPSE Ci; Nikon Solutions Co. Ltd., Tokyo, Japan). To estimate the extent of the regions, including neurons expressing hM4Di, digital images of mCherry-positive cells and neuronal marker NeuN-positive cells and their merged images were captured and created using a digital camera (DS-Fi3; Nikon Solutions Co. Ltd., Tokyo, Japan) with appropriate filter sets and software (NIS-ElementsD; Nikon Solutions Co. Ltd., Tokyo, Japan).

Statistics
Mice whose hM4Di-expressing neurons were not observed in either or both of the bilateral BNST regions were excluded from the behavioral data analysis, which was performed using Prism (GraphPad Software, CA, USA). To estimate the robustness of CTA, the CS intake of each group in the retrieval tests was compared with that in the conditioning test using paired t-tests. The effect of the CNO injection was assessed by comparing the group differences in the CS intake in the retrieval tests using a multiple unpaired t-test corrected by the Holm-Šídák method. In other tests, the mean intake after CNO injection was compared with that after vehicle injection using a paired t-test (water intake test after the CTA test) and multiple paired t-tests corrected by the Holm-Šídák method (water and taste solution tests using naïve animals).

Histology
We confirmed the expression of hM4Di in BNST cells by observing the expression of the red fluorescent protein, mCherry. mCherry-positive cells were observed in both the dorsal and ventral parts of the BNST in the experimental (n = 6) and control (n = 7) groups (Fig. 1). The other four experimental and three control mice showed mCherry-positive cells outside the BNST. We excluded these mice from the behavioral data analysis.

CTA experiment
The intake of CS in the acquisition test (0.139 ± 0.04 g; mean ± standard error) was significantly lower than that of the conditioning (0.7995 ± 0.062 g) ( Fig. 2a; paired t-test; t[12] = 8.142, p < 0.001). These results indicated that the mice acquired CTA. There was no group difference in CS intake in the conditioning (control, 0.789 ± 0.048 g; experimental, 0.812 ± 0.1295 g) and acquisition tests (control, 0.145 ± 0.057 g; experimental, 0.132 ± 0.062 g). These results suggest that the strength of aversion in both the groups was comparable.
In the first retrieval test, the experimental group received an intraperitoneal injection of CNO 30 min before the CS presentation. The comparison using a paired t-test showed that each group consumed significantly less CS in the first retrieval test than in the conditioning test ( Fig. 2b; control, t[6] = 4.246, p < 0.01; experimental, t[5] = 5.851, p < 0.01). These results suggest that both groups displayed CTA during the first retrieval test. However, a group difference in the robustness of the CTA was observed, potentially because the CS intake in the experimental group (0.062 ± 0.032 g) was significantly lower than that in the control group (0.369 ± 0.065 g) (multiple unpaired t-test; t[8.74] = 4.237, p < 0.01). Thus, the CNO injection likely enhanced the suppression of CS intake in the conditioned animals.
The second retrieval test aimed to examine whether enhanced suppression of CS intake was shown even two days after CNO injection. Each group drank significantly smaller CS in the second retrieval test than in the conditioning test (paired t-test; control, t[6] = 2.633, p < 0.05; experimental, t[5] = 2.883, p < 0.05) (Fig. 2c). We found no group differences in CS intake (control, 0.481 ± 0.082 g; experimental, 0.319 ± 0.109 g). These results indicate the absence of a lasting effect of CNO injected in the first retrieval test.

Effect of CNO injection on water intake in the conditioned animals
The CNO-induced suppression of CS intake in the first retrieval test raised the question of whether inhibition of BNST neurons might depress consummatory behavior. To address this issue, we assessed the effect of CNO injection on the water intake in the mice used for the CTA experiment. Water intake after CNO injection was significantly lower than that after vehicle injection ( Fig. 3; CNO, 0.491 ± 0.042 g; vehicle, 0.621 ± 0.036 g; paired t-test, t[12] = 2.604, p < 0.05). However, the conditioning experienced by these mice in the CTA test was not negligible because the learning process could induce any plastic changes in the activity of BNST neurons. Therefore, in the subsequent experiment, we used naïve animals to investigate the effect of BNST neuron inhibition on water intake.  , c). b, The CNO enhanced the suppression of the CS intake in the first retrieval test. c, No significant group difference was observed in the CS intake in the second retrieval test (two days after the first test). Cond., conditioning; Acq., acquisition test; CON, control group; EXP, experimental group. * **p < 0.001; †p < 0.05, † †p < 0.01, † † †p < 0.001 (vs. the CS intake of each group in the conditioning).

Effect of CNO injection on water intake behavior in naive animals
The mice injected with the virus were trained to drink water during a 15-min session in their home cages. They then received an injection of CNO or vehicle injections before the presentation of water. We used two doses of CNO (0.2 and 1 mg/kg, i.p.). The water intake after 0.2 mg/kg of CNO injection (0.946 ± 0.089 g) was not significantly different from that after vehicle injection (1.032 ± 0.115 g) (Fig. 4). No significant difference was observed in the comparison between 1.0 m/kg of CNO (0.8111 ± 0.074 g) and vehicle (0.882 ± 0.088 g). These results indicate that neither dose of CNO altered water intake.

Control experiment: the effect of CNO injection on the CS intake in sham-operated mice
Recent studies have shown that it is difficult for CNO to cross the blood-brain barrier, and its metabolite, clozapine, binds to endogenous receptors. Therefore, to examine whether the CNO injection affected the retrieval of CTA without the expression of hM4Di in BNST neurons, an additional experiment was performed to measure CS intake after CNO injection in sham-operated animals that were microinjected with a control AAV (not inducing hM4Di). The water intake on the final training day in the sham-operated animals (0.558 ± 0.093 g) was lower than that in the control and experimental groups in the CTA test (0.924 ± 0.062 g and 1.083 ± 0.208 g, respectively). One-way analysis of variance (ANOVA) showed a significant main effect of group (F[2,26] = 4.087, p < 0.05). Thus, we calculated the intake ratios by dividing the intake during the conditioning and tests by that on the final training day. A one-way ANOVA revealed a significant main effect of the group in the first retrieval test ( Fig. 6; F[2,16] = 6.747, p < 0.01), but not in the conditioning and other tests. Post hoc comparison using Tukey's HSD test in the first retrieval test revealed that the intake ratio in the experimental group significantly differed from that in the control and sham groups (p < 0.01 and p < 0.05, respectively), while the sham and control groups did not significantly differ. These results indicate that the injection of CNO did not alter CS intake without hM4Di receptor expression in BNST neurons.

Discussion
This study aimed to determine the involvement of BNST in the retrieval of CTA images. Conditioned animals with chemogenetically inhibited BNST neurons received a presentation of the CS in the first retrieval test. The CNO-injected mice showed a significantly lower CS intake than the vehicle-injected mice, suggesting that inhibition of BNST neurons suppressed the consumption of CS. No significant group difference in the second retrieval test performed two days after the first test indicated that the suppressive effect of the CNO injection was temporary. The water intake in the conditioned animals significantly decreased after CNO injection, but in naïve (non-conditioned) mice, the inhibition of BNST neurons did not alter the water intake. We also found that inhibition of BNST neurons decreased the intake of saccharin and sucralose, but not NaCl, MSG, QHCl, and citric acid.
The mice that did not show the expression of hM4Di receptors in the bilateral BNST were excluded from the behavioral data analysis. To assess whether the stimulation of the hM4Di receptors outside the BNST by the CNO injection affects the consumption behavior, we analyzed the difference in the CS intake in the first retrieval test between the control and the excluded mice (n = 4) in the CTA experiment using an unpaired t-test. We found no significant group difference (t[9] = 0.354, p = 0.731). The analysis of the excluded data in the experiments using the naïve mice showed no significant drug effects on saccharin (t[4] = 3.18, p = 0.157) and sucralose (t[4] = 2.682, p = 0.168). These results indicate that the suppressed CS intake in the experimental group in the CTA experiment and the decreased intake of saccharin and sucralose in the naïve mice were due to the manipulation of the BNST neurons by the CNO injection.
Since the BNST plays a pivotal role in anxiety and fear states [43], we assumed that CS intake may increase following the injection of CNO, which induces the inhibition of BNST neurons via hM4Di receptors. However, the CNO-injected mice exhibited suppressed CS intake, suggesting that inhibiting the BNST neurons does not rescue the suppression of CS intake caused by the acquisition of CTA. Another unexpected result  was that the inhibition of BNST neurons reduced the intake of artificial sweeteners (saccharin and sucralose) because no previous studies have shown the involvement of BNST neurons in the intake of taste solutions. Thus, to our knowledge, this is the first report suggesting that BNST neurons mediate sweet tastes. The present study did not use natural sugars, such as sucrose, to avoid associating post-ingestion effects with the intraperitoneal injection of CNO. Therefore, whether BNST neurons play a role in the ingestive behavior of natural sugars remains unclear. Future studies should address this issue.
BNST neurons project to the NTS and the PBN, which are gustatory relay nuclei in the brainstem [21,22,27,52]. Electrical stimulation of BNST neurons exclusively inhibits taste-responsive neurons in the PBN [29]. While PBN-projective neurons in the BNST are rarely immunoreactive with anti-glutamic acid decarboxylase [52], they are positive for corticotrophin-releasing hormone (CRH) and somatostatin [27,28]. Thus, the descending inhibitory modulation from the BNST to the PBN may be mediated by CRH and somatostatin, rather than by the inhibitory neurotransmitter GABA. Although it is unclear whether the hM4Di-positive neurons in the BNST in the present study were positive for CRH and somatostatin, CNO injection might have disinhibited the taste-responsive neurons in the PBN. PBN neurons exhibit more significant electrophysiological responses to aversive CS than to preferable tastes [53]. The presentation of a CS after CTA acquisition increases the expression of the neural excitation marker Fos protein in the PBN [11,54,55]. These findings suggest that the inhibition of BNST neurons may enhance the retrieval of CTA by disinhibiting PBN neurons.
A previous study revealed that the presentation of a preferable sucrose solution does not change neural activity in the PBN [53]. Therefore, the decreased intake of saccharin and sucralose by the CNO injections in this study was likely not due to modification of the PBN neurons. The BNST neurons also project to the ventral tegmentum area (VTA) [24]. Dopaminergic (DA) neurons in the VTA are deeply involved in the ingestive behavior of preferable taste solutions. The intake of sucrose solution elevates extracellular dopamine levels in the nucleus accumbens, which is a major target of VTA DA neurons [56,57]. The ablation of dopaminergic neurons in the VTA decreases sucrose consumption [58][59][60]. Conversely, lesion or chemogenetic silencing of GABAergic neurons in the VTA increases the preference for sucrose [61]. Since the stimulation of VTA GABAergic neurons decreases sucrose intake [62], the DA neurons involved in sucrose intake are purportedly inhibited by GABAergic interneurons, as advocated by Jennings et al. (2013) [63]. The projective neurons from the BNST to the VTA are mainly GABAergic and innervate GABAergic interneurons in the VTA [64,65]. These findings strongly suggest that the inhibition of GABAergic BNST-VTA projective neurons disinhibits GABAergic interneurons in the VTA, resulting in the inhibition of DA neurons. The inhibition of GABAergic BNST-VTA also decreases binge-like ethanol intake [66]. Thus, in the naïve animals of the present study, the inhibition of BNST neurons via hM4Di receptors might induce silencing of VTA DA neurons, resulting in decreased intake of sweet taste solutions (saccharin and sucralose).
CNO injection decreased the water intake in the mice used in the CTA experiment (Fig. 2). Since BNST neurons can modify VTA DA neurons as described above, we assumed that the decreased water intake might have been produced by a temporal decline in appetitive and consummatory behaviors. However, the water intake test results using naïve mice did not support this idea, because the inhibition of BNST neurons did not alter their water intake (Fig. 3). One study showed that silencing of projective neurons from the subfornical organ to the ventral part of the BNST did not alter water consumption in salt-deficient mice [67]. To the best of our knowledge, no reports have identified the involvement of the BNST in water consumption. Therefore, under water-restricted conditions, BNST neurons may not be involved in regulating body fluid balance and water intake behavior.
The fact that the CNO injection decreased water intake only in the animals used for the CTA experiment suggests that inhibition of BNST neurons might enhance anxiety and fear in the test chamber because the mice experienced the conditioning procedure there. We used a test chamber with a wire mesh floor to prevent the mice from eating food spillage and feces before and during the fluid presentation. We also assumed that the context of the fluid presentation may be associated with a negative experience (malaise), such as contextual fear learning, because mice can discriminate the features of the floors of the test chamber in a conditioned place preference/aversion learning paradigm   6. The intake ratio in the first retrieval test. To calculate the intake ratio of the control (CON) and experimental (EXP) in the CTA experiment, the intake in the tests (shown in Fig. 2) was divided by the intake in the final training. These data were compared with the intake of the group that received sham surgery (SHAM). * *p < 0.01, *p < 0.01. [44]. Therefore, the decreased water intake following the CNO injection suggests that the inhibition of BNST neurons retrieved anxiety or fear in the chamber. This idea is supported by a recent study that investigated the role of GABAergic and glutamatergic projections from the BNST to the PBN on feeding and anxiety-related behaviors [68]. They showed that the calcium activity of PBN-projecting GABAergic neurons is decreased by the presentation of threats. They also found that inhibition of these neurons decreases food intake. These findings raise the possibility that the CNO injection in the present study may inhibit GABAergic projective neurons to the PBN, inducing decreased water intake in a potentially threatening environment (the test chamber). However, because the single-bottle test used in the present study did not measure any indices of anxiety or fear, we will focus on this issue in future studies.
CNO injection did not significantly reduce the intake of 0.5 mM QHCl and 20 mM citric acid. The CS in the present study (saccharin) was a learned aversive, while quinine and citric acid were naturally aversive. Both aversive stimuli are rejected by animals but might be differentially processed in the brain. We had previously revealed that GABAergic transmission in the ventral pallidum mediates both naturally preferable and learned aversive saccharin, but not naturally aversive or learned extremely aversive quinine [69,70]. The VTA DA neurons also seem to respond to a preferable taste but not an aversive one because the lesion of the VTA by injecting 6-OHDA suppresses the intake of sucrose, but not QHCl and HCl [59]. Thus, BNST neurons mediate only naturally preferable tastes.
The taste solution intake test demonstrated that the inhibition of BNST neurons did not alter the intake of 0.075 M NaCl. Since a higher concentration of NaCl is aversive for animals [71], we selected 0.075 M by referring to a study showing that C57BL/6 mice prefer NaCl to water [48,49]. We also showed that mice considered 0.075 M NaCl palatable because its intake after vehicle injection was close to that of saccharin and sucralose. However, the NaCl intake did not significantly differ between the drug injections. The lesion of the BNST decreases the intake of 2% (approximately 0.34 M) of NaCl induced by salt deficiency (appetite) [72]. The inhibition of projective neurons from the subfornical organ to the ventral BNST suppresses the intake of 0.3 M NaCl in a salt appetite condition; however, the activation of those projective neurons elevates the preference for 0.15 M NaCl and reduces avoidance for 0.3 M NaCl under a water-deprived condition [67]. These findings suggest that BNST plays a critical role in sodium consumption. Since the concentration of NaCl in the present study was lower than those used in previous studies, this might have resulted in the absence of an alteration in NaCl intake by the inhibition of BNST neurons. These discussions may be applicable to MSG, because it is also preferable for mice. We determined the concentration (0.3 M) of MSG based on a previous study [49], in which mice demonstrated preference to this concentration of MSG. The present study showed that MSG intake was not altered after CNO injection. However, the amount of MSG intake was almost equivalent to that of water intake. Thus, 0.3 M MSG may not be as preferable for mice as saccharin and sucralose. However, since 600 and 1000 mM MSG were not preferable in a previous study [49], we did not use them in the present study. Although future studies may shed light on the relationship between BNST neurons and MSG intake, we speculate that BNST neurons are not involved in MSG ingestive behavior because the consumption of MSG solution is not altered by lesions in the VTA DA neurons [60].
The present study used an AAV vector containing a gene cassette of hSyn as a promoter, which causes transfection in neurons regardless of their profiles, to manipulate the neurons in the BNST non-specifically. The BNST comprises multiple subnuclei with different structures and functions [73][74][75]. From an anatomical perspective, the BNST is divided into medial, intermediate, lateral, and ventral subdivisions [76]. More exclusively, these three divisions contain different properties of cells, which divide the anteromedial division into six subnuclei (anterodorsal, anteroventral, dorsomedial, magnocellular, dorsolateral, ventral), anterolateral division into five subnuclei (anterolateral, juxtacapsular, oval, fusiform, rhomboid), and posterior division into three subnuclei (transverse, interfascicular, principal) [73]. These subnuclei contain cells expressing different characteristics of neurotransmitters, such as glutamate, GABA, acetylcholine, CRH, vasoactive intestinal peptide, somatostatin, and calretinin [77]. Since the neuron types inhibited by the CNO injection in the present study is unclear, future research should examine the effect of cell type-specific manipulation of BNST neurons on ingestive behavior.
A previous study showed that CNO injection excites rather than inhibits hM4Di-expressing neurons in the dorsal BNST by detecting the induction of Fos protein, a marker of excitation of cells [78]. In contrast, an in vivo electrophysiological recording has demonstrated that CNO injection inhibits rather than excites the hM4Di-expressing neurons in the dorsal BNST [45]. In these two previous studies, different types of AAVs were used: the serotypes and promoters were five and eight and CaMKIIα and hSyn, respectively. The AAV used in the present study was identical to that in a later study [45]. Therefore, we believe that the CNO injection in the present study inhibited BNST neurons.
The present study revealed that the inhibition of BNST neurons decreased saccharin intake. The suppression of CS (saccharin) intake in CTA retrieval was enhanced by chemogenetic inhibition of the BNST neurons. These results suggest that the strength of inhibition of BNST neurons may affect how much animals consume the CS in CTA retrieval. The present study newly reports that BNST is a part of neural mechanisms underlying CTA retrieval. However, exogenous manipulation of neural activity like chemogenetic inhibition might produce unnatural brain function. Therefore, the endogenous activity of BNST neurons in CTA retrieval should be recorded using an electrophysiological or calcium imaging technique in a future study.

Conclusion
The chemogenetic inhibition of BNST neurons via hM4Di receptors decreases the intake of both the learned aversive sweet taste and naturally preferable taste. These results indicate that BNST neurons mediate sweet taste and regulate their intake, regardless of whether they should be ingested or rejected. BNST neurons may be inhibited in the retrieval of CTA, inducing the suppression of CS intake.

CRediT authorship contribution statement
E. Kikuchi and T. Inui: designed and performed research; S. Su: performed research; E. Kikuchi, T. Inui, Y. Sato, and M. Funahashi: wrote the paper. All authors reviewed the manuscript.

Declaration of Competing Interest
The authors declare no competing financial interests.

Data Availability
Data will be made available on request.