Aversive Emotion Activates Orexin Neurons and Induces Tachycardia in Freely Moving Mice

Stress affects the sensory and negative emotional components in the brain and causes autonomic responses and aversive emotion. However, the possible roles for the negative emotional component remain largely unclear. The perifornical area of the hypothalamus has been known as the center for the defense response, or ght-or-ight response, which is characterized by a concomitant rise in arterial blood pressure, heart rate, and respiratory frequency. Orexin neurons located in that region are suggested to be a critical population responsible for that response. In this study, we examined the suggestion by recording orexin neuronal activity and heart rate in freely moving mice using an original dual-channel ber photometry system in vivo. Association analysis between orexin neuronal activity and aversive stress-induced autonomic responses revealed a rapid increase in neuronal activity just prior to changes in heart rate. In addition, we examined whether orexin neurons would be activated by a conditioned neutral sound that was previously associated with aversive stimulus. We show that the negative emotional memory indeed activated orexin neurons and increased heart rate. Our data suggest that orexin neurons are the key component required to receive aversive emotion and link it with an autonomic defense response. Our data also suggest that targeting orexin neurons may enable treatment of psychiatric disorders that result from chronic stress and occur long after the original sensory inputs are gone.


Introduction
Stress is often mistakenly thought of as being purely detrimental. While it is true that excessive stress generally has negative effects that can cause various mental disorders and emotional disturbances 1 , responses to acute stress can be bene cial and even indispensable for life because they protect against potential sources of danger. Acute stress induces a rise in blood pressure, heart rate, respiration, and stressinduced analgesia. These autonomic changes are collectively called the defense response and prepare for and support ght-or-ight behavior 2 . We previously found that orexin-producing neurons in the hypothalamus play a crucial role in this defense response because it is severely attenuated in orexin knockout mice 3 and orexin neuron ablated mice 4 .
Although orexin neurons are essential to autonomic defense responses induced by stress in general, orexin neuronal activity occurring in response to speci c stressors remains unclear. In the past, several technical limitations have prevented us from thoroughly examining these unknowns. First, the use of freely moving animals without anesthesia is necessary because stress responses are unable to occur under anesthesia.
Second, the hypothalamus contains many different cell types and targeting only the orexin-producing neurons of interest was di cult. Some researchers have reported on the activity of orexin neurons in awake rats by using traditional electrophysiological methods such as extracellular recording, and then subsequently identifying the neurons that were recorded from with immunohistochemistry 5,6 . However, this is a time-consuming approach and therefore is not realistic to use for studying the possible effects of several kinds of stressors while simultaneously recording autonomic stress responses. In addition, single cell recording techniques create a sampling bias: there are about 2000-3000 orexin neurons in the mouse brain 7 but the activity of only one or two neurons can be recorded at a time. To overcome these limitations, we set out to develop and improve a ber photometry system. A ber photometry system can track the real-time dynamics of genetically speci ed neuron populations located deep within the brain of freely moving mice by using a single channel ber and genetically encoded Ca 2+ indicators [8][9][10] . We recently utilized a single channel ber photometry system and successfully recorded orexin 11,12 and dopamine 13 neural activities using G-CaMP6. More recently, we upgraded the system to operate via dual channels. The dual channel system can detect more sophisticated data by virtue of twocolor measurement(s) by using one color for reference. The animal's movement sometimes affects the uorescent signal intensity so this improvement was essential for the simultaneous recording of neuronal activity and animal behavior. We used mCherry signal as a reference to G-CaMP signal. Although we brie y reported on this dual channel system in our previous paper 12 , it will be described in more thorough detail here. The system with dual-channel was previously published independently of us 14,15 . However, our system is jointless from the recording place to the photodetector. Thus, the light transmission e ciency of the excitation blue light of our single ber system is theoretically higher than compared to the ferrule systems already published that include one joint. Recently, some groups have published multi-ber photometry systems 16,17 . These systems use bundled bers. Bundling very thin bers provides better spatial resolution than our single ber system. Sometimes it seems that the activity of individual cells can be recorded.
However, since the ber is very thin, it is expected that the detection capability of changes in uorescence is low.
Stress processing in the brain can be described as being dictated by two components. The rst is a sensory discriminative component de ned by strength of pain and perception of places, and the other is an emotional component that is de ned by negative emotions that accompany pain such as anxiety, fear, and aversion 18,19 . Between these components, the emotional component is likely the main contributor to psychiatric disorders resulting from chronic stress because the symptom may occur long after the original sensory inputs are gone.
The neural mechanisms behind the emotional component of stress have not yet been fully elucidated. One of the reasons for this is that many stress studies utilized an acute stress paradigm in which one cannot distinguish between emotional and sensory components. These include restraint stress 20,21 , intruder stress 22,23 , aversive odor stress 24 , auditory stress 12,25 , and trembling stress 23,25 . Therefore, we do not know at present whether the above-mentioned possible contribution of orexin neurons to stress processing is within the branch of the emotional component or downstream of the sensory component.
In order to separate the components, memory-based Pavlovian testing is of interest because this method gives psychological stress to the animals after cessation of original sensory inputs. Examples are fearconditioning from social defeat 26 , electric shock 27 , and mate vocalization 28 . It is interesting to point out that previous research showed that orexin neurons contribute more preferentially to psychological stressors than to physical stressors 29 .
In order to clarify the activity of orexin neurons at the moment stress applied and to reveal possible relationship between orexin neuronal activity and the heart rate response, we measured orexin neuronal activity induced by aversive stress using a dual channel ber photometry system and heart rate using a telemetry system in freely moving mice. Furthermore, we attempted to clarify the possible effect of the negative emotional component of stress in the absence of the original sensory input on orexin neuronal activity using a fear-conditioning paradigm.

Results
Setup of dual-channel in vivo ber photometry system Fiber photometry systems can track the real-time dynamics of genetically speci ed neuronal populations in the deep brain of freely moving mice by using a single ber and genetically encoded Ca 2+ indicators 9 . We improved this system by adding a second channel in order to be able to simultaneously detect both G-CaMP6 and mCherry. To speci cally express G-CaMP6 and mCherry in orexin neurons, we prepared transgenic mice and AAV vectors as shown in Fig. 1A. The AAV mixture consisting of AAVdj-TetO-G-CaMP6 and AAVdj-TetO-mCherry was stereotaxically injected into the hypothalamus of ORX-tTA mice that express tetracycline transactivator (tTA) in orexin neurons (Fig. 1A). These AAVs can drive expression of G-CaMP6 and mCherry proteins in the presence of tTA. Three weeks after this mixture was injected, G-CaMP6 and mCherry expressed almost exclusively in hypothalamic orexin neurons (Fig. 1B). In 393 ± 60 orexin positive neurons (n = 6 animals), G-CaMP6(+) cells were 283 ± 46, mCherry(+) cells were 345 ± 59, and both G-CaMP6-and mCherry-expressing orexin neurons were 260 ± 43 (Fig. 1Bv). There was small number of ectopic expressions of G-CaMP6 (G-CaMP6(+) & orexin (-) in total G-CaMP6(+) was 8.3 %). Blue and yellow excitation lights for G-CaMP6 and mCherry were supplied by LEDs and delivered through a single optical silica ber (Fig. 1C). Fluorescence emission from G-CaMP6 and mCherry expression in the hypothalamic orexin neurons pass back through the single optical silica ber. The uorescent signals pass through two dichroic mirrors and separate into green uorescence derived from G-CaMP6 and red uorescence derived from mCherry. The signals then arrive at their respective photomultipliers (Fig. 1C). To place the optical single silica ber in the desired region of the hypothalamus, we implanted a guide cannula with dental cement (Fig. 1C). The guide cannula allowed the optical ber to be placed just dorsal to the hypothalamic area of interest (Fig. 1D). The single optical silica ber and guide cannula were made in our laboratory with the support of LUCIR Inc. (Tsukuba, Ibaraki, Japan) (Fig. 1E). These tools allowed us to obtain a stable, noiseless, and high-resolution signal. To measure electrocardiogram (ECG) and body temperature, we implanted a telemetry system transmitter (DSI) into the abdominal cavity at the same time as the guide cannula implantation (Fig. 1C). The complete system is shown in Fig. 1F. This system allowed for several types of physiological data to be measured in freely moving mice receiving food and water ad libitum. The data was measured simultaneously under a high temporal resolution (Fig. 1G). G-CaMP6 signals responded independently from shifts in mCherry signals, so we could determine whether the G-CaMP signal was real by comparing it to the levels of mCherry uorescence. The results shown in Fig. 1 demonstrate we were able to utilize a dual channel ber photometry system.
Association between orexin neuronal activity and aversive stress-induced autonomic responses In order to plot real-time orexin neural activity dynamics and autonomic response dynamics before and after various types of aversive stress, we simultaneously measured G-CaMP6 uorescence and ECG and heart rate was calculated from the ECG recordings. As an autonomic response to emotional stress, we analyzed heart rate response because change in body temperature was slow when compared to change in heart rate (see Fig. 1G). Three different stress paradigms with different sensory modalities and different afferent pathways were used. The paradigms were intruder stress, aversive sound stress, and aversive smell stress ( Fig. 2A).
All of the stimuli tested, namely intruder ( Fig. 2B), ultrasound (Fig. 2C), and predator odor (Fig. 2D), induced abrupt increases in heart rate ( Fig. 2B-D, left column) and G-CaMP6 uorescence ( Fig. 2B-D, middle column), while mCherry uorescence ( Fig. 2B-D, right column) did not change. As the control of stress stimuli, the physiological response was measured using a cotton swab without odor and an empty box without intruders (Fig. 2E). These control stimuli did not increase GCaMP uorescence or heart rate as much as ultrasound stress (Fig. 2F).
Although increases in G-CaMP6 uorescence and heart rate appeared to occur simultaneously ( Fig. 2B-D), closer examination revealed that there were differences in the responses. To examine possible differences in the time course, we de ned onset as the time when the signal more than doubled the standard deviation of the baseline uctuation (Fig. 3A). We then calculated ∆time to be the difference between the onset of the increase in heart rate and onset of the increase in G-CaMP6 signal. As a result, ∆time was a positive value in all the stimuli in every mouse ( Fig. 3B) meaning that the change in orexin neural activity preceded the increases in heart rate in all events. It is of interest to point out that ∆time in the TMT test was signi cantly longer than in the intruder test and in the ultrasound test while there was no difference between the intruder test and the ultrasound test, indicating non-uniform changes in heart rate depending on stress stimuli.
We also calculated slope of the signal increase (maximum value -baseline value / time at maximumtime at the onset). This analysis revealed that slope of G-CaMP6 uorescence change in the TMT group was signi cantly smaller than that in the intruder and ultrasound groups (Fig. 3C). Meanwhile, the slope of the heart rate change did not differ among the three stimuli ( Fig. 3D). For TMT stimulation, activation of orexin neurons was relatively gradual when compared to the other stimulation types, and the start of the heart rate response was delayed, but the increase in heart rate ended up being similar.

Conditioned neutral sound stimulation activated orexin neurons
Although we have shown through the previous result that activation of orexin neurons precedes heart rate increase thus indicating close association between orexin neuronal activity and autonomic outcome, those experiments did not separate between the sensory component and the negative emotional component of the stressors. Subsequently, we evaluated the change in orexinergic neuronal activity resulting from exposure to a neutral tone that was previously conditioned as a cue for an unpleasant experience. With this classical conditioning paradigm, it is possible to examine the effects of the pure emotional component because the aversive electrical sensory input was no longer present at the time of testing.
First, while recording the orexin neural activity and electrocardiogram, we gave a neutral tone without electrical stimulation to the mouse (Fig. 4A Pre). No signi cant changes in orexin neural activity and heart rate were observed after the sound exposure ( Fig. 4B left and Fig. 4C). After conditioning was established as in the protocol shown in Fig. 4A, the mouse was exposed to the sound stimulation that is identical to that was applied in the preconditioning period (Fig. 4A Post). As a result, orexin neuronal activity and heart rate increased immediately and markedly ( Fig. 4B right and Fig. 4C). Furthermore, we calculated sound-induced changes in heart rate and orexin neuronal activity and compared them between before and after the conditioning (Fig. 4D). The rate of increase in heart rate and orexin neural activity were signi cantly increased by fear conditioning. Conversely, orexin neuronal activity and heart rate did not change (Fig. 4F, G) when exposed to the neutral tone in un-conditioning protocol shown in Fig. 4E. We examined whether conditioning was established by measuring freezing time during the observation period. As expected, conditioned group animals spent ~50% of the time in freezing while control animals did not (Fig. 4H).

Discussion
In this study, we examined whether aversive emotion has an effect on the activity of hypothalamic orexin neurons in freely moving mice. We rst constructed the recording system to assess neuronal activity from the speci c cell type of interest deep in the brain of freely moving mice. Using in vivo recording of hypothalamic orexin neuronal activity and ECG, we showed that every examined stress type activated orexin neurons just before the change in heart rate took place indicating a possible causative relationship between orexinergic neuronal activity and the resultant autonomic out ow. In addition, we showed that the emotional component of stress also activated orexin neurons, although the magnitude of orexin neuronal activation seems smaller than when aversive stimulation with sensory input was given. We propose that both the sensory and emotional components of stress activate orexin neurons and hence induce autonomic activation. Activation of orexin neurons when actual sensory component is absent (conditioned situation) indicates a possible contribution of the orexin system in the etiology of psychiatric disorders resulting from chronic stress because symptoms may occur after the original sensory inputs no longer occur.
The dual-channel ber photometry system produces good signal even in freely moving mice Using our system, we were able to record and analyze neuronal activity and autonomic responses in freely moving animals with high temporal resolution (Fig. 1G). Two-color simultaneous recording allowed us to distinguish between signal and noise with con dence. Neuronal activity recording with head-xed mice 11,12 is an easy way to distinguish signal from noise because the animals' head is unable to move and affect the signal, but more natural behaviors cannot be measured. As a solution, measurements are taken via optical ber from freely moving mice. However, in this case, the ber may twist or move and the transmission e ciency at the connection between the ber and the uorescence detection device changes and large amounts of noise may occur. The device shown in Fig. 1 was able to eliminate these kinds of noise issues. The uorescence ratio of green and red may also be used when analyzing data, similar to ratiometry analysis employed when using a FRET system. However, we did not calculate the ratio and instead used the red signal as an indicator of stable recording. Although we were able to exclude unstable data that occurred during large or sudden animal movements as judged by mCherry signal, it is necessary to determine a method to completely eliminate instability in future experiments. Fiber photometry systems also often struggle with the weakness of their signals. In previous papers, authors have tried to e ciently extract uorescent signals with various methods such as time-correlated counting system, lock-in ampli er, averaging, and ratiometric measurement 8,9,30,31 . We were able to solve this issue by implanting only a single ber from the uorescence detection device into the brain (Fig. 1E). Systems that use standard ferrules will commonly lose some uorescence signals in their connections. The absence of such a joint is an advantageous feature of our system manufactured by LUCIR Inc.
Orexin neuronal activity and heart rate rapidly increase due to aversive stress stimulation irrespective of sensory modality We showed here that orexin neuronal activity and heart rate are increased instantaneously by aversive stress in freely moving mice. Giardino et al. already showed predator odor-induced activation of the lateral hypothalamic orexin neurons using ber photometry method 32 . Unfortunately, however, they did not assess possible autonomic consequence of the orexinergic activation. We simultaneously recorded heart rate and tried three kinds of stressors to induce aversive emotion and mice showed similar responses to each stimulus (Fig. 2). It has already been shown that heart rate increase due to intruder stress is severely attenuated in orexin de cient mice 3,4 . Therefore, orexin may be involved in the regulation of stress-induced autonomic responses. Furthermore, orexin neurons have dense innervation from the bed nucleus of the stria terminals (BNST) and the central nucleus of the amygdala (CeA) 33 . A recent study revealed that optogenetic/chemogenetic activation of GABAergic neurons in the BNST induces a rapid transition from NREM sleep to wakefulness through inactivation of melanin-concentrating-hormone neurons (which are located in the same hypothalamic region as orexin neurons 33 ) and activation of orexin neurons 34 . A previous study has also shown that stress-induced autonomic responses are attenuated by pharmacological suppression of the activity of the BNST and CeA in wild type mice 6 . The central medial nucleus of the amygdala (CeM) and BNST are known as areas that regulate emotion-related autonomic responses 35 . Therefore, it is suggested that when emotional stress is given to mice, the activity of the CeM and BNST is enhanced, the orexin neurons are subsequently activated, and the autonomic response will occur 2,36 . Actually, corticotropin-releasing factor expressing BNST neurons seem to be directly connected to orexin neurons, although their connections were primarily inhibitory and suggesting complex nature of stress-induced orexinergic activation 32 . In this study, we showed that orexin neurons are promptly activated at the time of stress just preceding an increase in heart rate, thus strengthening the "emotional stress circuit" hypothesis.
Three types of aversive stress were chosen for this study because we hypothesized that if the sensory component of stress directly activates orexin neurons, there might be a different response depending on the sensory modality. However, as shown in Fig. 2, all three stress types activated orexin neurons in a similar manner even though the heart rate response from the TMT test was slightly different from that in the intruder test and ultrasonic test. Therefore, activation of orexin neurons seemed to take place after the sensory inputs of different modalities were integrated as being aversive.
Possible causal relationship between orexin neuronal activity and heart rate increase When focusing on the differences between the starting point of the orexin neuronal activity increase and the starting point of the heart rate increase, we found that the onset of the increase in orexin neuronal activity always preceded changes in heart rate (Fig. 3B). Due to this nding, it seemed reasonable to conclude that there might be a causal relationship between them, namely that orexin neuronal activity may directly cause an increase in heart rate. Unfortunately, our data did not fully prove causality. It is necessary to directly manipulate neuronal activity, with optogenetics or pharmacogenetics for example, to conclusively prove it. Nevertheless, our data strongly support the possibility.
Differences in orexin neuronal activity depending on the type of stress stimulation Interestingly, as shown in Fig. 3B, the Δtime is only high in the group exposed to TMT. We further examined the difference in the rate of signal rise between the three types of stressors by examining the G-CaMP6 slope. This examination shows that the G-CaMP6 uorescence intensity increase in the TMT group is relatively moderate, that is, the neuronal activity seems to rise slowly (Fig. 3C). However, the slope of the heart rate increase in the TMT group is the same as that in the other groups (Fig. 3D). Due to these ndings, we hypothesized that orexin neuronal activity needs to reach a threshold in order to increase heart rate. Taken together, the Δtime in G-CaMP6 slope was prolonged because the TMT group took longer than the other groups to reach the orexin neuronal activity threshold required to initiate heart rate increase. We do not currently have an explanation for the reasons behind why the orexin neuronal activity rises slowly only in the TMT group. It may be that because olfactory information differs from visual and auditory information in that it takes longer for the chemical substance to diffuse and bind to the receptor, and that the experimental result re ects the time difference from when the TMT stress is given until the mouse processes and interprets the odorous substances.
The negative emotional component of stress activates orexin neurons and increases heart rate Sensory input usually increases vigilance levels in animals regardless of the valence of emotion. The orexin neuronal activity we observed (Fig. 2) may be due to an increase in awareness that accompanies an increase in attention. We were unable to distinguish whether the result of the stress-induced increases in orexin neuronal activity and the subsequent autonomic responses shown in Fig. 2 were caused by an increased vigilance caused by sensory input alone or by an "emotional change" accompanying the sensory input. Therefore, we performed an experiment to distinguish between them by utilizing a fear-conditioning paradigm. Observation of freezing behavior clearly showed that the neutral sound did not cause any aversive emotional changes unless the sound was associated with electric shock (Fig. 4B). We observed an immediate increase in orexin neuronal activity and heart rate response only when the neutral tone was previously associated with aversive emotion (Fig. 4C). This observation is in line with reports showing a possible relationship between orexin and fear behaviors 37,38 . Considering the results of Fig. 4 and the description in the previous discussion, it is possible that the changes in orexin neuronal activity observed in Fig. 2 might depend on changes in the emotion of the animals, and not the individual sensory stimulation itself.
Orexin aversive emotion response theory In our previous study 11 , we showed that orexin neurons were activated by painful stimuli (nociceptive stimuli). At that time, it was not possible to distinguish whether the response was caused by the ascending pathway from the nociceptive receptors or if it was caused by the descending pathway after the brain recognized the signal as a painful one. Although direct comparison was di cult, we now speculate that activation of orexin neurons via noxious stimulation in response to the aversive emotion that occurs in response to painful sensations was made similarly to the present study. This speculation is supported by the observation that the orexin neuronal response to noxious stimuli was eliminated under anesthesia 11 .

Conclusions
This paper shows that use of LUCIR's ber photometry system with our modi cation will give us a reliable signal even when the animal is actively behaving. This system can easily be used in combination with ECG, EEG, EMG recording and video tracking. It can provide simultaneous monitoring and analysis at a high temporal resolution of all the physiological pieces of information analyzed from neuronal activity deep in the brain. Although the relationship between stress and orexin neurons has been suggested, it was found that orexin neuronal activity is indeed increased due to aversive stress induced by the sensory modalities tested so far. The most important nding is that activation of orexin neurons was caused not by sensory inputs from stress but by changes in emotion to drive autonomic outputs. Aversive emotion serves a very important function as a warning system for keeping animals informed of a variety of dangers. Unfortunately, in many cases of modern society, repeated exposure to acute stress in unavoidable. For example, o ce workers may repeatedly suffer from aversive emotions, which may manifest in illnesses such as depression. While it may be di cult to completely escape from the stressor in these types of situations, it would be bene cial to be able to reduce the severity of the aversive emotional component. These data suggest that targeting orexin neurons may enable treatment of psychiatric disorders that result from repeated acute stress.

Animals
We used transgenic mice carrying a tetracycline-controlled transactivator transgene (tTA) under the control of the orexin promoter 11 (orexin-tTA mice, n = 45). All experimental procedures were performed in accordance with the guiding principles for the care and use of animals in the eld of physiological sciences published by the Physiological Society of Japan and were approved by the Institutional Animal Use Committee at Kagoshima University (MD15075, MD17090, MD18081). Mice were maintained under a strict 12-hour light/dark cycle (light period: 7:00-19:00; dark period: 19:00-7:00) in a temperature-controlled room (22 °C). Food and water were available ad libitum and all efforts were made to minimize animal suffering and discomfort and to reduce the number of animals used.
In vivo recordings of neuronal activity using ber photometry system and cardiovascular parameters using the radio-telemetry system

Implantation of optical ber for ber photometry system
Over 2 weeks after viral injection, mice were surgically implanted with a guide cannula (LUCIR, Tsukuba, Japan) for placing the optical ber just above the PeF to record orexin neuronal activity (Fig. 1D). At the start of the surgical procedures, mice were anesthetized with iso urane (2-3%, inhalation), and placed on a small animal stereotaxic instrument as described previously. The site of implantation for the guide cannula was as follows: from bregma anterior 1.5 mm, lateral 0.8 mm, ventral 5.0 mm from dura. During implantation, uorescence signal was continuously monitored so that optimal position of the ber tip was easily recognized by abrupt increase of the output signal. The guide cannula was xed by dental cement (Fuji LUTE BC, GC, Tokyo, Japan), gel type quick drying glue (LOCTITE 454, Henkel Japan, Yokohama, Japan), and a small anchor screw. After xation of the guide cannula, measuring ber was removed and a dummy cannula was inserted to prevent dust from entering the implanted guide cannula during the recovery period.
Implantation of the transducer for the radio-telemetry system Immediately after performing the guide cannula implantation, an additional implantation surgery was performed on the mice. To measure heart rate and body temperature, we used a radio-telemetry system (Data Sciences International, St. Paul, MN, USA)(DSI). This system consists of a radio-frequency transducer (TA11PA-C20) and a receiver (RLA1020). The negative electrode (-) of the transducer was stitched into the right side of the mouse's chest cavity, the positive electrode (+) was stitched into the left side of the abdomen, and the body of the transducer was implanted into the abdominal cavity. A temperature sensor in the body of the transmitter allowed abdominal temperature to also be measured. During all surgeries, care was taken to maintain body temperature. After surgeries, mice were treated with penicillin and an analgesic, buprenorphine. For recovery, mice were individually housed and monitored and had access to food and water ad libitum for at least 1 week.

Recording
The recording was started once mice completely recovered from all surgeries and a normal circadian rhythm was present. After the recovery period, the dummy cannula was removed, an optical silica ber (Fig.  1E) was inserted, and mice were moved to their measurement cages (Fig. 1F). Locomotor activity was recorded with a passive pyroelectric infrared motion sensor (AMN 1111, Panasonic Co., Osaka, Japan) that was attached to the ceiling of the experimental cage. Mice were housed in the cage individually during the recording period. Orexin neuronal activity and heart rate were recorded for 2-3 successive days using LabChart software version 8 (ADInstruments, New South Wales, Australia) in unrestrained, freely moving, and unanesthetized conditions.
Fiber photometry device A ber photometry system (COME2-FTR/OPT, LUCIR, Tsukuba, Japan) was used to record the activity of orexin neurons in freely moving mice (Fig. 1C). The system utilizes a single silica ber that can deliver two excitation lights and detect uorescence from G-CaMP6 and mCherry simultaneously. Blue excitation light for G-CaMP6 (470 nm, 0.5 mW at the tip of the silica ber) and yellow excitation light for mCherry (590 nm) were produced by a high-power LED system, blue; Tholab OPT/LED Blue_TT_FC, yellow; Tholab OPT/LED yellow_TT_FC (ThoLab Japan, Tokyo, Japan). Blue and yellow excitation light was re ected by a dichroic mirror, passed through an excitation bandpass lter, and was delivered via a 400 μm silica ber into the brain. G-CaMP6 and mCherry uorescence were collected by the same silica ber and guided to an individual photomultiplier for either G-CaMP6 or mCherry (PMTH-S1M1-CR131, Zolix instruments, Beijing, China) (Fig. 1F). The signal was digitized using a data acquisition system (PowerLab16/35, ADInstruments, New South Wales, Australia), and recorded by LabChart software version 8. Signals were collected at a sampling frequency of 100 Hz.

Stress stimulation
Three types of stressors were applied on the following day after the basal measurement was completed ( Fig. 2A). The rst was a socioemotional stressor via the resident-intruder stress test. This stressor was applied by placing an age-matched wild-type mouse (intruder mouse) contained in a small polypropylene cage into the experimental cage for 2 min. The polypropylene cage is constructed so that the intruder and resident (experimental) mice are unable to physically contact, but visual, auditory, and olfactory communication is available. The second stressor was aversive sound stress. For this test, an approximately 100 dB/25 kHz ultrasonic sound was applied to the mouse via an ultrasound-emitting device (PET-AGREE, K-II enterprise, New York, USA). The device was turned on above the cage for 2 seconds. The third stressor was aversive smell stress. For the aversive smell, 2,4,5-trimethylthiazoline, (TMT, Contech Enterprise, Victoria, Canada) a constituent of fox urine and feces, a common predator odor for mice 24,39 , was placed near the nose of the mouse for 2 seconds via cotton swab.
The rationale for choosing these particular types of aversive stress were carefully considered. First, intruder stress has been used in orexin knockout mice showing a possible role of orexin neurons in stress-induced autonomic responses 3 . Second, we are interested in separating the sensory component and the emotional component of stress. If the sensory component of stress directly activates orexin neurons, there might be a different response depending on sensory modality (optic, auditory, and olfactory) because the information transmitted to the brain from the respective receptors (eyes, ears, and nose) occurs via different sensory afferent pathways.
As the control of stress stimuli, the physiological response was measured using a cotton swab without odor and an empty box without intruders (Fig. 2E).

Fear conditioning
The animal was placed in an experimental chamber and acclimatized for 2 hours and then neutral tone sound (60 dB, 1 kHz) was given for 2 sec during ber photometry and the heart rate recording. In order to associate a neutral tone with an unpleasant experience (Fig. 4A), 1 sec electric shock was given to the hind paw of the animal 1 to 5 sec after a 2 sec neutral tone. This was repeated 15 times using a shock generator (CBX-CT) and a cycle timer (CSG-001, Muromachi Kikai, Tokyo, Japan). After a resting period of 1 hr., the identical neutral tone was given during ber photometry and heart rate recording. During the post-recording period, freezing behavior was also measured to con rm that classical fear conditioning had been successfully established. Freezing time was visually calculated by observing videotaped animal behavior by the experimenter who is blinded to the experimental group. No electrical stimulus was given to the control animals. All the procedures were performed in one chamber. Both behavioral test (sound stimulation and fear conditioning) including baseline recording were performed between 12:00 and 18:00. Immunohistochemistry.
Mice were deeply anesthetized with urethane (

Consent for publication
Not applicable.
Availability of data and materials The summary statistics are available within the article. The data that support the ndings of this study are available from the corresponding author upon reasonable request.

Funding
This work was supported by JSPS KAKENHI Grants (17K14936 to A. Yamashita; 17K16387 to S.M.; 16H05130 to T.K.) and CREST JST (JPMJCR1656 to A. Yamanaka) in the courses during the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions A.Yamashita and T.K. designed the study; A.Yamashita, S.M., R.N., J.K., A.Yamanaka and T.K. conducted the study and analyzed the data; and A.Yamashita and T.K. wrote the manuscript. All authors approved the nal version of the manuscript.