Stress—commonly defined as a disruption of homeostasis—impacts a variety of behavioral and physiological parameters, ranging from cardiac and respiratory patterns to arousal and emotional states [1-8]. While adaptive stress responses are essential for the survival and well-being of all animals, excessive and maladaptive stress responses contribute to the etiology of numerous disorders, including anorexia, depression, and anxiety disorder [7, 9-13]. As such, considerable effort has been extended in gaining a mechanistic understanding of stress responses, which is still an active research area [14-23].

A body of research has established that multiple forebrain structures coordinately orchestrate stress responses [24-27]. Among these brain regions, many distinct subregions and cell types of the LS have been implicated in diverse stress-related functions [28-36]. Recent studies have revealed a role of the LSd in stress-related behaviors [29, 30, 34, 36], where neurons expressing somatostatin (a neuropeptide also linked to stress in different parts of the brain [37-42]) are concentrated [28-30, 36, 43-46]. Indeed, LSd^Sst neurons have been shown to be directly innervated by the hippocampus and can modulate contextual fear discrimination [29]. These led us to hypothesize that LSd^Sst neurons may play an important role in regulating stress responses.

Norepinephrine (NE) is a key neuromodulator in stress responses [47-54]. The LS is densely innervated by noradrenergic fibers, and some LS neurons express adrenergic receptors [55-59]. Besides, studies over the past several decades have identified the functional relevance of NE signaling to LS neurons in stress [60, 61]. However, the specific identity of LS neurons that receive NE signals has not been identified. The locus coeruleus (LC) is a pontine brain area that serves as a major source of NE to many forebrain areas, including the LS [53-56]. Since noradrenergic neurons in the LC are well known to be activated upon diverse stressors (such as restraint stress, innate fear, and footshock) [49, 51-54], their connections to the LS may represent an important component of stress reactions.

Here, we explored the role of LSd^Sst neurons in stress responses and sought to determine the effect of NE on LSd^Sst neurons using in vivo fiber photometry, anatomical tracing, ex vivo electrophysiology, and optogenetic stimulations. Our data collectively show that LSd^Sst neurons respond to diverse stressors and their activity is modulated by NE, suggesting a link between LSd^Sst neurons, noradrenergic signaling (possibly arising from the LC), and stress responses.

Mice

All procedures were approved by the Seoul National University Institutional Animal Care and Use Committee. Adult wild-type or heterozygote mice from C57BL/6J background (C57BL/6J mice, JAX #000664; Sst^{tm2.1(cre)Zjh}/J, JAX #013044) were groups housed (except for the restraint experiment) with ad libitum access to food and water in a temperature- and humidity-controlled room, with reverse 12-hr light/dark cycle. Both male and female mice at least six weeks of age were used for data collection. Data from males and females were pooled as we found no evidence for sex differences in all our experiments. All mice used fiber photometry recordings were males. All behavior experiments were performed during the dark cycle.

Viral constructs

The recombinant adeno-associated virus (AAV) vector expressing GCaMP6m (AAV1-hSyn-FLEX-GCaMP6m, 1.2×10¹³ copies/ml) was purchased from the Penn Vector Core, and the AAVs expressing channelrhodopsin (AAV5-EF1α-DIO-hChR2(H134R)-eYFP, 6.2×10¹² copies/ml), eYFP (AAV5-EF1a-DIO-eYFP, 3.5×10¹² copies/ml), optimized rabies G protein (AAV8-CA-Flex-RG, 1.8×10¹² copies/ml) and TVA receptor (AAV8-EF1a-FLEX-TVAcherry, 5.4×10¹² copies/ml) were obtained from the UNC vector core. The AAV expressing mRuby-fused synaptophysin (AAV-DJ-hSyn-FLEX-GFP-2A-Synaptophysin-mRuby, 4.0×10¹³ copies/ml) was purchased from the Stanford Vector Core. The recombinant EnvA-pseudotyped G-deficient rabies virus vector expressing GFP (RV-EnvA-ΔG-GFP) was purchased from the Salk vector core or generously provided by B. K. Lim (UCSD).

Stereotaxic surgery

Mice were placed in a stereotaxic frame (Kopf Instruments) while resting on a heat pad under 1.5~3.0% isoflurane anesthesia. Following hair removal and alcohol disinfection, craniotomy was performed using a hand drill (Saeshin, 208B), and 250~300 nl of viral vectors were injected to the LS using a pressure injection system (Nanoliter 2000) with a pulled glass capillary at 50 nl/min. After injection, the capillary was retracted slowly (0.01 mm/s) to prevent the virus from flowing backward. The coordinates were +1.00 mm antero-posterior (AP), 0 mm medio-lateral (ML), -2.70 mm dorso-ventral (DV) for LSd stimulation, anterograde projection mapping, and rabies tracing experiments, except for the fiber photometry group.

For fiber photometry recordings from the LSd, recombinant AAVs expressing GCaMP6m were unilaterally injected into the LSd of Sst^cre/+ mice at -18 degree angle relative to the sagittal plane at +1.00 mm AP, -0.44 mm ML, -2.79 mm DV to avoid the lateral ventricles. Then a low-autofluorescence fiberoptic cannula (Doric Lenses, NA 0.48, 400 µm core) was implanted 50 μm above the virus injection site in the same manner. The cannulae were affixed to the skull with C&B Metabond (Parkell) and dental cement.

For optogenetic stimulation experiments, recombinant AAVs expressing channelrhodopsin (ChR2) were injected into the LSd of Sst^cre/+ mice, and fiberoptic cannulae (NA 0.22, 200 µm core) were bilaterally implanted. The coordinates for cannulae were +1.00 mm AP, ±1.05 mm ML, -2.00 mm DV with a 16-degree angle relative to the sagittal plane for the LSd. Angled implantation enabled bilateral implantation. For control cohorts, Sst^cre/+ mice were injected with the AAV expressing eYFP, or Cre-negative siblings of Sst^cre/+ mice were injected with the AAV expressing ChR2. Because there were no discernible differences in our behavioral measurements between the two control groups, we combined the data from both groups. For the experiments that used head fixation, a custom-made stainless steel bar (4.0×1.0×1.0 mm³) was attached to the dental cement to allow subsequent head fixation to a custom-made station.

For anterograde projection mapping, AAVs expressing GFP and mRuby-fused synaptophysin were injected, and after 3~4 weeks, mice were euthanized and processed for histology. For rabies tracing experiments, AAVs expressing Cre-dependent TVA and G were injected into the LSd of Sst^cre/+ mice. After 3 weeks, recombinant EnvA-pseudotyped G-deficient rabies virus (RVΔG) expressing eGFP was injected into the LSd. RVΔG selectively infects TVA-expressing LSd^Sst neurons, and co-expression of G enables LSd^Sst neurons to complement RVΔG and produce infectious rabies viruses that spread to their direct presynaptic partners. After 8-9 days, mice were euthanized and processed for histology.

The incision was sutured, and antibiotics and analgesics were given to the mice. Mice were kept in their home cages for four weeks to allow for recovery and adequate viral expression.

Fiber photometry

Fiber photometry recordings were performed as previously described [62, 63]. Briefly, excitation lights from 470-nm and 405-nm LEDs (Thorlabs, M470F3/M405F1) that were sinusoidally modulated by the RZ5P processor (Tucker Davis Technologies) at 211 Hz and 531 Hz, respectively, were delivered to the target region of mice via a low-autofluorescence fiberoptic patch cord and cannula (Doric Lenses, 400 μm-core, 0.48 NA). The light intensity was maintained at a maximum of 20 μW during recordings. The emitted fluorescence was detected by a femtowatt photoreceiver (Newport, 2151). The resulting signal was demodulated, amplified, and collected at ~1 kHz by the RZ5P processor. To correlate the photometry signals with behavior, behavioral experiments were recorded using a video camera, and the location and activity of the mice were automatically tracked by video tracking software (Noldus Ethovision). A TTL pulse generated by a pulse generator (Sanworks, Pulse Pal) was split and fed into the RZ5P processor and a TTL-triggered blue LED was placed in the field of view where mice could not see. For the footshock, event timestamps marking shock deliveries were used.

Optogenetic manipulations

For optogenetic stimulations, 5 or 10 mW blue light (159 mW/mm² at the tip of the patch cords) was generated by a 473-nm laser (MBL-III-473; OEM Laser Systems) and delivered to mice through fiberoptic patch cords (0.22 NA, 200 μm diameter; Newdoon) connected by a rotary joint (Doric Lenses). Light pulses (5 or 10 ms pulse trains at 15 Hz) were generated by controlling the blue laser with a pulse generator (Pulse Pal, Sanworks). Light pulse trains were delivered throughout the experimental session unless otherwise stated.

Behavioral assays

To reduce stress caused by experimenters, all mice were handled for at least 5 days prior to behavior experiments. To reduce the stress caused by the patch cord connection, mice were first connected to a patch cord and placed in a new cage for 5 min before being introduced to the behavior arena. To avoid interactions between experiments, different behavioral assays were performed at least two weeks apart. Behavioral tests that cause severe stress in mice (e.g., footshock, restraint stress, and tail suspension) were carried out at the end (specifically, for the footshock experiment) or in a separate cohort (for the restraint stress and tail suspension experiments). For all behavior assays where video analysis was appropriate, video tracking software (Noldus, EthoVision XT) was used to track the location and activity of mice.

For the elevated plus maze test, mice were placed in a plus-shaped plastic maze, consisting of two open and closed arms (30×5 cm) extending from a central platform elevated from the ground by 50 cm. Each mouse was initially placed in closed arms, and the behavior was recorded for 10 min for all experiments.

For the open field test, mice were placed in an open field chamber (50×50×50 cm), where the center zone was defined as a square at the center (20×20 cm). Each mouse was placed at the corner at the beginning of the session. Mouse behavior was recorded for 10 min for fiber photometry experiments. For stimulation experiments, mice were recorded for 20 min, in which laser stimulation was applied at the second and fourth 5-min epochs; the two laser-off and laser-on epochs were pooled for analysis.

To conduct the fiber photometry test during the restraint test, mice were introduced into a clear Plexiglas tube (3 cm inner diameter) and two black plastic gates with holes for the nose and tail were inserted to hold mice in place tightly. For the baseline recording, single-housed mice were connected to the patch cord, returned to the home cage, and the fiber photometry recording began after 5 min. After baseline recording in the home cage for 10 min, mice were briefly anesthetized with isoflurane and placed into the restraint apparatus. The restraint apparatus was then placed in a behavioral chamber (25×25×25 cm) for 45 min. Mice were then returned to the home cage and the recording was continued to monitor the post-restraint responses.

For the footshock test, the mice were placed in a behavioral chamber (18×20×36 cm) with a metal grid floor connected to an electric shock generator (Precision animal shocker, Coulbourn). After 10 min of baseline recording, a footshock (0.2 mA, 2 s) was delivered at a pseudo-random interval (90 s on average) 5 times.

For the tail suspension test, the tail of mice was attached to a bar that was 40 cm elevated from the ground. A 2-cm tube was inserted through the tail to prevent the mice from climbing up during the test. The session was divided into three 3-min epochs, where the laser was turned off, on, and off, respectively.

For the real-time place preference test, mice were placed in a white plastic arena (50×25×25 cm) consisting of two identical chambers with a slit to freely move across the chambers for 15 min. One chamber was paired with laser stimulation, and the choice of the stimulation-paired chamber was counterbalanced across mice.

To measure heart and respiratory rates, pulse oximetry (MouseOx Plus, Starr Life Sciences) was used. First, mice were shaved around the neck, then habituated to head fixation and the collar sensor for 30 min a day, at least for 3 days. For each recording session, mice were acclimated to the head fixation and collar sensor for 5 min, then the recording began. Heart and respiratory rates were measured by a collar sensor attached to the animal’s neck for 9 min, in which 3-min laser stimulation was applied from 3 min after the recording began.

Ex vivo electrophysiology

Mice were anesthetized with isoflurane and the brains were extracted. Acute 300 μm-thick coronal slices were obtained using a vibratome (Leica VT1200S) in an ice-cold dissection solution (in mM; 212 sucrose, 3 KCl, 26 NaHCO₃, 1.25 NaH₂PO₄, 7 MgCl₂, and 10 glucose, gassed with 95% O₂, and 5% CO₂). Slice was recovered in artificial cerebrospinal fluid (aCSF) (in mM; 118 NaCl, 2.5 KCl, 11 glucose, 1 NaH₂PO₄, and 26.2 NaHCO₃, gassed with 95% O₂ and 5% CO₂) at 35℃ for an hour, and then maintained at room temperature. All electrophysiological recording was made under the constant perfusion of aCSF heated to 30℃. Neurons were visualized with an upright microscope (Nikon, Eclipse FN1) equipped with both DIC optics and a filter set for visualizing eYFP (ChR2) and tdTomato, using a 40× water-immersion objective and an sCMOS camera (Andor, Zyla 4.2).

Whole-cell recordings were made from fluorophore-labeled or non-labeled LSd neurons, using patch pipettes (2~6 MΩ) filled with either potassium-based internal solution (for current-clamp analyses; in mM; 130 K-gluconate, 10 HEPES, 0.6 EGTA, 5 KCl and 2.5 MgCl₂, pH 7.3) or cesium-based internal solution (for voltage-clamp experiments; in mM; 115 Cs-methanesulphonate, 20 CsCl, 10 HEPES, 2.5 MgCl₂, 0.6 EGTA, 5 QX314, 4 Na2-ATP, 0.4 Na2-GTP, and 10 Na-phosphocreatin, pH 7.3). Series resistance was typically 10-15 MΩ. Neuronal activity was filtered at 2 kHz, sampled at 20 kHz, and recorded to disk using Multiclamp 700B and Clampex 10.3 (Molecular Devices).

To record optogenetically evoked action potential, 3.5~5.5 mW blue light generated by a 473-nm blue laser was delivered to slices through fiberoptic patch cords, while recording from ChR2-expressing neurons at a holding current of 0 pA. Pulsed input signals (5-ms pulse trains at 15 or 30 Hz) were generated by pClamp (Molecular Devices). To record miniature excitatory postsynaptic current (mEPSCs), LSd neurons were recorded at a holding potential of -70 mV, with aCSF containing 50 μM picrotoxin and 1 μM tetrodotoxin. Recorded data were analyzed using Minianalysis 6.0.7 (Synaptosoft).

Data analysis

All data were analyzed with custom-written Matlab (Mathworks) code. The photometry signal was analyzed as previously described [62, 63]. Briefly, data were low-pass filtered at 2 Hz, downsampled to 100 Hz, and a linear function scaled the 405-nm signal to the 470-nm signal to obtain the fitted 405-nm signal. The ΔF/F was calculated as (raw 470 nm signal – fitted 405 nm signal) / (fitted 405 nm signal). Peri-event time plots were created using either the TTL timestamps generated by shocker-triggering pulse generators or timestamps marked by manual video analysis.

In anxiety tests, ΔF/F for specific zones was normalized using average and standard deviation of ΔF/F values from the whole-session data. For peri-event plots, ΔF/F was normalized using mean and standard deviation from -5 s to -2.5 s relative to zone entrance events. Only zone entrance events with intervals longer than 5 s were included in the analysis, to ensure that another entrance event does not influence the defined baseline. To examine the correlation between the calcium activity and the movement velocity of mice, normalized ΔF/F and velocity values from the open field test were binned into 1-s intervals.

In the restraint stress experiment, ΔF/F values were normalized by subtracting the average ΔF/F of the baseline recording (from the first homecage recording) and dividing the difference by the baseline standard deviation. For the analysis of the restraint epoch, recordings from the last 30 min were used to exclude the periods that could have been potentially affected by brief anesthesia preceding restraint. The peak time and amplitude of individual calcium transients were identified based on the maximum value of the thresholded signal within each transient, using the code generously provided by Rothschild [64].

In the footshock test, ΔF/F during shock was normalized using the mean and standard deviation of ΔF/F values from the 10-min baseline recordings. For peri-event time plots, the baseline was defined as 10 s preceding shock delivery in footshock tests.

Histology and microscopy

Mice were anesthetized and transcardially perfused using ice-cold 1× PBS and 4% paraformaldehyde (PFA) solution in PBS. Brains were extracted and fixed overnight in PFA solution, and equilibrated in 30% sucrose solution, before cutting into 50 µm-thick sections using a freezing microtome (Leica, SM2010R). Slices were stored in a cryoprotectant solution (a 5:6:9 mixture of glycerol, ethylene glycol and PBS) at 4℃. Sections were then washed in PBS, incubated for >25 min in 1: 50,000 DAPI solution, washed again in PBS and mounted on microscope slides with PVA-DABCO. Confocal images were obtained on a Zeiss LSM 880 laser scanning microscope using 10×/0.45 NA objective lens. Only mice with restricted expression of the desired transgene (GCaMP and ChR2) in the LSd were included in the study. Note that while we did not demonstrated the specificity of Cre and Cre-dependent transgene expressions in Sst-expressing LSd neurons of Sst^cre/+ mice (i.e., Sst^{tm2.1(cre)Zjh}/J, JAX #013044), one previous study used the same mouse line to Cre-dependently express transgenes in LSd^Sst neurons [45], and another study demonstrated the specific expression of Cre in Sst-expressing neurons in the dorsomedial striatum [65].

Statistical analysis

Statistical analyses and linear regressions were performed using Matlab (Mathworks) or Prism (GraphPad). We used a two-tailed Wilcoxon rank-sum test, one-way repeated measures ANOVA, two-way repeated measures ANOVA with subsequent Bonferroni post-tests, or Pearson correlation depending on the experimental paradigm. *p<0.05, **p<0.01, ***p<0.001. Data were presented as mean±s.e.m. unless otherwise noted. No statistics to determine sample size, blinding or randomization methods were used. Viral expression and implant placement were verified by histology before mice were included in the analysis.

LSd^Sst neurons are activated by diverse stressors

To directly assess if LSd^Sst neurons are activated in response to stressors in vivo, we examined their real-time activity using fiber photometry. For this experiment, we injected Cre-inducible AAV vectors carrying calcium reporter GCaMP6 in the LSd of Sst^cre/+ mice (Fig. 1A) and implanted a fiberoptic cannula slightly above the injection site. Sinusoidally modulated 470 and 405 nm lights were delivered to each LS subregion to stimulate GCaMP6 in a Ca²⁺-dependent and independent manner, respectively, and the signals from emitted photons were detected, de-modulated, and collected (Fig. 1B). This experimental design enabled ratiometric measurements of GCaMP activity, yielding signals minimally compromised by motion artifacts and ambient light [62, 63].

We first examined if the activity of LSd^Sst neurons correlates with anxiety levels by recording from these neurons during an open field test, a well-established behavioral assay for anxiety. We observed that this population exhibited a significant increase in activity in the center zone of the open field (Fig. 1D, E), which is the anxiogenic, stress-inducing area of the open field arena. Consistently, the peri-event plot of average calcium activity revealed that activities of LSd^Sst neurons increase upon the center zone entry (Fig. 1F). In contrast, we observed only weak correlations between the normalized calcium activity and the velocity of the animals (Fig. 1C, Pearson’s r=0.25). To further examine the correlative functions of LSd^Sst neurons in anxiety, we next recorded from these neurons during the elevated plus maze test, another extensively validated anxiety assay. In line with the result from the open field test, activities of LSd^Sst neurons were significantly increased in, and upon the entry of, the open arms of the maze (Fig. 1G~I). Thus, LSd^Sst neurons are activated in anxiogenic and stressful environments.

We then asked if LSd^Sst neurons are activated by other types of stressors, such as physical restraint for a prolonged time, or brief electric footshocks. We found that physical restraint in a cylindrical tube induced a strong increase in the amplitude of the calcium transients for the entire duration of the restraint (Fig. 1J, K). LSd^Sst neurons were also potently activated by brief two-second electric footshocks (Fig. 1L~N). Notably, the degree of activation by electric footshock was considerably larger than the activation of LSd^Sst neurons in anxiogenic and stressful environments (i.e. the center zone of the open field or the open arms of the elevated plus maze), indicating that the response amplitude is likely correlated with the intensity of the stress stimulus. These results also show that LSd^Sst neurons can be activated by diverse stressors, of which durations range from a few seconds to tens of minutes.

LSd^Sst neurons are connected with regions implicated in stress response

To gain insight into how LSd^Sst neurons process the stress signals, we explored the output and input connectivity of LSd^Sst neurons. First, to probe the anatomical projection targets of LSd^Sst neurons, we infused the AAV vector carrying a Cre-dependent construct encoding GFP and synaptophysin-fused mRuby (which labels the neural processes and terminal boutons, respectively) in the LSd of Sst^cre/+ mice (Fig. 2A, B). We found dense GFP and mRuby signals in the forebrain area that are known to regulate stress response, for example, the ventral basal forebrain (vBF) (including the nucleus of the horizontal limb of the diagonal band (HDB), magnocellular preoptic nucleus (MCPO), and substantia innominata (SI)), anterior hypothalamus (AH), LH, and supramammillary nucleus (SuM), as well as the LSv (Fig. 2C).

Next, to probe the monosynaptic inputs of LSd^Sst neurons, we injected Cre-inducible AAV vectors expressing avian-specific retroviral receptor TVA and a rabies glycoprotein (RG) in the LSd of Sst^cre/+ mice. After three weeks, we injected EnvA-pseudotyped RG-deficient rabies virus (RVΔG) carrying eGFP in the LSd, selectively infected TVA-expressing LSd^Sst neurons and produced reconstituted viral particles that can spread to their direct presynaptic inputs (Fig. 2D) [66]. The resulting eGFP-labeled neurons, which correspond to the monosynaptic inputs of LSd^Sst neurons, were densely located in the hippocampus (HPC), but were also found in in many areas known to be activated by stress, such as the ventral pallidum (VP), lateral preoptic area (LPO), lateral hypothalamus (LH), periaqueductal gray (PAG) and LC, albeit more sparsely than the hippocampus (Fig. 2E, F). These results, which are consistent with a previous report of retrograde tracing from the LSd [67], show that LSd^Sst neurons are connected with numerous structures functionally implicated in stress response.

NE increases excitatory synaptic transmission onto LSd^Sst neurons

Since LSd^Sst neurons are directly innervated by LC neurons (Fig. 2E, F), NE release from LC may modulate the activity of LSd^Sst neurons. To test if NE can modulate the basal synaptic activity of LSd^Sst neurons, we injected a Cre-inducible AAV vector carrying eYFP in the LSd of Sst^cre/+ mice and performed whole-cell electrophysiological recordings from LSd^Sst neurons that are virally labeled with eYFP (Fig. 3A). We recorded and analyzed miniature excitatory postsynaptic currents (mEPSCs) from LSd^Sst neurons before and after the bath application of NE (Fig. 3B). Strikingly, we observed that the amplitude of mEPSCs was increased after the application of NE (Fig. 3C), but the frequency remained unaltered (Fig. 3D). These results demonstrate that NE increases spontaneous excitatory synaptic transmission onto LSd^Sst neurons, possibly via a postsynaptic mechanism. In contrast, bath application of corticosterone did not affect either mEPSC frequency or amplitude (Fig. 3E~G). These results show that NE increases excitatory synaptic transmission onto LSd^Sst neurons, and support the possibility that NE release from noradrenergic LC neurons in response to stressors may contribute to the excitatory response of LSd^Sst neurons to diverse stressors as observed in Fig. 1.

Optogenetic stimulation of LSd^Sst neurons does not affect stress-related behaviors or autonomic functions

We next sought to determine the causal function of LSd^Sst neurons using optogenetics. To validate the optogenetic stimulation, and also to determine the optimal illumination frequency that enables reliable triggering of action potentials in these neurons, we performed whole-cell electrophysiological recordings from LSd^Sst neurons that are virally transduced with a light-gated cation channel channelrhodopsin-2 (ChR2) (Fig. 4A). Delivery of light pulse trains at either 15 or 30 Hz evoked action potentials in ChR2-expressing LSd^Sst neurons (Fig. 4B), but the success rate (action potential fidelity) was only ~55% for 30 Hz stimulations, whereas the fidelity was >93% in case of 15 Hz stimulations (Fig. 4C). Because of this, as well as previous findings that the 15 Hz frequency is in the range of natural firing rates of LS neurons during stress [31, 68], we chose to use 15 Hz stimulation, although we note that a previous study used a variant of ChR2 optimized for ultrafast stimulation to perform reliable stimulations at even higher frequency ranges [45].

To explore the behavioral effect of activating LSd^Sst neurons, we Cre-dependently expressed ChR2 in this population and bilaterally implanted fiberoptic cannula slightly above the injection site (Fig. 4D). We then delivered light pulses through the fiberoptic cannula to stimulate LSd^Sst neurons, while subjecting mice to stress-related behavioral assays. However, we found no effect in any of the assays that we performed. Optogenetic stimulation of LSd^Sst neurons did not affect anxiety-like behaviors in the elevated plus maze test (Fig. 4E) or the open field test (Fig. 4F). In the open field test, during the light-on epoch, there was no significant difference in velocity between ChR2 and control mice (Fig. 4G). In addition, the number of the center zone crossing was not significantly changed (n=11 ChR2, n=11 Ctrl, two-way repeated measures ANOVA interaction, F(1,20)=0.095, p=0.7614). Next, activating these neurons left depression-like symptoms in the tail suspension test unaltered (Fig. 4H), and did not induce place preference or avoidance in the real-time place preference test (Fig. 4I). Given the negative results from all of our behavioral assays, we further sought to determine if optogenetic stimulation of LSd^Sst neurons might affect physiological parameters, such as heart and respiratory rates. However, this manipulation did not influence either the heart rate (Fig. 4J) or the respiratory rate (Fig. 4K). Thus, under our experimental conditions, simultaneous activation of the entire LSd^Sst population using optogenetic means has no effect on stress-related behavior or physiology.

In this study, we investigated the role of LSd^Sst neurons in stress responses and their connectivity, as well as the effect of NE on these neurons. We first demonstrated that LSd^Sst neurons are activated by a variety of stressors, the duration of which can range from seconds to minutes, and the amplitude of which likely correlates with the intensity of the stress stimulus, using fiber photometry (Fig. 1). Anatomical tracing in both anterograde and retrograde directions revealed that LSd^Sst neurons receive inputs from key stress-responsive regions including the LC and in turn project to many areas that are known to mediate stress-related functions (Fig. 2). Next, we found that, in the ex vivo slice patch-clamp setting, NE increases spontaneous excitatory synaptic transmission onto LSd^Sst neurons (Fig. 3). Optogenetic stimulation of the LSd^Sst population, on the other hand, had no effect on stress-related behavior and physiology—at least not in the tests performed in this study (Fig. 4).

Several recent studies have also investigated the function and anatomy of LSd^Sst neurons, each from a distinct perspective [29, 30, 45]. Remarkably, among these, an elegant study by Besnard and colleagues has employed miniaturized microscopes to characterize the responses of individual LSd^Sst neurons under various stressful situations [29]. This study found that approximately 47% of individual LSd^Sst neurons were activated upon footshock in at least one experimental session and that the calcium transient frequency of LSd^Sst neurons was higher in the open arms of the elevated plus maze than in the closed arms, which is consistent with the overall responses of the LSd^Sst population that we observed using fiber photometry (Fig. 1G~I). However, the calcium transient frequency of LSd^Sst neurons decreased in the center of the open-field, which appears inconsistent with our data (Fig. 1D~F) and contradicts the previous result from their elevated plus maze test. This discrepancy may be due to the difference in analysis parameters. In our study, we used fiber photometry to measure and analyze the amplitudes of averaged calcium signals of the LSd^Sst population. On the other hand, Besnard and colleagues employed miniature endoscopes to monitor the calcium transients of individual LSd^Sst neurons, and primarily focused on analyzing the frequency of the calcium transients, which may not correlate with the amplitude of the bulk calcium signals measured in our study. Notably, they reported that the maximum calcium amplitude of LSd^Sst neurons was higher in the open arms of the elevated plus maze than in the closed arms, and was also higher in the center of the open field than the periphery, which are consistent with each other and also in agreement with our results. Further research into the amplitude of calcium transients of LSd^Sst neurons in stressful environments would be useful in resolving this issue.

In addition, Besnard and colleagues [29] found that optogenetic stimulation of LSd^Sst neurons reduced anxiety-like behaviors in the elevated plus maze test, whereas we found no effect (Fig. 4E), despite the similar targeting and optogenetic stimulation conditions (note that another recent study also found that chemogenetic stimulation of LSd^Sst neurons had no effect on anxiety-like behavior in the elevated plus maze test [30]). We speculate that this disparity might be attributable to the different circadian timing used for the behavioral assays (light cycle in [29] vs. dark cycle in our study) and/or the light intensity during the behavioral assays (700 lux in [29] vs. 15 lux in our study), which considerably impacts the baseline stress levels of the animals [69-71]; one possible explanation is that LSd^Sst neurons may regulate anxiety-like behaviors only in stressful situations, which could be formally tested in the future studies.

Simultaneously, the study by Besnard and colleagues [29] revealed significant functional heterogeneity within the LSd^Sst neuron population, which may help to explain to other reported roles of LSd^Sst neurons, such as modulation of contextual fear discrimination [29] or food-seeking behavior [45]. Distinct subpopulations of LSd^Sst neurons may play divergent roles. A recent paper that dissected the connectivity of LSd neurons and delineated the functionally different neural subpopulations in this region lends support to this viewpoint [36].

Numerous studies have revealed that NE is released upon stressors and modulates a variety of behavioral and physiological functions [49, 72-76]. We observed that NE, but not corticosterone, increased excitatory synaptic transmission of LSd^Sst neurons likely via a postsynaptic mechanism (Fig. 3). These results suggest that modulation of adrenergic receptors expressed on LSd^Sst neurons may elicit behavioral stress-related functions. Indeed, microinjection of NE into the LS has been reported to drive stress-related behaviors, while injecting either α1- or β-receptor antagonists into LS attenuates defensive behavior following acute footshock stress [60, 61]. It would be interesting to see if adrenergic receptors are expressed on LSd^Sst neurons and if their activation causally regulates stress responses. Furthermore, noradrenergic LC neurons project widely to multiple brain areas, including the LS [28, 55, 56]. Using retrograde tracing, we found that LSd^Sst neurons receive inputs from the LC neurons, indicating that the LSd^Sst neurons likely receive NE from noradrenergic LC neurons. Given the well-established role of noradrenergic LC neurons in stress responses, it is tempting to speculate that activation of noradrenergic LC neurons releases NE to LSd^Sst neurons, and a subpopulation of LSd^Sst neurons expressing adrenergic receptors is activated by NE and elicits stress-related behaviors such as anxiety-like behaviors.

However, we observed that optogenetic stimulation of all LSd^Sst neurons has no effect in stress-related behavior or physiology, at least under our experimental conditions (Fig. 4). We speculate that this is owing to the functional heterogeneity as discussed above. More specific targeting of functionally discrete subpopulations of LSd^Sst neurons would likely reveal their causal roles. On the other hand, we suppose that the loss-of-function experiments are unlikely to elicit noticeable effects, since multiple pathways supporting stress responses may work in parallel and redundant manners. For example, noradrenergic LC neurons are considered to be an important circuit node for stress-related behaviors, such as stress-induced anorexia and anxiety-like behaviors. Indeed, activation of this circuitry elicits anorexia or anxiety-like behaviors [49, 75, 77]. However, inhibition of these neurons has no effect on basal feeding or anxiety-like behaviors [49, 77]. In addition, decades of studies have indicated that many stress hormones and neuropeptides modulate the HPA axis, dopamine system, and frontal cortical areas to elicit stress-related behaviors [2, 78].

Collectively, our results have shown the activity pattern of LSd^Sst neurons in stress responses and suggested the role of the noradrenergic signaling in this process. Further studies are required to provide an understanding of the functional role of the noradrenergic signaling in lateral septum neurons in stress responses, which may also bring clinical insights into stress-related disorders.

M. A., H. -K. K., and H. P. contributed equally. M. A., H. -K. K., H. P., C. C. and S. -Y. K. designed the project, interpreted the data, and wrote the paper with input from all authors. M. A., H. -K. K., and K. K. performed behavior, pharmacology, fiber photometry, and histology experiments with contributions from G. H. and H. -E. P., H. P. and C. C. performed ex vivo electrophysiology experiments. H. -E. P. established the fiber photometry setup. S. -Y. K. supervised all aspects of the work.