An amygdala to anterior hypothalamic circuit gates stress sensitivity

Animals

All mice were bred on a C57BL/6J genetic background and were approximately 2–6 months old at the start of testing. Male C57BL/6J mice were obtained from Jackson Laboratories for whole-brain immediate early gene imaging as well as pan-neuronal and axon terminal silencing (Jackson Laboratories, catalogue no. 000664). Studies with homozygous GAD2-IRES-Cre mice (Jackson Laboratories, catalogue no. 028867) used an equal mixture of female and male mice that were bred in house. Critically, we did not observe any sex differences in the behaviours under study, consistent with our previous work¹¹, nor did we observe any sex differences in our neuronal recordings. After initially observing no effects of sex, sex was dropped as a variable in the final reported analyses. The number of male and female mice per experiment can be found in Supplementary Table 3. Animals were housed in a temperature- and humidity-controlled vivarium on a 12–12 h light–dark cycle (lights on at 7.00 a.m.), and all handling and behavioural testing took place during the light phase. All experimental procedures were approved by the Icahn School of Medicine at Mount Sinai IACUC. Sample sizes were determined based on previous research from the laboratory using the described stress sensitization procedure¹¹ and the laboratory’s published work on calcium imaging^36,55. To the extent that it was possible, experimenters were blind to the group membership of mice. During experimentation, experimenters were blind to viral condition but could not be blinded from stress group membership or laser condition when these manipulations were ongoing. All analyses were automated and performed blind to experimental group, including calcium imaging analysis. All mice were assigned randomly to groups.

Surgery

For all surgeries, anaesthesia was induced with 5% isoflurane and subsequently kept at 1–2%. Body temperature was maintained during surgery and recovery with a heating pad below the animal, and ophthalmic ointment was applied to lubricate the eyes. All surgeries followed aseptic surgical technique. Following surgery, animals were given 20 mg kg⁻¹ ampicillin and 5 mg kg⁻¹ carprofen (sub-cutaneously) per day for 7 days and body weight and general disposition were monitored daily. Animals that underwent GRIN lens implantation were also treated with 0.2 mg kg⁻¹ dexamethasone (sub-cutaneously) for 7 days following surgery. All viruses were infused through a glass micropipette at 2 nl s⁻¹ using a Nanoject III (Drummond Scientific). Following viral infusion, an additional 5–10 min were allowed to pass before retracting the micropipette. Micropipettes were slowly retracted at 0.1 mm min⁻¹ over 3–5 min before being retracted at a faster rate. All coordinates listed below are relative to bregma.

Calcium imaging

For imaging AHN^GAD neurons, 300 nl of AAV1-syn-flex-GCaMP6f-WPRE (1.9 × 10¹³ genome copies (GC) ml⁻¹) was infused into the AHN of GAD2-IRES-CRE mice (antero-posterior (AP), −0.5; medio-lateral (ML), 0.5; dorso-ventral (DV), −5.2). For imaging BLA^AHN neurons, the same GCaMP6f virus was injected into the BLA of wildtype mice (AP, −1.6; ML, 3.1; DV, −5.25) and a cocktail of AAVrg-ef1α-Cre (7.3 × 10¹² GC ml⁻¹) and AAV5-hSyn-DIO-mCherry (7.3 × 10¹² GC ml⁻¹) was infused into the AHN. mCherry was used to localize placement in the AHN. A 0.6-mm diameter needle was then lowered slowly to 0.2 mm above the injection site and then retracted to create space for the GRIN lens (needle lowered three times). Finally, a GRIN lens (0.6 × 7.3 mm, Inscopix, catalogue no. 1050-004597) was lowered to the same coordinates and affixed to the skull using cyanoacrylate and dental cement. The lens was subsequently covered with Kwik-Sil (World Precision Instruments) followed by a thin layer of dental cement to protect the lens before baseplating. Four to 6 weeks later, while animals were under anaesthesia, the lens was uncovered and a baseplate connected to a UCLA Miniscope (v.4.4, OpenEphys) was lowered toward the lens until the optimal field of view was identified. The baseplate was then affixed to the skull with cyanoacrylate and dental cement. The Miniscope was then detached and a dust cover was connected to the baseplate to protect the lens when not imaging. For BLA^AHN animals, a headbar was also attached to the baseplate using superglue.

Optogenetics

For optogenetic silencing of AHN neurons, 150 nl of AAV5-hSyn-SIO-stGtACR1-FusionRed (2.1 × 10¹³ GC ml⁻¹), or AAV5-hSyn-DIO-mCherry (1.1 × 10¹² GC ml⁻¹), was infused at the following coordinates at a 20-degree angle: AP, −0.7; ML, 0.5; DV, −5.5. For pan-neuronal silencing, 150 nl of AAV9-hSyn-Cre-hGH (1 × 10¹² GC ml⁻¹) was co-infused. Subsequently, optic fibres (200-µm diameter, 0.5 numerical aperture, RWD Life Science) were lowered to 0.6 mm above the injection site, also at a 20-degree angle. Optic fibres were affixed to the skull with cyanoacrylate and dental cement. For optogenetic silencing of BLA^AHN terminals, 125 nl of AAV5-CaMKII(0.4)-eOPN3-mScarlet-WPRE (1.05 × 10¹³ GC ml⁻¹) or AAV5-hSyn-EGFP (1.1 × 10¹² GC ml⁻¹), was infused into the amygdala (AP, −1.4; ML, 3; DV, −5.3). Optic fibres were implanted overlying the AHN as described above.

Chemogenetics

For chemogenetic activation of AHN neurons, 100 nl of AAV5-hsyn-DIO-HM3Dq-mCherry (2.5 × 10¹³ GC ml⁻¹) or AAV5-hSyn-DIO-mCherry (4.5 × 10¹² GC ml⁻¹), was infused at the following coordinates at a 20-degree angle: AP, −0.5; ML, 0.5; DV, −5.5. For chemogenetic silencing of inputs to the AHN, a 100-nl mixture of rgAAV-efla-Cre (7.3 × 10¹² GC ml⁻¹) and AAV8-hSyn-EGFP (1.5 × 10¹² GC;ml⁻¹) was infused at the following coordinates: AP, −0.7; ML, 0.5; DV, −5.5. Separately, 150 nl of AAV5-hSyn-DIO-hM4Di (2.4 × 10¹³ GC ml⁻¹) or AAV5-hSyn-DIO-mCherry (7.3 × 10¹² GC ml⁻¹) were infused into the amygdala (AP, −1.4; ML, 3.1; DV, −5.2) or the vHC (AP, −3; ML, 3.2; DV, −4.5).

Retrograde tracing

rgAAV-hSyn-EGFP (75 nl; titre, 1.36 × 10¹³ GC ml⁻¹) was infused unilaterally at the following coordinates: AP, −0.5; ML, 0.5; DV, −5.5.

Behavioural testing

Single-housing and habituation

All animals were housed singly 1–2 weeks before beginning behavioural testing. During this time, they were handled in the vivarium for 1 min per day for 3–5 days. Furthermore, animals were habituated to transport to the laboratory for 3–5 days, where they were also handled. Animals in chemogenetic experiments were habituated to restraint three times during this time. Animals in Miniscope and optogenetic experiments were habituated to the connection to the Miniscope and/or patch cords around five times. During each of these habituation sessions, animals were connected to the Miniscope and/or patch cords and returned to their home cage with the lid off while still connected for 3–5 min. For experiments involving head fixation, in addition to the habituation procedures described above, animals were habituated to head restraint over the course of 5–6 days, beginning with 5 min and going up to 20 min.

Experimental contexts

Stressor 1 (footshock stress) and Stressor 2 (auditory stressor), as well as their respective recall sessions, took place in unique experimental contexts, consisting of highly distinct visual, olfactory, auditory and spatial cues. For Stressor 1, animals were transported from the vivarium in their cages on a cart to the experimental testing room, which was well lit and had an air filter providing ambient sound. Animals were then placed in a brightly lit experimental testing chamber with a grid floor (Med Associates) scented with 5% Simple Green solution. For Stressor 2, with the exception of the iDISCO experiment (see details in description of Stressor 2, below), animals were transported from the vivarium in P1000 pipet tip boxes and carried in a dark cardboard box to the experimental testing room, which was dark except for a dim red light. Animals were then placed inside a dark testing chamber (Med Associates) with a flat Plexiglass floor and a curved rear wall. The chamber was scented with 1% acetic acid solution.

Stressor 1

After a 5-min period of baseline exploration, animals received ten 1 s, 1 mA, scrambled footshocks, with an inter-shock interval of 30 s. Animals were taken out of the testing chamber 30 s after the last shock. For the iDISCO experiment, animals received the same number of shocks, pseudorandomly distributed over a 60-min session. When optogenetic inhibition of cell bodies was applied during Stressor 1, blue laser light was administered continuously, beginning 30 s before the first shock and continuing until the end of the session. See ‘Chemogenetic and optogenetic actuation’ for laser parameters.

Stressor 2

For most experiments, after a 3-min baseline period, animals were exposed to a single loud auditory stimulus (3 s, 125–130 dB white noise, 0 ms rise time) that was delivered by a speaker attached to the chamber wall. Animals were removed 10 s later and returned to the vivarium. For the eOPN3 experiment, mice were instead removed from the chamber after 180 s, given the potentially slow nature of metabotropic signalling. For the iDISCO experiment, Stressor 2 was administered in the home cage to avoid c-Fos evoked by transport. For calcium imaging, a series of five auditory stimuli were presented, enabling us to define whether cells reliably respond to the auditory stimulus. Here, the baseline period was 2 min and the inter-stimulus interval was 2 min. For optogenetic inhibition of AHN neurons during Stressor 2, laser light was administered continuously throughout the session. For optogenetic inhibition of amygdala to AHN terminals, light illumination began 15 s before stimulus onset and continued until the end of the session. See ‘Chemogenetic and optogenetic actuation’ for laser parameters.

Recall

For recall sessions, animals were transported to the same experimental testing chamber where the stressor was administered for an 8- to 10-min test session. When optogenetic inhibition of cell bodies was applied during Stressor 1 recall, light was administered in 2-min intervals (alternating off–on–off–on). Light-off and light-on epochs were collapsed after similar patterns were observed across the first and second off–on cycles. Because of the slow kinetics of eOPN3, for terminal inhibition experiments we instead used a 5-min baseline period followed by 5 min of laser light illumination. See ‘Chemogenetic and optogenetic actuation’ and ‘Light-sheet image acquisition and quantification’ for laser parameters.

Valence assay (with shocks and sounds)

To assess responses to shocks and sounds of various intensities in freely moving animals (recording AHN^GAD neurons), animals were placed in a perceptually distinct conditioning chamber and received a series of eight 2-s footshocks (four of each amplitude, 0.25 mA and 1 mA), and eight 2-s white noise stimuli (four of each amplitude, 100 dB and 130 dB). Stimuli were pseudorandomly intermixed such that in each block of four stimuli, every type of stimulus was experienced. The first stimulus occurred 180 s after being placed in the chamber and each subsequent stimulus was scheduled with an inter-stimulus interval of 60–100 s.

To assess responses to shocks and sounds of various intensities in head-fixed mice (recording BLA^AHN neurons), animals’ headbars were screwed to a small post and the animals sat within a cylindrical tube, approximately 2.5 inches (in.) in diameter. To administer shock, one electrical lead was taped to their tail and another was connected to a piece of aluminium foil beneath their forepaws. These leads were connected to the same scrambler used to administer footshock. Sound was delivered by a speaker mounted to the chamber wall. The animals first received a series of eight 2-s white noise stimuli (four of each amplitude, 100 dB and 130 dB), followed by eight 2-s shocks (four of each amplitude, 0.25 mA and 0.5 mA). The first stimulus occurred 180 s after being placed in the chamber and each subsequent stimulus was scheduled with an inter-stimulus interval of 60–100 s.

Valence assay (with liquids and shocks)

Across 7 days, water-deprived animals were trained to collect water from three liquid dispensers located along one of three adjacent walls in a rectangular conditioning chamber. Each dispenser consisted of a blunt, 18-gauge needle, protruding approximately 1 in. from the wall (1 in. high), connected by polyethylene tubing to a 60-ml syringe. Dispensers were programmed to produce 5 µl of water upon solenoid activation and were connected to a capacitive touch sensor to record licks. Each 15-min session began with water being dispensed at the first port (Port 1). Water would then be dispensed at the next port when water at the previous port was collected, and this occurred in a circular pattern (Port 1,2,3,1,2,3…). This pattern guaranteed that animals received an approximately equal number of liquid rewards from each port.

After training, animals were exposed to a test session in which they experienced liquids and shocks of various valence and/or palatability. During the first 15 min, Port 1 delivered water, Port 2 delivered 20% sucrose and Port 3 delivered a weak (0.5 mM) quinine solution. After 15 min collecting the liquids, a shock period began. During this period, animals received a series of 12 2-s footshocks (six 0.25-mA shocks and six 1-mA shocks), each separated by 45–75 s. Because virtually all licking stopped following the first footshock, we examined licking behaviour only before the first footshock.

Light–dark test

The light–dark test was conducted using two interconnected square compartments with an open top (each compartment measured 7.5 in. width × 11.25 in. height), separated by a 1.5-in.-wide passageway that could be closed with an opaque sliding divider. One chamber was made of all white acrylic, whereas the walls of the other were covered in matte black wallpaper and had a red acrylic floor. Overhead lighting provided luminance of 50 lux on the light side. After a 1-min baseline period in which animals were confined to the dark side, the central divider was raised, and the animals could explore freely both sides of the light–dark box. For optogenetic experiments, this period was 8 min, and light was administered at 2-min intervals (alternating off–on–off–on). Light-off and light-on epochs were collapsed after similar patterns were observed across the first and second off–on cycles. For all other experiments, animals were allowed 5 min to explore both sides of the chamber. When animals were tested more than once, test sessions were separated by 2 days.

Behaviour quantification

For analysis of freezing and motion in conditioning chambers when no fibres and/or cables were attached to the mice, Med Associates Video Freeze software was used⁵⁶. For measuring freezing and motion in chambers when cables were attached to the animals, as well as position in the light–dark test, ezTrack was used⁵⁷. Freezing was defined as the cessation of movement lasting at least a half a second in length. Pre-stress freezing (‘Pre’ in the figures) refers to freezing occurring from the start of the session until the onset of the stressor. Likewise, post-stress freezing (‘Post’ in the figures) refers to freezing occurring from the time of stressor offset until the end of the session. When several stressors were given in the same session, freezing measurements reflect the inter-stimulus periods.

Chemogenetic and optogenetic actuation

Actuation of HM4Di and HM3Dq was achieved through intraperitoneal administration of 3 mg kg⁻¹ CNO-dihydrochloride (Tocris), 30–45 min before behaviour, at a volume of 10 ml kg⁻¹ (dissolved in saline).

For stGtACR1-mediated inhibition, laser light was delivered with a 473 nM laser (OptoEngine LLC) connected through a patch cord to a bifurcating rotary joint (Doric). Consistent with previous reports that inhibited AHN neurons with stGtACR1 (ref. ³²), we used a 20-Hz, 20-ms pulse width with 5 mW, illumination. For eOPN3-mediated terminal inhibition, 473 nM laser light was pulsed at a 0.1-Hz, 500-ms pulse width with 10 mW illumination, consistent with previous reports⁴⁴. For all optogenetic experiments, light intensity was measured from a fibre tip using a light meter (PM100D with S130C attachment, ThorLabs) at the beginning of each day.

Histology

For confirmation of viral placement, GRIN lens placement and fibre-optic placement, animals were perfused transcardially with 10 ml PBS followed by 10 ml 4% paraformaldehyde (PFA). Brains were then extracted, stored in 4% PFA overnight at 4 °C, and then transferred to a 30% sucrose–PBS solution before being sectioned at 50-µm coronal sections on a cryostat and mounted on slides. Tissue was then imaged on a Leica DM6 epifluorescent microscope, with the exception of amygdala terminals, which were imaged using a confocal microscope (Leica LSM 780). Viral expression and cannula placement were evaluated using the mouse brain atlas of Franklin and Paxinos⁵⁸.

iDISCO immunostaining

Following perfusion with 10 ml of 1× PBS and 10 ml of 4% PFA, brains were extracted, any remaining dura and vasculature were dissected carefully off the brain with fine tweezers, and brains were stored overnight in 4% PFA at 4 °C. Brains were then washed three times in PBS before the iDISCO+ procedure²³. All subsequent steps were performed in 5 ml black Eppendorf tubes (fully filled to reduce air). Except where stated, all steps were done while being rotated at 10 rpm.

The general outline of staining and clearing is as follows:

Day 1: tissue was dehydrated in escalating concentrations of methanol in H₂O, for 1 h each, at room temperature (20%, 40%, 60%, 80%, 100%). After an additional 2 h in 100% methanol, brains were chilled on ice for 10 min and then incubated overnight in a solution of 66.5% dichloromethane (DCM) and 33.5% methanol at room temperature.

Day 2: tissue was washed twice in 100% methanol, 3 h each wash, then chilled at 4 °C for around 10 min. Subsequently, brains were incubated in chilled 5% hydrogen peroxide (one volume of 30% hydrogen peroxide to five volumes methanol) and kept at 4 °C overnight, without rotation.

Day 3: after allowing samples to warm to room temperature, tissue was rehydrated in decreasing concentrations of methanol in H₂O (60%, 40%, 20%), 1 h each, followed by 1 h in PBS. Subsequently, tissue was incubated twice in PTx.2 solution (0.2% Triton X-100 in PBS), 1 h each wash. Finally, tissue was permeabilized across 2 days at 37 °C in the following solution: 80% PTx.2, 20% dimethylsulfoxide, 22% glycine (w/v).

Day 5: tissue was transferred to blocking solution (84% PTx.2, 6% donkey serum, 10 ml dimethylsulfoxide) and incubated across 2 days at 37 °C.

Day 7: tissue was transferred to primary antibody (1:2,000 rabbit anti-c-Fos; Abcam catalogue no. 190289) in PTwH (0.2% Tween-20, 0.1% of 10 mg ml⁻¹ heparin solution, in PBS) plus 3% donkey serum and incubated for 1 week at 37 °C.

Days 14–15: tissue was washed four to five times in PTwH at room temperature across 2 days.

Day 16: tissue was incubated in secondary antibody (1:1,000 donkey anti-rabbit IgG Alexa Fluor 790; Invitrogen, catalogue no. A11374) in PTwH plus 3% donkey serum for 7 days at 37 °C.

Days 23–24: tissue was washed four to five times in PTwH at room temperature across 2 days.

Day 25: tissue was dehydrated in escalating concentrations of methanol in H₂O, for 1 h each, at room temperature (20%, 40%, 60%, 80%, 100%).

Day 26: tissue was incubated for 3 h in 66.5% DCM and 33.5% methanol at room temperature. After two 15-min washes in 100% DCM, tissue was transferred to 100% dibenzyl ether.

Light-sheet image acquisition and quantification

Cleared brains were imaged using a LaVision UltraMicroscope II light-sheet microscope using a ×1.3 objective lens coupled with 488-nm and 785-nm lasers for imaging autofluorescence and c-Fos, respectively. The 12-bit horizontal brain images were collected using 3.89 µm numerical aperture light-sheet width in 4.5 µm step sizes, spanning from the dorsal-most to ventral-most portion of the brain.

Light-sheet images were processed using custom Python-based scripts written in-house to identify c-Fos cell positions (https://github.com/ZachPenn/2026_Nature). Images for each animal were preprocessed as follows: two-dimensional images were first converted to three-dimensional arrays, permitting three-dimensional kernel operations. Images were then smoothed using a small median kernel, followed by subtraction of local background fluorescence, estimated using a morphological opening. Small intensity pixel fluctuations were then thresholded and set to zero. To identify c-Fos puncta in preprocessed images, a distance transform was then calculated, followed by identification of local maxima, corresponding to cell centroids. The same image processing parameters were applied to every subject and parameters were selected based upon visual inspection across several regions. Images and cell positions were then aligned to the Allen Brain Atlas²⁴ using the ClearMap2 software package^23,59.

Miniscope imaging and cell extraction

Calcium imaging was performed using the UCLA Miniscope, v.4.4 (ref. ³⁶). Parts were obtained from Open Ephys and subsequently assembled in-house. Following baseplating, animals were habituated to wearing the Miniscope over the course of several days while in their home cage. Animals were restrained lightly (held in a cupped hand), the dust cover was removed from the baseplate and the Miniscope was attached to the baseplate and locked into position with a set screw. The mice were then placed back in their home cage for several minutes with the Miniscope attached and turned on. During this time, the optimal focal plane was identified and this focal plane was held constant throughout the duration of experimental testing by maintaining the setting of the electrowetting lens. Furthermore, on each day, the focal plane and field of view were compared manually with those of the previous day. Miniscope videos were acquired at 30 frames per second.

Calcium imaging videos were processed using the open-source software, Minian⁶⁰. In brief, for each session, videos were downsampled to 15 fps and corrected for translational motion. To improve signal quality, a minimum projection for each video was then subtracted from each frame to remove vignetting, a median filter was applied to remove granular noise and a morphological opening was performed to identify—and subsequently filter out—local fluctuations in background. To identify potential cell locations, local maxima in the field of view were defined across the video. These local maxima were then used as input to the constrained non-negative matrix factorization (CNMF) algorithm. CNMF parameters were chosen for each animal based on the Stressor 2 session (when several sessions were recorded) in a visually guided manner using Minian. These parameters were then applied to every video for that animal. Parameters were selected blind to group membership. For all analyses, we used the raw signal for each cell, obtained by multiplying the spatial footprint resulting from CNMF by its temporal activity (each pixel has a weight for each cell, and the weighted average of fluorescent intensity values across time for these pixels comprise a cell’s activity).

Statistical analysis

All statistical analyses were performed in R and can be found in Supplementary Table 3. Group sizes are listed in each figure caption. More complete information regarding the breakdown of sample size by sex can be found in Supplementary Table 3. The data from each figure can found in the Source Data files. All data and statistical analyses are available at https://github.com/ZachPenn/2026_Nature. Group means were used as the primary measure of central tendency and are displayed in the figures along with the s.e.m. All statistics employ two-sided tests. Within the figures, significance is marked as follows: *P < 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Whole-brain c-Fos analysis

Within the Allen Mouse Brain Atlas²⁴, regions are classified in a nested hierarchy (for example, infralimbic layers 1 and 2 are nested within the infralimbic cortex, which is in turn nested within the isocortex). For all c-Fos analyses, the lowest available level of this hierarchy was used, providing the most granular regional information. The olfactory bulbs and cerebellum were excluded from all analyses due to frequent damage resulting from dissections. Any region that had very low c-Fos counts across groups was also excluded (median counts per group all less than 1). This left a total of 452 brain regions across which subsequent analyses were conducted.

Group differences in regional c-Fos counts

The iDISCO data presented in Fig. 1 and Extended Data Figs. 1 and 2 arise from a single experiment with three groups, consisting of animals that did not receive Stressor 2 (Control (Ctrl)), those that did not receive Stressor 1 but received Stressor 2 (Group NS), and those that received both stressors (Group S). For the sake of simplicity, data from the Ctrl group are omitted from Fig. 1 but are presented in Extended Figs. 1 and 2 and included in Supplementary Tables 1 and 2.

To compare c-Fos counts across groups, negative binomial regression was performed using the MASS package in R⁶¹. For each region, count data were modelled as a function of group (y = B₀ + B₁ × NS + B₂ × S + e). We first screened whether group membership was a significant predictor of c-Fos counts by comparing the full model listed above to an intercept only model (y = B₀ + e). As can be appreciated from Extended Data Fig. 1; this served largely to detect regions activated by Stressor 2 relative to controls given no stress at this time point, and to restrict our subsequent comparison of groups NS and S to regions that showed some evidence of activation by Stressor 2, increasing statistical power for this comparison. Critically, the control group was included from the outset of this experiment to limit the number of regions examined between our main groups of interest. Amongst the 90 regions that displayed group variation in c-Fos counts following false discovery rate (FDR) correction (across all 452 regions, Benjamini–Hochberg correction), we then assessed whether S and NS groups displayed differential c-Fos counts using post hoc contrasts with the multcomp package⁶². For this contrast, we report regions which surpass FDR correction (across the 90 regions that displayed some variation in c-Fos in our initial screen). The results of statistical tests for each region can be found in Supplementary Table 1. Summary statistics for each region and group can also be found in Supplementary Table 2. For full count information, please see the Source Data files.

Correlational analysis and hierarchical clustering

For all correlational and hierarchical clustering analyses, c-Fos counts were z-scored separately for each group. Because analyses focused on NS and S animals, which had a substantial spread of counts and moderate sample size (n = 10 per group), Pearson correlations were used as the bases for these analyses.

For hierarchical clustering, the region–region correlation matrix relating counts between brain regions (R) was first converted to an adjacency matrix (A = 1 − R), and subsequently to a set of distances using the stats package in R⁶³. Agglomerative hierarchical clustering was then performed on these distances using the average distance between clusters. Cluster thresholds were set using the ‘elbow’ method, where cluster number is increased until within-cluster variance no longer decreases substantially when increasing cluster number (Extended Data Fig. 2d). We also attempted to keep within-cluster correlations high (R ≥ 0.5).

To compare cluster ‘strength’ across groups, we initially performed hierarchical clustering on animals exposed to Stressor 1 (S), as described above. After identifying the cluster containing the AHN in these animals, we examined the correlation matrix exclusively for regions within this cluster for S, NS and Ctrl mice. Specifically, we compared the number of significant positive correlations on the off-diagonal of this matrix using Fisher’s exact test. A similar approach was taken to compare the overall correlation matrix from each group, as well as the number of correlations between the AHN and other brain regions between animals that received Stressor 1 and those that did not. Of note, although we used uncorrected significance (P = 0.05) to illustrate functional connectivity of the AHN more broadly, using FDR correction provides a similar result (for example, Fisher Exact test of number of AHN correlations between S versus NS, P < 0.01).

Calcium imaging analysis

For each cell and session, activity was normalized by first computing a z-score. For comparing responses of cells to various amplitude stimuli, activity was binned into 200-ms intervals and cells were treated as individual datapoints. Furthermore, to examine cellular responses to each stimulus more readily, peri-stimulus activity for each cell was normalized relative to the average activity in the 5 s preceding each stimulus (by subtracting this average). Accordingly, for all activity traces, data are in standard deviation units (z), with zero reflecting average pre-stimulus activity. Whenever mean activity preceding or following a stimulus is referenced, the window for this average is 5 s.

To classify whether a cell was responsive to a particular stimulus (for example, shock), the average change in activity in the 5 s following stimulus onset relative to the 5 s preceding stimulus onset was first assessed. This interval was chosen based upon visual inspection of calcium transients. A null distribution was then obtained by circularly shuffling neural activity across the session and recalculating this average difference 1,000 times. Stimuli that had a response above the 95th percentile of this null distribution were considered responsive.

Behavioural analysis

For analysis of behavioural data, omnibus ANOVA were conducted using the package ezANOVA with type 3 d.f. The white adjustment was implemented to correct for heterogeneity of variance using heteroscedasticity corrected s.e. values (‘hc3’). For repeated measures ANOVA, the Greenhouse–Geisser correction was implemented when the assumption of sphericity was not met. F and t values are rounded to the nearest tenth and hundredth, respectively. Where F values were less than 0.1, F is listed as 0.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.