Central neurons encode interleukin-1β signals and mediate stress-induced inflammation

The brain continuously monitors peripheral inflammatory states and coordinates adaptive physiological responses to optimize host defense, protect against tissue injury, and promote recovery (Dantzer, 2018; Kipnis, 2018; Pavlov et al., 2018; Sammons et al., 2024). Through specialized sensory pathways, including the vagus nerve, the central nervous system receives information about inflammatory mediators and responds by orchestrating sickness behaviors and immunomodulatory responses (Maier et al., 1998; Salvador et al., 2021; Silverman et al., 2023). This bidirectional communication between the nervous and immune systems operates through discrete neural circuits that maintain homeostatic control over inflammatory processes, preventing both inadequate responses to threats and excessive inflammation that damages healthy tissue.

Among the molecular messengers that link peripheral inflammation to central neural responses, cytokines serve as critical signaling molecules that inform the brain about the nature and magnitude of inflammatory challenges (Salvador et al., 2021). We and others have demonstrated that specific cytokines activate distinct populations of vagus sensory neurons in a cytokine-specific manner, transmitting signals that inform the brain of emerging inflammatory states (Huerta et al., 2025; Jin et al., 2024; Steinberg et al., 2016; Zanos et al., 2018). This cytokine-specific neural communication represents a fundamental mechanism by which the brain encodes information about peripheral immune challenges and subsequently modulates systemic inflammatory responses through efferent neural pathways.

IL-1β, a pivotal inflammatory mediator, acts both locally to amplify immune responses and systemically to signal the brain via specific neural pathways. Peripheral administration of IL-1β elicits characteristic sickness behaviors including fever, reduced food intake, and cardiovascular changes, which are abrogated by IL-1 receptor antagonists or vagotomy, demonstrating vagus nerve dependence of these responses (Gaykema et al., 2000; Goehler et al., 2000; Maier et al., 1998; Watkins et al., 1995). Previous work from our laboratory established the inflammatory reflex, a neural circuit wherein vagal sensory neurons detect inflammatory mediators and relay this information to brainstem nuclei, which in turn activate efferent pathways that terminate in the spleen to suppress cytokine production (Borovikova et al., 2000; Chavan et al., 2017; Kressel et al., 2020; Rosas-Ballina et al., 2008; Tracey, 2002; Zanos et al., 2018). This prototypical neuroimmune circuit demonstrates how peripheral inflammatory signals are encoded in neural activity and translated into precise physiological responses. However, the circuit organization for specific cytokines, like IL-1β, and the specific neural populations responsible for encoding cytokine-specific information, have not been defined.

The bed nucleus of the stria terminalis (BNST) represents a compelling candidate brain region for cytokine signal integration and stress response coordination (Dantzer et al., 2008; Kim et al., 2013). This neurochemically diverse structure contains multiple GABAergic and glutamatergic neuronal subtypes, each expressing distinct neuropeptides, including corticotropin-releasing hormone (CRH), which plays central roles in orchestrating stress responses (Lebow and Chen, 2016; Ortiz-Juza et al., 2021). The BNST receives inputs from multiple brain regions involved in threat detection and stress processing and projects to hypothalamic and brainstem areas that coordinate physiological stress responses (Barbier et al., 2021; Dabrowska et al., 2016; Huang et al., 2021). Importantly, CRH-expressing BNST neurons have been implicated in cardiovascular responses to stress and in anxiety-related behaviors, suggesting they may serve as critical integrators of inflammatory and psychological stressors (Nijsen et al., 2001).

Here we used targeted recombination in active populations (TRAP) and whole-brain serial two-photon tomography (STPT) to map IL-1β–responsive neurons. Single-nucleus RNA sequencing (snRNA-seq) identified CRH-expressing neurons in the BNST. Chemogenetic reactivation of these CRH⁺ BNST neurons recapitulates the physiological signature of IL-1β, including cardiovascular changes and cytokine production. These neurons respond to both inflammatory and psychological stressors and are required for stress-induced inflammatory responses through a BNST→paraventricular nucleus→β receptor adrenergic pathway, revealing cytokine-responsive neural circuits as fundamental organizing principles of neuroimmune communication.

Systemic IL-1β administration induces characteristic physiological responses, including alterations in core body temperature, heart rate, circulating cytokines, and corticosterone levels. We confirmed that i.p. injection of IL-1β (5 μg/kg) in mice elicits significant changes across these physiological responses within 2 h (Fig. S1, A–F). To systematically map brain regions activated by peripheral IL-1β, we employed TRAP methodology combined with whole-mount STPT. TRAP2-tdTomato mice were injected with 4-hydroxytamoxifen (4-OHT) followed by either IL-1β or vehicle (PBS), resulting in permanent tdTomato labeling of IL-1β–activated neurons. STPT imaging generated complete 290-section coronal datasets that were analyzed using automated cell counting algorithms (Fig. 1 A). IL-1β administration activated multiple brain regions with established roles in inflammatory signaling, including the nucleus tractus solitarius, the paraventricular nucleus of the hypothalamus (PVN), and the BNST (Fig. 1, B and C; and Fig. S1 G). Notably, the BNST exhibited the most robust activation following IL-1β exposure as compared with control conditions, primarily within the dorsal region (Fig. 1 C).

To determine whether BNST neurons encode IL-1β–specific information, we combined TRAP methodology with chemogenetic reactivation. IL-1β–responsive BNST neurons were captured using bilateral stereotaxic injection of Cre-dependent hM3Dq-mCherry AAV followed by IL-1β administration (Fig. 1 D). Viral targeting specificity was confirmed through postmortem analysis of mCherry expression co-localized with c-Fos activation following clozapine N-oxide (CNO) administration (Fig. 1 E). Chemogenetic reactivation of IL-1β–TRAPed BNST neurons with CNO precisely reproduced the physiological signature of systemic IL-1β exposure. These manipulations induced significant hypothermia, with core body temperature decreasing over 6 h (Fig. 1, F and G), concurrent tachycardia lasting 60 min (Fig. 1, H and I), and significant increases in circulating IL-6 (from 26.2 + 8.3 to 275.5 + 74.1 pg/ml) and corticosterone levels (Fig. 1, J and K). These responses were indistinguishable from those produced by peripheral IL-1β administration (Fig. S1, A–F), demonstrating that IL-1β–responsive BNST neurons are sufficient to activate the physiological program triggered by this cytokine. Importantly, CNO alone did not change serum IL-6 levels and heart rate (Fig. S1, H and I). To assess whether these neurons are required for IL-1β responses, we genetically ablated IL-1β–TRAPed BNST neurons using Cre-dependent diphtheria toxin A (dtA) virus (Fig. S2 A). Targeted ablation significantly attenuated IL-1β–induced tachycardia (Fig. S2 B) while leaving temperature, cytokine, and corticosterone responses intact (Fig. S2, C–E), indicating that BNST neurons are specifically required for this cardiovascular component of IL-1β responses, with parallel pathways mediating other physiological effects of systemic administration of IL-1β.

The BNST contains neurochemically diverse cell populations that coordinate distinct aspects of stress and anxiety responses (Ortiz-Juza et al., 2021). To identify which BNST neuronal subtypes mediate IL-1β responses, we performed snRNA-seq of the entire BNST, capturing 11,564 individual cells with neuronal and five nonneuronal populations (Fig. S3 A). Louvain clustering analysis revealed 17 distinct neuronal clusters (9,348 neurons total) (Fig. 2 A). As expected, GABAergic neurons (Slc32a1/Vgat expressing) comprised the vast majority of BNST cells compared with glutamatergic neurons (Slc17a6/Vglut2 expressing) (Fig. 2 B). Analysis of neuropeptide expression patterns identified four major BNST neuronal subtypes: crh, somatostatin (sst), prodynorphin (pdyn), and preproenkephalin (penk)-expressing populations (Fig. 2 B and Fig. S3 B). To determine whether specific subsets mediate IL-1β responses, we performed chemogenetic activation experiments using Cre-dependent hM3Dq-mCherry virus in corresponding transgenic mouse lines.

Selective activation revealed distinct physiological profiles across BNST neuronal subtypes. Sst⁺ and pdyn+ neurons induced robust tachycardia and corticosterone responses but no IL-6 response, while penk⁺ neurons triggered IL-6 elevation without affecting heart rate (Fig. 2, C–E). Remarkably, Crh⁺ neuron activation precisely recapitulated the dual cardiovascular and inflammatory responses observed with IL-1β–TRAPed neuron reactivation: significant increases in both heart rate and circulating IL-6 levels (Fig. 2, C and D). Immunohistochemical analysis confirmed CRH expression in IL-1β–TRAPed BNST neurons (Fig. S3 C), establishing CRH⁺ BNST neurons as the specific population that encodes IL-1β information and orchestrates its physiological responses.

Inflammatory stressors activate catecholamine release through sympathetic and adrenal pathways (Goldstein, 2021; Nance and Sanders, 2007). Prior work (Morimoto et al., 1992) has shown that IL-1β administered i.p. within the range used in our experiments (5 μg/kg) increased circulating norepinephrine and activated tachycardia. Further, Mota and Madden (2022) have shown that the neural circuits controlling the response of core body temperature, cutaneous responses, and heart rate depend upon activation of the sympathetic nervous system. Based on these prior observations, we hypothesized that the tachycardia induced following direct chemoactivation of IL-1β–responsive BNST neurons depends upon catecholaminergic activation. To determine whether IL-1β–responsive BNST neurons utilize adrenergic signaling, we performed chemogenetic reactivation experiments in the presence of β-adrenergic receptor (AR) antagonists. Both IL-1β–TRAPed BNST neurons (Fig. 3, A–E) and CRH⁺ BNST neurons (Fig. 3, F–J) produced identical results: propranolol (5 mg/kg i.p.; β1/β2 AR antagonist) completely blocked both tachycardia and IL-6 responses, while SR59230A (5 mg/kg i.p.; β3 AR antagonist) had only a minor effect. Neither antagonist affected corticosterone responses, indicating that different signaling pathways mediate distinct components of the IL-1β response pattern. Together with the above results, these findings demonstrate that CRH⁺ BNST neurons activate cardiovascular and inflammatory responses through β1/β2-adrenergic signaling, contrasting with previous reports of β3-adrenergic dependence for stress-induced cytokine production (Qing et al., 2020).

The PVN serves as a critical integration hub for stress responses, coordinating autonomic, endocrine, and inflammatory outputs (Poller et al., 2022). To investigate whether IL-1β–responsive BNST neurons function through BNST–PVN connectivity, we performed comprehensive circuit mapping using multiple viral tracing strategies. As expected, anterograde tracing with Cre-dependent EYFP and synaptophysin-mRuby in IL-1β–TRAPed BNST neurons revealed robust fiber projections and synaptic terminals throughout the PVN, but no projections from PBS-TRAPed BNST neurons (Fig. 4, A and B). Transsynaptic connectivity was confirmed using dual viral approaches (Fig. 4 C): Cre-dependent Flp expression in IL-1β–TRAPed BNST neurons combined with Flp-dependent EYFP in the PVN-labeled axonal projections in the rostral ventral lateral medulla (RVLM) but not to the NTS, as well as postsynaptic PVN targets (Fig. 4 D). Additionally, we confirmed that reactivation of IL-1β–responsive BNST neurons results in c-Fos expression in the RVLM (Fig. 4 E). To test functional connectivity, we selectively targeted PVN neurons receiving IL-1β–responsive BNST inputs using Cre-dependent Flp in the BNST and Flp-dependent hM3Dq-mCherry in the PVN (Fig. 4, F and G). Chemogenetic activation of these BNST-connected PVN neurons reproduced key elements of IL-1β responses, including increased heart rate, circulating IL-6, and corticosterone levels (Fig. 4, H–K), demonstrating that the BNST–PVN pathway is sufficient to drive IL-1β–like physiological responses.

To establish necessity of these functional circuits, we ablated IL-1β–responsive PVN neurons with bilateral stereotaxic injection of Cre-dependent dtA virus while preserving BNST neurons for reactivation. This circuit-specific disruption completely abolished BNST-mediated increases in IL-6, heart rate, and corticosterone (Fig. 5, A–F), proving that PVN signaling is required for IL-1β–responsive BNST neuron function. We further tested this circuit requirement using CRH-specific manipulations. Genetic ablation of CRH+ PVN neurons while activating CRH⁺ BNST neurons significantly attenuated all physiological responses (Fig. 5, G–L), establishing that CRH⁺ BNST–PVN signaling constitutes an essential neural circuit for IL-1β–mediated cardiovascular, inflammatory, and stress responses.

Given the established role of BNST in stress processing (van de Poll et al., 2023) and the overlap between inflammatory and psychological stress responses, we hypothesized that IL-1β–responsive BNST neurons might also mediate responses to psychological stressors. Acute restraint stress (4 h) produces physiological responses identical to IL-1β administration: tachycardia, elevated IL-6, and increased corticosterone (Fig. S4, A–E). Furthermore, acute restraint stress also strongly induced neuronal activity in the BNST regions (Fig. S4 F). We found that IL-1β stimulation activated the same BNST neurons activated by acute restraint stress (Fig. S4 G). Genetic ablation of IL-1β–TRAPed BNST neurons significantly reduced restraint stress responses, decreasing IL-6 levels by 67% and heart rate responses by 94% while preserving corticosterone elevation (Fig. 6, A–E). Importantly, baseline heart rate remained unaffected (Fig. S5 A), indicating specific roles in stress-induced rather than basal cardiovascular regulation. Parallel experiments using CRH⁺ BNST neuron ablation yielded identical results: eliminated inflammatory and cardiovascular responses to restraint stress while preserving HPA axis activation (Fig. 6, F–J), while baseline heart rate was similarly unaffected (Fig. S5 B). These findings reveal that IL-1β–responsive CRH⁺ BNST neurons serve as a convergent neural population that encodes both inflammatory cytokine signals and psychological stress, translating these inputs into shared cardiovascular and inflammatory output patterns while maintaining distinct pathways for glucocorticoid responses.

These findings establish the first cytokine-specific neural circuit with defined physiological readouts, demonstrating that CRH-expressing neurons in the BNST encode IL-1β signals and orchestrate precise cardiovascular and inflammatory responses through a BNST→PVN→RVLM→β receptor adrenergic pathway. Our results reveal that these same neurons serve as a convergent integration point for both inflammatory cytokine signals and psychological stress, providing a mechanistic foundation for understanding how immune challenges and stress converge on shared physiological outputs while maintaining circuit specificity for distinct response components.

The identification of CRH⁺ BNST neurons as IL-1β–responsive elements aligns with decades of research establishing these neurons as master regulators of stress responses and autonomic function (Hammack et al., 2021; Horvath et al., 2023; Maguire, 2014). The extensive CRH literature has consistently demonstrated that BNST CRH⁺ neurons integrate multiple stressors and coordinate complex physiological outputs, including cardiovascular responses, anxiety behaviors, and neuroendocrine activation (Laryea et al., 2012; Silberman and Winder, 2013). Our finding that BNST is the most prominently activated region identified in whole-brain mapping following IL-1β administration was initially surprising, given the established focus on brainstem circuits in neuroimmune signaling (Ilanges et al., 2022; Jin et al., 2024). However, our subsequent demonstration that chemogenetic activation of these IL-1β–responsive BNST neurons produces tachycardia validates this established framework while extending it to encompass inflammatory signaling. Previous studies showing CRH⁺ BNST neurons regulate heart rate during psychological stress (Nijsen et al., 2001) directly parallel our observations of IL-1β–induced cardiovascular responses, suggesting these neurons represent a conserved circuit architecture for threat detection and physiological mobilization.

The BNST–PVN circuit we define operates through established neural pathways that have been extensively characterized in stress neurobiology. CRH⁺ BNST neurons project robustly to the PVN, where they influence both autonomic outflow and hypothalamic–pituitary–adrenal axis activation (Aguilera and Liu, 2012; Becker, 2018). Our demonstration that this circuit is both necessary and sufficient for IL-1β responses provides a mechanistic basis for the long-recognized overlap between stress and inflammatory physiology. In addition to the critical role of the PVN in stress responses, it is well known that stress also activates the sympathetic nervous system via the RVLM as a central control point to increase sympathetic outflow to control multiple physiological functions, e.g., cardiac function (Lamotte et al., 2021). As prior work has shown that IL-1β administration causes the release of norepinephrine, our observations have identified a neural circuit whereby IL-1β–responsive neurons residing within the BNST initiate an adrenergic response.

Unexpectedly, the pharmacological experiments revealed that IL-1β–responsive circuits depend predominantly on β1/β2-adrenergic signaling rather than a β3 pathway, contrasting with a previous report of β3-adrenergic dependence for stress-induced cytokine production by brown adipose tissue (Qing et al., 2020). The observation that the β3 antagonist SR59230A exhibits partial efficacy in blocking the effects of IL-1β on heart rate and IL-6 production may indicate the existence of an alternative pathway remaining to be defined. Alternatively, this effect could result from an incomplete β3 specificity for SR59239A, as prior work has shown a reduction in heart rate following systemic injection of this molecule in the amounts used in this study (Gill et al., 2012). Further work is required to determine whether one of these possibilities may explain the current observations.

Additional prior work has also shown that restraint or hypoxic stress activates vasopressin- and IL-6–producing magnocellular neurons located in the PVN and supraoptic nuclei to release IL-6 directly into the systemic circulation via the neurohypophysis (Jankord et al., 2010). Our study identifies the BNST as the upstream component of this neural circuit, which utilizes a different β receptor AR for signaling than does brown adipose tissue. These two different sources of IL-6 activated by stress highlight the specificity of cytokine-responsive circuits within the broader stress architecture and answer a longstanding question about how different stressors might engage distinct downstream effector mechanisms despite activating overlapping central neural populations.

These findings position IL-1β–responsive CRH⁺ BNST neurons within a hierarchical organization of neuroimmune circuits. Peripheral inflammatory signals detected by cytokine-sensitive vagal afferents are relayed to brainstem nuclei, which in turn activate forebrain circuits, including the BNST. This central processing layer integrates inflammatory information with other threat-related signals before activating downstream effector pathways that coordinate systemic responses. A particularly striking result was our discovery that among the four major neuropeptide-expressing BNST populations we examined, only CRH+ neurons could recapitulate both the cardiovascular and inflammatory components of IL-1β responses. While SST⁺ and PDYN⁺ neurons induced tachycardia and PENK⁺ neurons elevated IL-6, the unique dual capability of CRH⁺ neurons reflects their established role in orchestrating multisystem stress responses and provides definitive evidence that inflammatory signals specifically engage the neural machinery evolved for comprehensive threat responses.

These results further reveal that IL-1β–responsive CRH⁺ BNST neurons function as a neural engram for inflammatory memory, consistent with recent advances in understanding how specific experiences are encoded in discrete neural populations (Gogolla, 2021; Koren et al., 2021). Similar to classical memory engrams that can be reactivated to retrieve behavioral responses in the absence of original stimuli, chemogenetic reactivation of IL-1β–TRAPed neurons produces the complete physiological signature of cytokine exposure without peripheral IL-1β administration. The precision of this recall was remarkable: reactivation produced a 10-fold increase in circulating IL-6 along with hypothermia, tachycardia, and elevated corticosterone, responses that were indistinguishable from those produced by peripheral IL-1β administration. This represents a form of an “immune engram” where inflammatory experiences are encoded in neural circuits and can be recalled through specific population reactivation, extending the engram concept beyond traditional learning and memory paradigms to encompass physiological state representations (Koren and Rolls, 2022).

This organization follows principles of labeled line coding, where specific sensory inputs are processed through dedicated neural pathways that maintain signal fidelity from periphery to central representation. This framework aligns with the Zuker group’s recent demonstration that discrete populations of vagal sensory neurons exhibit cytokine-specific responses, with individual nodose ganglia neurons selectively encoding IL-1β, TNF, and IL-10 through distinct neural activity patterns that relay inflammatory information to brainstem nuclei (Jin et al., 2024). Our present results reveal that IL-1β engages a dedicated forebrain circuit downstream of this vagal-brainstem axis, ensuring that inflammatory cytokine signals maintain their specificity throughout neural processing and enabling precise physiological responses tailored to specific immune challenges.

These findings provide a mechanistic foundation for understanding the bidirectional relationship between inflammation and stress disorders. The convergence of inflammatory and psychological stress signals on identical neural populations offers a circuit-level explanation for why inflammatory diseases are associated with increased rates of depression and anxiety and why psychological stress exacerbates inflammatory conditions (Prescott and Angus, 2018; Song et al., 2018). Perhaps the most clinically relevant finding was our demonstration that genetic ablation of IL-1β–responsive BNST neurons dramatically reduced restraint stress responses, decreasing IL-6 levels by 67% and heart rate responses by 94% while preserving corticosterone elevation. This dissociation addresses a fundamental question in stress biology by showing that inflammatory and glucocorticoid responses are mechanistically separated, suggesting that targeted interventions affecting BNST function might modulate pathological inflammation without necessarily disrupting adaptive stress responses mediated through parallel circuits. Finally, it should be noted that in contrast to the large effects following ablation of IL-1β–responsive BNST neurons on restraint stress–induced heart rate and IL-6 changes (Fig. 6), only the heart rate response was affected following administration of IL-1β (Fig. S2). This discordant result is likely explained by the fact that systemically administered IL-1β interacts with multiple nonnervous system targets, whereas restraint stress acts centrally to increase IL-1β within the brain (Shintani et al., 1995) and does not lead to IL-1β increases in the periphery (Qing et al., 2020).

One remaining issue concerns the electrophysiology of the BNST→PVN circuit. Namely, the vast majority of CRH⁺ BNST neurons are of an inhibitory GABAergic phenotype (Dabrowska et al., 2013) (Fig. 2 B), and how these could provide excitatory input to the PVN is currently unclear. There are several possible explanations. First, and more likely, a small minority (∼3%) of BNST neurons are glutamatergic (Ortiz-Juza et al., 2021), and these could mediate the observed effect of activation of PVN-mediated stress responses by BNST neurons. A second potential explanation could be that GABAergic BNST projections may disinhibit PVN excitatory neurons through inhibition of local inhibitory interneurons, yielding net excitation despite inhibitory neurotransmission. This mechanism has precedent in multiple brain neural pathways (Letzkus et al., 2015; Myers-Joseph et al., 2024; Wang et al., 2021). A third possibility could be that CRH release from BNST terminals may modulate PVN activity and neuroimmune output independent of GABA, as both CRH and urocortin can modulate PVN targets. Importantly, it is known that CRH can exert context-dependent pro-inflammatory or anti-inflammatory actions and well-described excitatory effects on PVN neurons through CRHR1 (Agelaki et al., 2002; Wlk et al., 2002; Zhang et al., 2023). An alternative potential explanation could be in the setting of stress, chloride ion gradients in GABAergic neurons are altered, resulting in a reversal of the ion gradient by downregulation of the transmembrane anion transporter KCC2, thereby increasing the activity of target neurons (Hewitt et al., 2009; Maguire, 2014; McArdle et al., 2023). This possibility seems less likely given that the physiological and immune responses activated by the PVN did not increase following BNST neuronal ablation, which would be expected if inhibitory input was eliminated. Thus, further study will be necessary to determine the mechanism by which PVN neurons are activated by CRH⁺ BNST neurons.

In summary, by demonstrating that CRH⁺ BNST neurons transform IL-1β signals into precise physiological responses through a defined neural circuit, this work provides a mechanistic foundation for the integration of inflammatory and stress physiology. The complete silencing of BNST-mediated responses following PVN ablation establishes that this BNST→PVN→RVLM →adrenergic pathway is an essential neural circuit for cytokine-responsive cardiovascular and inflammatory responses, resolving longstanding questions about the central neural substrates underlying neuroimmune communication. These findings also establish a new framework for understanding the inflammatory component of stress responses.

All experimental procedures involving animals were approved by the Institutional Animal Care and Use Committee and the Institutional Biosafety Committee of the Feinstein Institutes for Medical Research, Northwell Health, Manhasset, NY, USA, in accordance with National Institutes of Health (NIH) guidelines. Animals were housed under controlled conditions at a temperature of 25°C, with a 12-h light–dark cycle, with free access to food and water. C57BL/6J mice (stock #000664; The Jackson Laboratory), TRAP2-transgenic mice (stock #030323; The Jackson Laboratory), crh-cre (stock #012704; The Jackson Laboratory), sst-cre (stock #018973; The Jackson Laboratory), pdyn-cre (stock #027958; The Jackson Laboratory), penk-cre (stock #025112; The Jackson Laboratory), and Ai14D transgenic mice (stock #007908; The Jackson Laboratory) were purchased from Jackson Laboratory and maintained in fully accredited facilities at the Feinstein Institutes for Medical Research. For visualization of active neurons, TRAP2 mice were crossed with Ai14D mice (TRAP2-tdTomato mice). Male mice aged between 6 and 16 wk were used in these studies.

For chemogenetic studies, either AAV-hSyn-DIO-hM3D(Gq)–mCherry (cat #44361; Addgene) or AAV-hSyn-fDIO-hM3D(Gq)-mCherry-WPREpA (cat#154868; Addgene) was used. For the ablation studies, either AAV-EF1a-mCherry-flex-dtA (cat # NRZP-0622-ZP616; Creative Biolabs), or AAV-hSyn-DIO-EGFP (cat #50457; Addgene) was utilized. For the tracing studies, either AAV-hSyn-DIO-EGFP (cat #50457; Addgene), AAV-hSyn-FLEx-mGFP-2A-Synaptophysin-mRuby (cat# 71760; Addgene), AAV-Ef1a-fDIO-EYFP (cat# 55641; Addgene), or AAV pEF1a-DIO-FLPo-WPRE-hGHpA (cat# 87306; Addgene) was utilized. All used virus titers were in the range of 2–3 × 10¹³ genome copies per ml.

Recombinant and tag-free human IL-1β were produced in-house. Following expression in Escherichia coli, IL-1β was purified using a cation exchange column, and endotoxins were removed by phase separation with Triton X-114. IL-1β (5 μg/kg) was administered i.p., followed by blood collection via cardiac puncture, as indicated in the corresponding figures. Serum cytokines were analyzed using ELISA kits (Invitrogen).

Acute restraint stress was initiated by placing mice in a well-ventilated 50-ml conical tube with multiple holes without food and water supply for a duration of 4 h. After 4 h of restraint stress, blood was collected by cardiac puncture, and tissues were collected for evaluation. Serum cytokines after the restraint stress were analyzed using a customized highly sensitive U-PLEX assay platform for IL-6 and MCP-1 (Meso Scale Discovery).

TRAP2 mice were utilized for viral injections. In brief, the mice were deeply anaesthetized with 2.5% isoflurane, positioned in a stereotaxic frame, and maintained at 2% isoflurane during the surgical procedure. A middle incision was made from above the forehead to the ear line, and a craniotomy was performed above the designated target regions. 200 nl of viruses were microinfused bilaterally through a glass micropipette, connected with a Nanoliter Injector (Nanoliter 2020, World Precision Instruments) at a rate of 50 nl min⁻¹ into the dorsal BNST (anterior/posterior (AP) +0.15; medial/lateral (ML), dorsal/ventral (DV) ±1.10; DV −4.30) and PVN (AP −0.7; ML ±0.15; DV −4.8). Following the injection, the micropipette was left in place for 5 min to allow for virus diffusion before cautiously withdrawing the pipette. As a sham operation, a glass micropipette was injected into the dorsal BNST and left in place for 5 min. Mice were allowed to recover for at least 2 wk.

4-OHT (Sigma-Aldrich) was dissolved at a concentration of 20 mg/dl in ethanol. Prior to utilization, 4-OHT was heated to 37°C for 15 min and suspended in Chen Oil (a 1:4 mixture of castor oil: sunflower seed oil) at a final concentration of 10 mg/ml. Following the evaporation of the ethanol under vacuum centrifugation, a final dose of 50 mg/kg of the 4-OHT was injected i.p., 30 min prior to cytokine injections. Brains were harvested after 7 days following cytokine administration.

Animals were perfused transcardially with ice-cold PBS followed by 4% paraformaldehyde (PFA). After 24-h post-fix in 4% PFA, the brains were washed and stored in 0.05 M phosphate buffer at 4°C. STPT imaging was performed as previously described (Ragan et al., 2012). The brains were embedded in 4% oxidized agarose in 10 mM sodium periodate. The agarose block was cross-linked in a 0.5% sodium borohydride solution for 3–4 h at room temperature. The brains were imaged with a high-speed multiphoton microscope with integrated vibratome sectioning with 290 sections of the whole brain over 50-μm intervals (TissueCyte 1000, TissueVision). The brain images were stitched together to create a high-resolution image of each individual brain section. The images are registered to the Allen Mouse Brain Common Coordinate Framework (CCFv3) to outline 508 distinct brain regions along all 290 sections. ImageJ software was used to detect and count activated neurons in each brain section. Coordinates of each individual neuron were extracted and were overlayed to the area of each brain region present through that brain section using MATLAB software to extract the neuronal count of each individual brain region.

A single i.p. injection of CNO (5 mg/kg; Sigma-Aldrich) dissolved with DMSO in in sterile saline or saline with DMSO (control) is given. 2 h after CNO injection, animals were euthanized, and blood samples were collected by cardiac puncture. Cytokines in the blood were measured using commercially available ELISA kits (Invitrogen), following the manufacturer’s instructions.

Heart rate was monitored in anesthetized mice after administration of cytokines, CNO, or adrenergic antagonists (propranolol: 5 mg/kg, #P0884; Sigma-Aldrich, SR59203A: 5 mg/kg, # 21407; Cayman Chemical). Mice were anesthetized using isoflurane at 2.5% in 100% oxygen at a flow rate of 1 L/min and maintained in a supine position at 1.0% isoflurane. Heart rate was monitored in anesthetized mice using a small animal physiological monitor system (ST2 75-1500; Harvard Apparatus), which also allowed for the maintenance of the animal’s body temperature at 37°C. The heart rate was monitored for 10 min before injecting cytokines, CNO, or adrenergic antagonists, and then monitored for an additional 60 min.

Heart rate was monitored in awake mice using a telemetry system in restraint stress experiments. Mice were induced with general anesthesia using isoflurane at 2.5 in 100% oxygen at a flow rate of 1 L/min and maintained in a supine position at 2.0% isoflurane. A midline incision was made, and an ETA-F10 implant (Data Science International) was placed in the peritoneal cavity, tacked to the peritoneal wall. After a minimum 5-day recovery period, the mice were placed onto the Data Science International receiver. Data were collected and analyzed using Ponemah 6.51 software (Data Science International). For restraint stress experiments, mice were recorded for 4 h during their active cycle in the restraint model.

To perform snRNA-seq on the BNST, we isolated single cells from the BNST. Briefly, mice were euthanized with CO₂ asphyxiation and transcardially perfused with 20 ml in ice-cold HEPES-ACSF (119 mM NaCl, 2.5 mM KCl, 1.3 mM MgCl₂, 2.5 mM CaCl_2, 1 mM NaH₂PO₄, 11 mM glucose, and 20 mM HEPES; PH 7.4). The brains were rapidly isolated and sliced into three consecutive sections beginning at +0.38 mm Bregma with 300 μm thickness on a vibratome. 10 punches (0.5 mm each, 15113-50; Ted Pella, Inc.) were isolated from the BNST and frozen in liquid nitrogen. The frozen tissue was processed by the Cold Spring Harbor Laboratory genome core to isolate and barcode individual nuclei using the 10X Genomics Chromium system. Single-cell gene expression libraries were prepared using the Single Cell 3′ Gene Expression kit version 3.1 (#1000268; 10X Genomics) according to the manufacturer’s instructions. Libraries were sequenced on an Illumina NextSeq 2000 to a mean depth of ∼23,000 reads per cell.

Demultiplexed FASTQ files were mapped to the mouse reference genome (“mm10-2020A” GRCm38/GENCODE vM23, 10X Genomics) using Cell Ranger version 7.1.0 (10X Genomics) and default parameters. Unambiguously mapped intronic reads were counted in the gene expression matrix. Quality control analysis was performed by removing the reads corresponding to ribosomal and mitochondrial genes, resulting in 11,612 cells that were used for further analyses.

Analysis of snRNA-seq data, including the generation of cell clusters and identification of neuronal cluster markers, was carried out using Seurat (5.1.0) and R Studio (4.2.0).

For immunostaining, TRAP2-tdTomato or TRAP2 mice were euthanized with CO₂ asphyxiation, then transcardially perfused with cold PBS followed by chilled 4% PFA. The brain was dissected and postfixed in 4% PFA solution at 4°C overnight. 50 μm free-floating sections were prepared using a vibratome (Leica VT1200S; Leica Microsystems) and blocked (3% goat serum, 0.1% Triton X-100) for 1 h, followed by incubation with primary antibodies at 4°C for 3 days. After washing with PBS, sections were incubated with secondary antibodies for 1 h at room temperature. Sections were mounted using Fluoromount-G with DAPI (Southern Biotech) and imaged with BZ-X810 All-in-One Microscope (Keyence) and confocal microscope (Zeiss LSM880; Zeiss), with subsequent analysis performed via FIJI software. Primary and secondary antibodies were diluted in a mixture of 1% goat serum and 0.1% Triton X-100 as follows: anti-CRH rabbit antibody (#10944-1-AP; Proteintech) at 1:500, anti-mCherry rat antibody (#M11217; Thermo Fisher Scientific) at 1:500, Alexa Fluor 488-conjugated anti-rabbit gout antibody (#A21109; Thermo Fisher Scientific) at 1:1,000, and Alexa Fluor 594-conjugated anti-rat goat antibody (#A11007; Thermo Fisher Scientific). Manual outlining of brain areas was conducted with guidance from local structural markers, anatomical landmarks, and a comprehensive mouse brain atlas. Quantification of tdTomato-positive cells and colocalization was performed using ImageJ software. The average number of tdTomato-positive cells (three to five consecutive sections/brain region/mouse) was used in the analysis. The degree of colocalization between tdTomato and cFos was calculated based on Manders’ coefficient, which ranges from 0 (no overlap) to 1 (100% colocalization) between the two images.

All statistical tests were carried out with GraphPad Prism 9 software (GraphPad). Data are represented as individual data points with box and whisker plots showing the minimum, maximum, median, and 25th and 75th percentiles. Statistical analysis of mean differences between groups was performed using two-way ANOVA, one-way ANOVA, and Welch’s t test. Comparison between two or three groups over multiple time points in heart rate was analyzed using a Mixed-Effects ANOVA; a Šidák correction was used when making multiple comparisons. For all analyses, P ≤ 0.05 (two-tailed) was considered statistically significant.

Fig. S1 shows characterization of immune responses after IL-1β injection. Fig. S2 shows the analysis of mice with ablation of IL-1β–TRAPed BNST neurons after IL-1β injection. Fig. S3 shows additional snRNA-seq data analysis used in Fig. 2. Fig. S4 shows inflammatory, cardiovascular, and corticosterone responses to acute stress conditions. Fig. S5 shows the additional data related to Fig. 6.

The data that support the findings of this study are available from the authors upon request.

The authors would like to thank Dr. Tea Tsaava and Dr. Carlos E. Bravo-Iniguez for their help in animal experiments and assays. The authors would like to acknowledge the Cold Spring Harbor Laboratories Core Facilities, particularly Jon Preall, Claire Reegan, and Sanjeev Kaushalya, for their guidance and expertise in performing these experiments.

This study was supported by grants from the National Institute of Health: National Institute of General Medical Sciences (R01GM132672 to S.S. Chavan and R35GM118182 to K.J. Tracey); National Institute of Arthritis and Musculoskeletal and Skin Diseases (R01AR083159 to K.J. Tracey and S.S. Chavan, multi-principle investigator [MPI]); and funding from Bill and Melinda Gates Foundation (INV-051750 to S.S. Chavan and K.J. Tracey, MPI).

Author contributions: Okito Hashimoto: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, and writing—original draft, review, and editing. Tyler D. Hepler: data curation, formal analysis, investigation, methodology, and validation. Aisling Tynan: formal analysis, investigation, and methodology. Alejandro Torres: data curation, formal analysis, investigation, and software. Jian Hua Li: methodology and resources. Michael Brines: conceptualization and writing—original draft, review, and editing. Kevin J. Tracey: conceptualization, funding acquisition, project administration, supervision, and writing—review and editing. Sangeeta S. Chavan: conceptualization, formal analysis, funding acquisition, methodology, project administration, supervision, validation, visualization, and writing—original draft, review, and editing.