Neuro-immune regulation of sepsis-associated delirium via the PBN-CeA-spleen axis

Abstract

Background

Sepsis associated delirium (SAD) is the most prevalent manifestation of acute brain dysfunction in sepsis, yet its central regulatory mechanisms remain incompletely understood. As researchers have progressively explored the brain-spleen axis and neuroimmunity, this study aimed to investigate the role of the parabrachial nucleus (PBN)-central amygdala (CeA)-spleen axis in the pathogenesis of SAD.

Methods

In a mouse model of SAD induced by intraperitoneal injection of lipopolysaccharide (LPS, 10 mg/kg), we identified activation patterns in both the PBN and CeA using immunofluorescence analysis. We subsequently investigated direct anatomical connections between these regions through bidirectional neural tracing, followed by functional interrogation using fiber photometry and electrophysiological recordings. Finally, we evaluated the effect of targeted inhibition of the PBN-CeA pathway on delirium in septic mice through behavioral assays. Additionally, enzyme-linked immunosorbent assay (ELISA) and flow cytometry were employed for immunological profiling.

Results

We observed a significant increase in c-Fos expression in both the PBN and CeA, and confirmed a direct anatomical connection between these regions. Fiber photometry and electrophysiological recordings revealed that LPS stimulation activated CeA neurons and splenic sympathetic nerves. In addition, targeted inhibition of the PBN-CeA pathway mitigated CeA calcium dynamics and reduced spontaneous splanchnic nerve discharge. Behavioral assays showed that splenic denervation and inhibition of the PBN-CeA pathway attenuated delirium-like behaviors in septic mice. ELISA and flow cytometry demonstrated that these interventions reversed splenic proinflammatory cytokines (tumor necrosis factor-alpha [TNF-α], interleukin [IL]-1β, IL-6, and IL-10, interferon-gamma [IFN-γ]) and norepinephrine (NE) while restoring immune cell composition and enhancing natural killer (NK) cell function.

Conclusion

The PBN-CeA-splenic axis plays a critical neuroimmune conduit linking central neural circuits to peripheral immune modulation, offering new mechanistic insight into the neural regulation of systemic inflammation and the pathogenesis of SAD.

Introduction

Sepsis, defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection, remains a major global health challenge, with mortality rates of 30%–50% [1, 2]. Its rising incidence, driven in part by the aging population, is accompanied by increased vulnerability to multiorgan dysfunction syndrome (MODS) and enduring cognitive impairment [3, 4]. Among its complications, sepsis-associated delirium (SAD) is the most prevalent acute brain dysfunction, affecting 30%–70% of septic patients. SAD is strongly associated with increased in-hospital mortality and persistent neurocognitive deficits [4–6]. Although the pathogenesis of SAD involves multifactorial processes, such as neuroinflammatory activation of the blood–brain barrier (BBB), and immunometabolic dysregulation, the core regulatory circuitry orchestrating these events remains poorly defined [7, 8].

Emerging evidence from authoritative studies [9–11] shows that the nervous and immune systems engage in bidirectional communication across molecular, cellular, and organ levels, profoundly influencing immune homeostasis. These interactions are increasingly recognized as critical modulators of immune dysfunction in sepsis and represent potential therapeutic entry points for resolving sepsis-induced immunopathology. In particular, the brain-spleen axis has gained prominence as a critical interface in central-peripheral immune communication. As the largest secondary lymphoid organ, accounting for 25% of the total lymphoid tissue, the spleen orchestrates both cellular and humoral immune responses and plays a crucial role in shaping the host's immunological trajectory during sepsis [12]. Neuroanatomical tracing has revealed that the central amygdala (CeA) contains the densest splenic nerve projections among brain regions innervating the spleen [13]. Optogenetic stimulation of corticotropin-releasing hormone (CRH)-expressing neurons in the CeA induces rapid splenic nerve activation, subsequently enhancing B cell differentiation into plasma cells via α9 nicotinic acetylcholine receptor (α9nAChR)-mediated signaling and promoting the antigen-specific antibody production [9, 14]. Simultaneously, CeA-originating cholinergic projections to the dorsal motor nucleus of the vagus (DMV) dynamically regulate splenic lymphocyte composition [15]. γ-aminobutyric acid (GABA)-ergic CeA neurons can suppress cholinergic DMV activity, thereby reducing splenic Th2 cell populations and attenuating anti-inflammatory responses [16]. This bidirectional neuromodulation may be particularly relevant in sepsis, where the temporal shift from early hyperinflammatory states to late immunosuppression may reflect dysregulation within CeA-splenic neural circuits [17]. These insights identify the CeA as a candidate hub for neuroimmune regulation and a promising therapeutic target for reversing sepsis-induced immunosuppression and alleviating SAD [18].

The parabrachial nucleus (PBN), located in the brainstem, is a critical integrative center for processing nociceptive and visceral sensory inputs. It comprises a heterogeneous population of neurons, primarily glutamatergic, with smaller subsets of GABAergic and calcitonin gene-related peptide (CGRP)-expressing neurons. Glutamatergic neurons constitute approximately 85% of PBN neurons and are crucial for mediating arousal-related behavioral and electroencephalographic responses [19–24]. GABAergic neurons, though less abundant, exert inhibitory control over local excitatory circuits and are modulated by upstream inputs from regions such as the bed nucleus of the stria terminalis (BNST), which promote ingestion and alleviate anxiety [22, 25]. CGRP-positive PBN neurons, many of which co-express glutamate, serve as generalized alarm responders to a variety of aversive stimuli, including visceral discomfort [26–28]. The PBN projects extensively to the CeA. This PBN-CeA connectivity has emerged as a key modulator in neuroimmune signaling [26, 29–31]. In cecal ligation and puncture (CLP)-induced models of sepsis, both the PBN and the CeA exhibited increased c-Fos protein expression [32]. Under acute stress, PBN neurons can transiently enhance GABAergic transmission in the CeA [33]. Approximately 50% of CRH-expressing CeA neurons receive direct PBN input [34, 35], and CGRP released from PBN terminals may regulate splenic sympathetic output via CeA-CRH signaling pathways [33, 36]. This PBN-CeA-spleen pathway may constitute a mechanistic basis for the pathological feedback loop between peripheral immune activation and central neuroinflammation in SAD.

Given this context, we hypothesized that the PBN-CeA-splenic axis represents a central neuroimmune conduit contributing to SAD pathogenesis. To test this, we examined the anatomical and functional connectivity of the PBN-CeA projection and its modulation of splanchnic nerve activity in a LPS-induced mouse model of sepsis. Using chemogenetic inhibition of the PBN-CeA projection and surgical splenic denervation, we observed significant attenuation of delirium-like behaviors. Complementary immunological analyses using flow cytometry and enzyme-linked immunosorbent assay (ELISA) further revealed significant restoration of peripheral inflammatory cytokine levels and immune cell homeostasis. Our findings identify the PBN-CeA-splenic axis as a critical neuroimmune regulatory pathway in sepsis, providing mechanistic insights and a potential therapeutic target for mitigating SAD.

Materials and methods

Animals

Adult male C57BL/6 mice (6–8 weeks old, 18–22 g) were purchased from Hunan SJA Laboratory Animal Co., Ltd. (Changsha, China). The mice were group-housed (4–5 per cage) under controlled conditions (temperature: 25 ± 1 °C; relative humidity: 60 ± 5%; a 12-h light/dark cycle with lights on from 09:00 to 21:00) and provided with food and water ad libitum. All behavioral tests were performed during the light phase. Prior to testing, mice were allowed to acclimate for seven days and habituated to the testing conditions for at least 30 min. Littermates were randomly allocated to experimental groups before viral injections. All animal procedures were approved by the Medical Ethics Committee of Xiangya Hospital of Central South University (No. 2023020009). All biohazard materials were disposed of in compliance with the Medical Waste Management Regulations of China.

Drugs and viral constructs

Lipopolysaccharide (LPS; Escherichia coli O55:B5, Sigma-Aldrich, USA) was prepared in 0.9% saline and administered via intraperitoneal (i.p.) injection (10 mg/kg) or lateral ventricular cannulation (2 μg/2 μL) [37]. Clozapine-N-oxide (CNO; MedChemExpress, USA) was dissolved in 5% dimethyl sulfoxide (DMSO) in saline and administered i.p. at 5 mg/kg [38]. 6-Hydroxydopamine hydrobromide (6-OHDA; Sigma-Aldrich, USA; Cat# 162,957) was prepared at 120 μg/60 μL in saline and administered via intrasplenic injection [39]. Dimethyl fumarate (DMF; Sigma-Aldrich, USA; Cat# 50,744) was prepared at 5 mg/mL in 5% DMSO and saline, and was administered i.p. at 30 mg/kg [40].

To inhibit the PBN-CeA pathway, we employed CRE-dependent viral constructs for chemogenetic suppression. The following recombinant adeno-associated viruses (rAAVs) were obtained from BrainVTA Technology Co., Ltd. (Wuhan, China): rAAV2/1-hSyn-CRE, rAAV2/retro-hSyn-CRE, rAAV2/9-hSyn-DIO-mCherry, rAAV2/retro-CaMKIIα-CRE and rAAV2/9-CaMKIIα-DIO- hM4D(Gi)-mCherry, rAAV-CAG-GCaMp6s, PRV-CAG-EGFP.

Stereotaxic surgery and viral injection

Mice were anesthetized with sodium pentobarbital (60 mg/kg, i.p.) and secured in a stereotactic frame (RWD Life Sciences, Shenzhen, China). For anterograde tracing, 0.2 μL of rAAV2/1-hSyn-CRE was injected bilaterally into the PBN (anteroposterior [AP]: –5.45 mm; mediolateral [ML]: ± 1.25 mm; dorsoventral [DV]: –3.65 mm), and 0.2 μL of rAAV2/9-hSyn-DIO-mCherry into the CeA (AP − 1.25 mm; ML ± 2.85 mm; DV − 4.15 mm) via a microinjection pump (RWD Life Sciences). For retrograde tracing, 0.2 μL of rAAV2/9-hSyn-DIO-mCherry was injected into the PBN and rAAV2/retro-hSyn-CRE into the CeA. For chemogenetic inhibition, mice received bilateral injections of rAAV2/9-CaMKIIα-DIO-hM4D(Gi)-mCherry into the PBN and rAAV2/retro-CaMKIIα-CRE into the CeA. For lateral ventricular administration, a cannula was implanted at the following coordinates: AP: − 0.5 mm; ML: ± 1.0 mm; DV: − 2.25 mm.

PRV retrograde tracing from the spleen

Mice were anesthetized with sodium pentobarbital (60 mg/kg, i.p.). The spleen was accessed through a midline opening to the peritoneal cavity. One microlitre of the viral stock solution was injected into each of the upper, middle, lower tips of the spleen using a capillary glass needle. The surgical wound was closed by standard sutures and the indicated numbers of mice were killed 96 h later to retrieve the brain.

Immunofluorescence and image analysis

Mice were anesthetized with sodium pentobarbital (100 mg/kg, i.p.) and transcardially perfused with 0.9% saline followed by 4% paraformaldehyde (PFA). Brains were fixed overnight in 4% PFA at 4 °C, cryoprotected, and sectioned coronally at 7 μm thickness at the PBN and CeA levels. The tissue sections were blocked in 3% bovine serum albumin (BSA) with 0.1% Triton X-100 for 50 min and incubated overnight at 4 °C with primary antibodies: anti-c-Fos (1:1000, ab222699, Abcam), anti-mCherry (1:500, ab125096, Abcam), anti-CGRP (1:200, 14,959, Cell signaling), anti-vesicular GABA transporter (VGAT) (1:500, 131,004, SYSY), and anti-vesicular glutamate transporter 2 (VGLUT2) (1:1000, ab2251-I, Sigma). On the following day, the sections were washed three times with PBS and incubated with the corresponding secondary antibodies conjugated to Alexa Fluor 488 (1:500, ab150113, Abcam), Alexa Fluor 594 (1:500, ab150080, Abcam), and Alexa Fluor 647 (1:500, A-21450, ThermoFisher) for 1 h at room temperature. After washing three times with PBS, nuclei were counterstained with 4′,6-Diamidino-2-phenylindole (DAPI) and imaged using a confocal microscope. The number of c-Fos positive cells was quantified using ImageJ software.

Electrophysiological recording

Under anesthesia (5% induction and 1.5–2% maintenance isoflurane with 1 L/min oxygen), mice were placed on heating pads. A midline abdominal incision was made to expose and isolate the splenic artery and associated nerve. A cuff electrode (KD-cuff-2, KedouBC) was positioned around the splenic nerve, ensuring that the nerve rested securely atop the metallic surface of the electrode probe [9, 41]. The electrode was connected to a BNC 2090 A interface (National Instruments) and Model 1700 amplifier (A-M Systems). Electrophysiological signals were recorded using WaveSurfer software. Signals were bandpass filtered (10–60 Hz), and spike activity was automatically detected using MATLAB (R2021b) via the findpeaks function, with thresholds determined by median absolute deviation (MAD). Mean firing rates were calculated for 5 min before and after LPS or CNO stimulation were calculated and statistically. Data were analyzed using two-way repeated measures analysis of variance (ANOVA).

Fiber photometry

After injecting AAV-GCaMP6m into the right CeA, a 1.25-mm-diameter optical fiber (Thinkertech, Nanjing, China) was implanted. A lateral ventricular cannula was implanted on the contralateral side. Mice were allocated individually for three weeks post-surgery. Fluorescent calcium transients were captured using a tri-color multi-channel fiber photometry system (QAXK-FPS-SS-MC-LED, Thinkertech) using 405, 470, and 560 nm excitation lasers. Data were processed by custom-written MATLAB scripts to extract calcium-dependent fluorescence dynamics.

Splenic denervation

Under isoflurane anesthesia, an abdominal incision was made to expose the spleen. A 30-gauge insulin needle was used to inject 6-OHDA directly into the spleen, with the needle retained in situ for 3 min post-injection [39]. Mice recovered for seven days before proceeding to immunological or behavioral assessments.

Behavioral tests

Behavioral tests were conducted 24 h after the LPS infusion between 09:00 and 18:00. Each mouse was tested in a single experimental group. Preceding behavioral assessments, mice were acclimated for 1 h in the testing conditions. Apparatuses were cleaned with 75% medical alcohol between sessions. The Confusion Assessment Method (CAM) adaptation for mice was used to assess the delirium behaviors, which include open field test (OFT), Y-maze test, buried food-seeking test (BFST) [42, 43]. BFST and OFT were employed to evaluate alterations in natural behavior, while the Y-maze test was utilized to assess changes in learned behaviors in mice.

Open field test (OFT)

Behavioral assessment commenced 30 min post-CNO administration. Mice were placed in a 40 × 40 × 40 cm chamber and allowed to explore for 10 min. The total distance traveled, freezing duration, time and entries into the center zone (20 × 20 cm), and latency to the center zone were recorded using an automated tracking system.

Y-maze test

The rodent spatial cognition assay was conducted following an optimized Y-maze paradigm adapted from the previous studies [42, 44]. Each maze consisted of three arms arranged at 120° angles between each arm. In the training trial (10 min), conducted 30 min after CNO administration, murine subjects were allowed to explore two unobstructed arms while the third (novel) arm was blocked. Arm designations (start, familiar, novel) were randomized across animals to control for spatial bias. Subsequent to a 2-h inter-experimental interval, the retention trial (5 min) was conducted, during which mice had free access to all arms. Time spent in and the number of entries into the novel arm were recorded as indicators of spatial working memory. Behavior was tracked and analyzed using the Tracking-Master video tracking system.

Buried food-seeking test (BFST)

The experiment was administered following previous studies [42, 45]. Mice were fasted the day prior. For testing, a single feed nugget was concealed 8 cm below bedding in a clean cage. murine subjects were positioned in the cage and allowed 5 min to locate the food. Latency was recorded from cage entry until the mouse grasped the pellet. Mice failing to locate the food were assigned a latency of 300 s.

Composite Z score

To assess integrated behavioral changes, we computed a composite Z-score (C_ZX) by synthesizing data from multiple behavioral tests, adapting established methodologies [21, 22]. For each mouse, the change (ΔX) between post-injection measurements and baseline values was computed. Using the ΔX values from the control (CON) group, we calculated the mean (M_ΔXCON) and standard deviation (SD_ΔXCON). The composite Z-score for individual mice was defined as:

Where Z_X represents the Z-score for each behavioral indicator. The C_ZX metric was then used in subsequent statistical analyses.

Flow cytometry

Spleen samples were collected from each group of mice, and splenocytes were obtained by mechanical dissociation in 10 mL of RPMI 1640 medium (GIBCO, Invitrogen) supplemented with 5% FBS, which was then filtered through a 70 mm sterile mesh (352,350, Falcon). Staining of spleen cell suspensions was initiated by split-red (555,899, BD) processing. Staining reagents included CD19 (557,399 or 563,148, BD), B220/CD45R (552,771, BD), CD138 (558,626, BD), FAS (557,653, BD), GL-7 (144,614, Biolegend), CD86 (105,040, Biolegend), CD38 (102,728, Biolegend), CD45 (557,659, BD), CD3e (553,061, BD), CD4 (553,051 or 563,232, BD), CD8 (561,109 or 566,985, BD), F4/80 (565,411 BD), NK1.1 (557,391 or 562,062, BD), CD69 (553,237 or 552,879, BD), CD27 (124,212, Biolegend), and CD11b (557,397 or 550,993, BD). Cells were typically stained on ice with Staining reagents for 45 min. Experimental datasets were acquired via BD LSR Fortessa cytometer and analyzed with FlowJo. Doublets and dead cells were excluded by Fixable Viability Stain (564,406 or 564,996, BD) and FSC-A/FSC-H characteristics. Non-specific signal blockade was achieved using mouse BD Fc block (553,141, BD).

ELISA for cytokines and norepinephrine (NE)

Serum and spleen homogenates from mice. Spleen samples were minced into approximately 1 mm³ fragments and incubated in ice-cold PBS containing 1% EDTA, agitated at 100 rpm in a 35-mm dish on a shaker for 15 min. Supernatants were collected by centrifugation. Inflammatory cytokine (tumor necrosis factor-alpha [TNF-α], interleukin [IL]−1β, IL-6, IL-10, and interferon-gamma [IFN-γ]) and NE levels were then measured using TNF-α ELISA kit (EK0527, BOSTER), IL-1β ELISA kit (EK0394, BOSTER), IL-6 ELISA kit (EK0411, BOSTER), IL-10 ELISA kit (EK0417, BOSTER), IFN-γ ELISA kit (EK0375, BOSTER), and NE ELISA kit (H091-1–2, Njjcbio) according to the manufacturer’s instruction.

Statistical analyses

Data were presented as mean ± standard error of the mean and analyzed using GraphPad Prism 9.0 software (GraphPad, San Diego, CA, USA). Behavioral, c-Fos, flow cytometry, and ELISA data were analyzed using one-way ANOVA followed by Tukey–Kramer post hoc tests. Electrophysiology and Fiber photometry results were analyzed using two-way repeated measures ANOVA. Normality was assessed using a D'Agostino-Pearson test. Preset significance level (α) was established at P < 0.05 for all inferential analyses.

Results

Validation of the PBN-CeA connection in C57BL/6J mice

The anatomical pathway from the PBN to the CeA was confirmed using bidirectional trans-monosynaptic tracing techniques. For anterograde tracing, rAAV2/1-hSyn-CRE was stereotactically injected into the PBN, and rAAV2/9-hSyn-DIO-mCherry was injected into the CeA. rAAV2/1 vectors are known to support paracrine trans-synaptic transport at high titers [46]. At 21 days post-injection, robust mCherry expression was observed in neuronal somata within the CeA, indicating successful anterograde trans-synaptic labeling of PBN-originating projections (Fig. 1C). For retrograde tracing, rAAV2/retro-hSyn-CRE was administered into the CeA, while rAAV2/9-hSyn-DIO-mCherry was delivered to the PBN. Immunofluorescence revealed substantial mCherry expression in PBN neurons co-labeled with VGAT (Fig. 1A), VGLUT2 (Fig. 1B), and CGRP, consistent with neural coding of injurious stimulation in the PBN [24]. In addition, we injected the retrograde transsynaptic tracer pseudorabies virus (PRV-CAG-EGFP) expressing enhanced green fluorescent protein into the spleen to validate the neural control of the CeA over splenic function. Robust PRV fluorescence was subsequently detected in the CeA region (Fig. 1D).

Fig. 1.

Retrograde and anterograde viral tracing of the PBN-CeA-spleen neural circuit. A, B In retrograde viral tracing, representative immunofluorescence micrographs showing co-localization of mCherry with c-Fos, VGAT (A), VGLUT2 (B), and CGRP in the PBN. C In anterograde viral tracing, representative micrographs showing mCherry expression in the CeA. D Typical micrographs of neurons labelled by PRV retrograde tracing from the spleen to the CeA. PBN, parabrachial nucleus; CeA, central amygdala

Time-dependent activation of the PBN-CeA axis after LPS stimulation

To assess the temporal dynamics of neuronal activation within the PBN-CeA pathway during systemic inflammation, mice received intraperitoneal LPS (10 mg/kg), and brain tissues were collected at baseline (control), 6 h, 24 h, and 7 d post-LPS stimulation (n = 3 per time point). Coronal sections of the PBN and CeA were processed for c-Fos immunofluorescence staining. Quantitative analysis revealed a significant increase in c-Fos-positive neurons in the PBN at 6 h post-LPS stimulation (P < 0.05), with maximal activation observed at 24 h (Fig. 2A–B: P < 0.01). In the CeA, c-Fos distribution was significantly elevated at 6 h (P < 0.001), with a further increase at 24 h (Fig. 2C–D: P < 0.05). By 7 d post-injection, c-Fos levels in both regions had returned to baseline, indicating a lack of sustained activation. These findings suggest that the PBN-CeA circuit may participate in acute neuroimmune signaling.

Fig. 2.

c-Fos expression in the PBN and CeA after LPS-induced sepsis. A, C Representative images showing c-Fos immunoreactivity in the PBN (A) and CeA (C) at baseline (control) and after LPS administration (6 h, 24 h, and 7 d). B, D Quantification of c–Fos positive cells in the PBN (n = 3) and CeA(n = 3), respectively. LPS significantly increased c-Fos expression in the PBN at 6 h and 24 h and in the CeA at 24 h, with no significant change at 7 d. Data are presented as mean ± SEM, *P < 0.05, **P < 0.01, ***P < 0.001. PBN, parabrachial nucleus; CeA, central amygdala

Chemogenetic inhibition of the PBN-CeA pathway reverses LPS-induced CeA calcium hyperactivity

To functionally validate the role of the PBN-CeA circuit in sepsis-associated neural activation, we combined fiber photometry with chemogenetic modulation. CRE-dependent hM4D(Gi)-AAV was injected into the PBN, while retrograde CRE-expressing vectors were administered into the CeA. GCaMP6s was co-expressed in the CeA for real-time calcium imaging (Fig. 3A). We confirmed the regulatory effect of chemogenetic viruses on the PBN-CeA pathway through immunofluorescence validation. The results indicated that chemogenetic inhibition of the PBN-CeA pathway significantly reduced c-Fos expression in CeA neurons induced by sepsis (Fig. 4B, E; P < 0.05). After three weeks of viral expression and lateral ventricular cannula implantation, mice were randomized into two experimental groups: 1) the LPS group receiving intraperitoneal injections of LPS (10 mg/kg) followed by CNO or saline and 2) the Sham group receiving saline and subsequently challenged with LPS or saline via ventricular administration. Baseline calcium signals in the CeA were recorded in three consecutive phases (Fig. 3B): a 10-min baseline (0–10 min), a 5-min administration window (10–15 min), and a 30-min post-intervention period (15–45 min). In the Sham group (Fig. 3C), significant increases in CeA calcium transients were observed post-LPS compared to baseline and saline-treated controls (n = 4; P < 0.05 vs. 20 min; P < 0.01 vs. 30 min), indicating LPS-induced hyperactivation of CeA neurons. In the LPS cohort (Fig. 3D), CNO administration (5 mg/kg i.p.) led to a significant reduction in CeA calcium activity relative to both baseline and saline-injected controls (n = 4; P < 0.0001 vs. 20 min; P < 0.001 vs. 30 min). These findings confirm the functional engagement of the PBN-CeA pathway while establishing a mechanistic foundation for developing targeted therapeutic strategies against sepsis-associated neurological dysfunction.

Fig. 3.

Suppressed neuronal activity in the CeA and splenic sympathetic nerve after PBN-CeA pathway inhibition in septic mice. A Schematic diagram illustrating fiber photometry targeting the CeA. B GCaMP6s calcium signal traces in response to intracerebroventricular injection of saline and LPS (1.0 mg/mL) in sham-operated mice, and intraperitoneal injection of CNO (5 mg/kg) or saline in LPS-treated mice. C Statistical analysis results of calcium transients in Sham + Saline group and Sham + LPS group (n = 4). D Statistical analysis results of calcium transients in LPS + Saline group and LPS + CNO group (n = 4). E Workflow of in vivo electrophysiological recordings. F Representative traces of splenic nerve firing at baseline (Initial), after intracerebroventricular injection of LPS, and after intraperitoneal injection of saline or CNO in LPS-treated mice. Spikes are marked in red. G Quantification of spike frequency in Saline + Saline group, LPS + Saline group, LPS + CNO group (n = 4). Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Fig. 4.

Experimental design of behavioral tests and validation of Splenic denervation or chemogenetic interventions. A Schematic timeline of experimental design grouping and behavioral testing, including BFST, OFT, and Y-maze. B Typical micrographs of each group demonstrating c-Fos immunoreactivity in the CeA (C) Typical micrographs of each group demonstrating tyrosine hydroxylase (TH) distribution within splenic sections. D Quantification of mean TH fluorescence intensity in the spleen of sham-operated or splenic denervated mice. E Quantification of c–Fos positive cells in the CeA(n = 3), respectively. LPS significantly increased c-Fos expression in the CeA at 24 h, with significant decreased after chemogenetic inhibition of PBN-CeA pathway. Data are presented as mean ± SEM, *P < 0.5, ****P < 0.0001. PBN, parabrachial nucleus; CeA, central amygdala; BFST, buried food-seeking test; OFT, open field test

PBN-CeA pathway modulates LPS-induced elevation in splenic sympathetic nerve activity

To evaluate the PBN-CeA pathway’s influence on peripheral autonomic output, we conducted in vivo electrophysiological recordings of splenic nerve activity(Fig. 3E). Mice received stereotaxic injections of retrograde CRE vectors into the CeA and CRE-dependent hM4D(Gi)-AAV into the PBN, followed by lateral ventricular cannulation. After three weeks of recovery, baseline recordings of spontaneous splenic nerve discharge were obtained using implanted cuff electrodes under isoflurane anesthesia. LPS (2 μL; 1.0 mg/mL) was microinjected into the lateral ventricle [47], and splenic nerve activity was recorded at defined intervals. Baseline signals were collected prior to LPS exposure, followed by recordings taken 2 h after LPS stimulation and again 30 min after intraperitoneal injection of CNO or saline. After two hours post-LPS-injection, a significant increase in splenic nerve firing rate was observed compared to both baseline and saline-injected controls (Fig. 3F-G: P < 0.001), consistent with known activation of sympathetic output in response to systemic inflammation [48]. Chemogenetic inhibition of the PBN-CeA pathway via CNO administration leads to a robust suppression of splenic nerve activity (P < 0.001), reversing LPS-induced sympathetic excitation.

Behavioral deficits in sepsis-associated delirium are ameliorated through chemogenetic inhibition of the PBN-CeA pathway or splenic denervation

To assess whether modulation of the PBN-CeA-spleen axis could alleviate behavioral abnormalities associated with SAD, splenic sympathetic innervation fibers were ablated using intrasplenic injections of 6-OHDA. A week post-injection, immunofluorescence staining confirmed effective denervation, evidenced by significantly reduced tyrosine hydroxylase (TH) expression in splenic sections compared to untreated controls (Fig. 4C-D: P < 0.0001). Concomitantly, a marked decline in norepinephrine levels within the splenic parenchyma was observed (Fig. 6L, P < 0.001), indicating attenuated sympathetic innervation.

Fig. 6.

Reduced inflammatory cytokines in serum and spleen of septic mice following splenic denervation or chemogenetic inhibition of PBN-CeA pathway. A–L ELISA-based quantification of TNF-α (A, B), IL-6 (C, D), IL-1β (E, F), IL-10 (G, H), IFN-γ (I, J), and NE (K, L) levels in serum and splenic tissue at 24 h post-LPS in serum and spleen. (M-P) ELISA-based quantification of serum pro-inflammatory cytokines in the Sham, LPS and DMF + LPS groups. Data are presented as mean ± SEM, n = 5 per group, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. TNF-α, tumor necrosis factor-alpha; IL-1β, interleukin-1β; IL-6, interleukin-6; IL-10, interleukin-10; IFN-γ, interferon-gamma; NE, norepinephrine

Subsequently, five experimental groups were established (Fig. 4A): control (DIO-mCherry) + sham surgery + saline (CSS), control + sham surgery + LPS (CSL), control + splenic denervation + saline (CDS), control + splenic denervation + LPS (CDL), and chemogenetic inhibition of the PBN–CeA pathway (DIO- hM4D[Gi]-mCherry) + sham surgery + LPS (GiSL). After 21 days of recovery from viral injections and 7 days post-denervation, SAD was induced with intraperitoneal LPS (10 mg/kg). Delirium-associated behavioral tests, assessed using the CAM in mice, were conducted 24 h later to evaluate spontaneous motivation (BFST), emotional regulation (OFT), and cognitive performance (Y maze) [42, 44].

In the buried food-seeking test (Fig. 5J), splenic denervation in the CDS group significantly shortened latency relative to the CSS group (P < 0.0001). Mice in the CSL group exhibited a profound delay in locating the buried food pellet, with an average latency exceeding 300 s relative to CSS controls (P < 0.01). In contrast, septic mice with splenic denervation (CDL) did not significantly differ from CSL. However, chemogenetic inhibition of the PBN-CeA pathway (GiSL) significantly improved food retrieval latency compared to the CSL group (P < 0.01).

Fig. 5.

Attenuation of sepsis-induced delirium-like behaviors by splenic denervation, chemogenetic inhibition of PBN-CeA pathway, or anti-inflammatory treatment. A Representative movement traces in the OFT. B-F Quantitative analysis of total distance traveled, center entries, center dwell time, latency to center, and immobility duration in the OFT. G Representative exploration paths in the Y-maze test. H, I Quantification of novel arm entries and dwell time, showing partial rescue of spatial memory deficits by PBN-CeA inhibition and splenic denervation (n = 6). (J) BFST latency measurements showing improved spontaneous behavior following PBN-CeA inhibition. K Calculation of the composite Z-score for various behavioral indicators assessed the level of delirium. L-S Quantitative analysis of changes in various behavioral indicators(OFT/Y maze/BFST) of delirium in sepsis mice treated with dimethyl fumarate to inhibit inflammatory factors. T Composite Z-score of the LPS and DMF + LPS treatment groups. All data are presented as mean ± SEM, n = 6 per group, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

In the Y-maze (Fig. 5G-I), LPS-exposed mice (CSL) demonstrated significant reductions in novel arm entries (P < 0.0001 vs. CSS) and dwell time (P < 0.001). Both GiSL and CDL interventions partially reversed these impairments. Specifically, the GiSL group exhibited significantly greater novel arm entries than the CDL group (P < 0.01) and CSL mice (P < 0.05), while novel arm dwell time was also improved (P < 0.05 vs. CDL).

In the OFT (Fig. 5A-F), septic mice (CSL) exhibited significantly lower movement distance (P < 0.0001 vs. CSS), fewer center entries (P < 0.0001), and reduced center dwell time (P < 0.0001), along with significantly increased latency to enter the center (P < 0.05) and prolonged freezing time (P < 0.0001). Splenic denervation in septic mice (CDL) improved locomotor distance (P < 0.01 vs. CSL), number of entries into the center (P < 0.01), center dwell time (P < 0.05), and freezing time (P < 0.001), though latency to enter the center remained unchanged. In contrast, chemogenetic inhibition of the PBN-CeA pathway (GiSL) significantly restored all behavioral parameters to near-control levels (P < 0.05). Composite Z-score analysis revealed that splenic denervation and chemogenetic PBN-CeA pathway inhibition significantly improved delirium-like behaviors in septic mice (Fig. 5K).

To investigate whether amelioration of delirium is associated with a reduction in inflammatory cytokines, we pre-treated mice with intraperitoneal dimethyl fumarate (DMF, 30mg/kg) 1 h prior to LPS-induced sepsis modeling to attenuate pro-inflammatory cytokine production [49]. Behavioral tests assessing delirium-like symptoms were performed 24 h post-sepsis induction. Septic mice treated with DMF demonstrated significantly improved delirium-related behaviors compared to untreated septic controls (Fig. 5L-T). ELISA analysis of serum samples confirmed significant reductions in levels of key pro-inflammatory cytokines (TNF-α, IL-1β, IL-6, and IL-10) in the DMF-treated septic mice compared to their untreated counterparts (Fig. 6M-P).

Chemogenetic inhibition of PBN-CeA pathway or splenic denervation could influence the levels of inflammatory cytokines in septic mice

To investigate the role of the PBN-CeA-spleen axis in modulating peripheral inflammation, cytokine concentrations in serum and spleen were quantified using ELISA at 24 h post-sepsis induction. Mice from all experimental groups were assessed to characterize both systemic and splenic immune responses, focusing on five key cytokines: TNF-α, IL-1β, IL-6, IL-10, and IFN-γ.

In saline-treated mice, splenic denervation (CDS) significantly reduced circulating levels of IL-6 (Fig. 6C, P < 0.05) and IL-1β (Fig. 6E, P < 0.001) compared to the CSS group. No significant differences were observed in TNF-α (Fig. 6A), IL-10(Fig. 6G) and IFN-γ (Fig. 6I) levels between these groups. LPS administration (CSL) triggered significant elevations in all five cytokines. Both splenic denervation (CDL) and chemogenetic inhibition of the PBN-CeA pathway (GiSL) significantly decreased serum TNF-α(P < 0.01) and IL-1β (P < 0.01) levels relative to the CSL group. Additionally, PBN-CeA pathway inhibition also reduced serum IL-6 (P < 0.05) and IL-10(P < 0.05) levels.

In the splenic microenvironment, basal cytokine levels were also modulated by denervation. Compared to CSS controls, splenic denervation (CDS) exhibited significant reductions in TNF-α (Fig. 6B: P < 0.05), IL-1β (Fig. 6F: P < 0.05), and IL-10 (Fig. 6H: P < 0.01) levels. Septic mice (CSL) showed significant elevations in splenic TNF-α (P < 0.01 vs CSS), IL-6 (Fig. 6D: P < 0.01), and IL-1β (P < 0.0001), accompanied by decreased IL-10 (P < 0.05) and IFN-γ (Fig. 6J: P < 0.01) levels. Both splenic denervation (CDL) and chemogenetic inhibition of the PBN-CeA pathway (GiSL) effectively attenuated LPS-induced upregulation of TNF-α (P < 0.0001 vs. CDL; P < 0.01 vs. GiSL), IL-6 (P < 0.0001), and IL-1β (P < 0.001 vs. CDL; P < 0.05 vs. GiSL), while restoring IFN-γ (P < 0.05) expression in the spleen. Interestingly, both interventions further reduced IL-10 (P < 0.05) levels in the splenic compartment. These findings collectively suggest that support a model in which the PBN-CeA-spleen axis regulates peripheral cytokine production and splenic sympathetic tone during sepsis.

To further elucidate the functional link between neural activity and immune modulation, NE levels were measured in serum and spleen. In serum, LPS stimulation significantly elevated NE levels compared to CSS controls (Fig. 6K: P < 0.05 vs. CSS). However, neither splenic denervation nor PBN-CeA pathway inhibition significantly altered circulating NE levels. In contrast, splenic NE levels were significantly reduced in denervated mice (Fig. 6L: P < 0.001 vs. CDS) and septic mice (P < 0.0001 vs. CSL). Both splenic denervation (CDL) and chemogenetic inhibition (GiSL) reversed this reduction (P < 0.05), resulting in significantly higher splenic NE levels compared to controls.

PBN-CeA pathway inhibition and splenic denervation restore immune cell composition and enhance NK cell maturation in septic mice

To determine whether modulation of the PBN-CeA-spleen axis influences immune cell profiles in sepsis, splenic immune cell subpopulations were analyzed by flow cytometry at 24 h after LPS administration. The results showed that macrophages were not significantly different among lymphocyte populations in all groups of mice (Fig. 7A-B). Sepsis resulted in marked alterations in splenic lymphocyte composition. Compared to controls (CSS), septic mice (CSL) exhibited significant reductions in splenic T cells (Fig. 7C, E: P < 0.0001), NKT cells (Fig. 8F, G: P < 0.0001), and NK cells (Fig. 8F, H: P < 0.001), accompanied by a paradoxical increase in B cells (Fig. 7C, D: P < 0.0001). Both interventions (CDL and GiSL) significantly restored splenic NK cell (P < 0.01 vs. CDL; P < 0.001 vs. GiSL) and T cell proportions (P < 0.01) and normalized B cell frequencies (P < 0.0001 vs. CDL; P < 0.05 vs. GiSL) relative to septic controls. However, only the PBN-CeA pathway inhibition group (GiSL) significantly elevated NKT cell levels (P < 0.01).

Fig. 7.

Modulation of splenic lymphocyte subsets by splenic denervation or PBN-CeA pathway inhibition in sepsis. A Flow cytometry gating strategy for identifying macrophages. B Quantification of macrophage frequencies across groups. C Representative bivariate plots showing gating of CD19⁺CD3e⁻ B cells and CD19⁻CD3e⁺ T cells within CD45⁺ splenocytes. D–G Statistical comparison of B cells (D), total T cells (E), CD8⁺CD4⁻ T cells (F), and CD8⁻CD4⁺ T cells (G). Data are presented as mean ± SEM, n = 5 per group. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Fig. 8.

Enhanced maturation of splenic NK cells after PBN-CeA pathway inhibition or splenic denervation in septic mice. A Flow cytometry gating strategy for identifying germinal center B cells (GL-7⁺FAS⁺), plasmablasts (CD19⁻CD138⁺), and memory B cells (CD38⁺CD86⁺). B–E Quantification of respective B cell subtypes. F Gating strategy for NK1.1⁺CD3e⁻ NK cells, NK1.1⁺CD3e⁺ NKT cells, and mature (CD27⁺CD11b⁺ and CD27⁻CD11b⁺) NK cells, G–K Quantitative analysis of NKT cells (G), NK cells (H), CD27⁺CD11b⁺ NK cells (I), CD27⁻CD11b⁺ NK cells (J) CD69⁺ NK cells (K). Data are presented as mean ± SEM, n = 5 per group. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Subsequent subset analysis of B cells, T cells, and NK cells revealed differential effects between CDL and GiSL groups. B cells were subdivided into germinal center, plasma, and memory populations based on their developmental stages and functions, with CD86-high germinal centers suggesting robust activation [50]. Splenic denervation (CDL) significantly reduced germinal center numbers in septic mice (Fig. 8D: P < 0.01 vs. CSL), while the chemogenetic inhibition of the PBN-CeA pathway (GiSL) further increased memory B cell levels (Fig. 8C: P < 0.05 vs. CSL) in sepsis. Additionally, GiSL treatment declined CD4⁺ T cell populations (Fig. 7G: P < 0.0001 vs. CSL) and enhanced CD8⁺ T cell populations (Fig. 7F: P < 0.05 vs. CSL).

To assess the maturation of NK cells, we used CD11b and CD27 markers to define four developmental stages: CD11b⁻CD27⁻ → CD11b⁻CD27⁺ → CD11b⁺CD27⁺ → CD11b⁺CD27⁻ [51]. The CD27⁺CD11b⁺ represents a transitional stage with high proliferative and cytokine-secreting potential, whereas terminally differentiated CD11b⁺CD27⁻ subset is cytotoxic and marked by expression of inhibitory receptors such as KLRG1 [52, 53]. In the CSL group, transitional CD27⁺CD11b⁺ NK cells were significantly increased (Fig. 8F, I: P < 0.01 vs. CSS), while terminally differentiated CD11b⁺CD27⁻ NK cells were significantly decreased (Fig. 8F, J: P < 0.0001 vs. CSS), suggesting that sympathetic modulation influences NK cell terminal differentiation. Both interventions (CDL and GiSL) corrected this imbalance by enhancing the CD27⁺CD11b⁺ subset and restoring the terminally differentiated CD27⁻CD11b⁺ subset. Meanwhile, early NK cell activation, assessed by CD69 expression was selectively reduced in the GiSL group (Fig. 8K: P < 0.01 vs. CSL) but not in CDL mice.

Discussion

Neuroimmune imbalance is increasingly recognized as a critical mechanism underlying immune dysregulation in sepsis [8]. As a clinical manifestation of this imbalance, SAD is characterized by disrupted bidirectional neuroimmune communication between the central nervous system (CNS) and peripheral immune responses [54]. In this study, we employed three complementary methodologies—neuroanatomical tracing, chemogenetic intervention, and multimodal immune profiling, to identify the PBN-CeA-Spleen axis as a novel neural pathway implicated in SAD pathogenesis. These findings provide experimental validation for the theory of bidirectional neuroimmune regulation [9, 10, 16].

As the largest peripheral immune organ, the brain-spleen axis-mediated regulation of splenic immune function plays a pivotal role in immune dysregulation during sepsis [55–57]. Our experimental data demonstrated that LPS exposure potentiated CeA calcium signaling and significantly increased splenic nerve discharge. Chemogenetic inhibition of the PBN-CeA glutamatergic pathway reversed both effects, confirming its upstream regulatory role. Furthermore, LPS-induced NE depletion in the spleen—potentially reflecting neurotransmitter exhaustion from sympathetic overactivation—was rescued by both splenic denervation and inhibition of the PBN-CeA pathway. Collectively, these findings validate the role of the PBN-CeA-spleen axis in coordinating splenic sympathetic output and maintaining NE homeostasis, a critical modulator of neuroimmune tone.

This study elucidates the previously unrecognized role of the PBN-CeA-splenic axis in SAD pathogenesis. In behavioral paradigms, both PBN-CeA inhibition and splenic denervation ameliorated sepsis-induced deficits in spatial memory (Y-maze) and exploratory behaviors (OFT). Notably, only PBN-CeA inhibition improved spontaneous motivational behavior (BFST) in septic mice. These results align with previous studies identifying that the CeA serves as a critical hub for integrating sensory-emotional signals and coordinating behavioral responses [58] while also functioning as a key node in brain-splenic axis connectivity [59]. The CeA pathway modulation influences both motor (OFT) and cognitive (Y-maze) functional domains [60, 61]. Mechanistically, splenic denervation may attenuate neuroinflammation by suppressing proinflammatory cytokine production, thereby improving spatial memory consolidation [9].

Our cytokine profiling reveals distinct regulatory patterns induced by PBN-CeA inhibition versus splenic denervation. Both interventions suppressed key proinflammatory cytokines (TNF-α and IL-1β) in serum and spleen. However, PBN-CeA inhibition exerted broader effects, significantly reducing IL-6 and IL-10 in both compartments. This differential cytokine regulation suggests that central modulation of this pathway may coordinate a more global rebalancing of systemic immune tone. These findings complement recent discoveries of neural-immunological crosstalk mechanisms in the brain-spleen axis [9, 16, 59]. Mechanistically, PBN-CeA loop activation drives dual pathological effects: splanchnic nerve-mediated proinflammatory factor release (e.g., TNF-α, IL-1β) and IL-10-dependent anti-inflammatory pathway suppression. Targeted pathway inhibition may thus disrupt this"immune vicious cycle"in sepsis. Notably, the contrasting IL-10 expression patterns between serum and spleen likely reflect spatiotemporal competition between cholinergic anti-inflammatory pathways [16, 61] and sympathetic activation [9]. Sepsis is characterized by the simultaneous interplay of pro- and anti-inflammatory mechanisms. At different stages of sepsis, the levels of cytokines undergo dynamic changes. IL-10, as an anti-inflammatory cytokine, increases its release with the production of pro-inflammatory cytokines. Studies have shown that the levels of IL-10 in the serum of sepsis patients are elevated [62], and are associated with the severity of inflammatory injury and the concentration of pro-inflammatory cytokines. Compared with surviving patients, the serum IL-10 levels in deceased patients are significantly higher [63]. In contrast, the levels of IL-10 in the spleen may reduce due to massive apoptosis and functional exhaustion of local immune cells. Declines in CD4 T cells and elevations in lymphocyte apoptosis was found in the spleens of cecum ligation and puncture (CLP)-induced septic rats [64]. LPS stimulation reduces splenic NE concentration [65], which may be attributed to NE depletion and sustained activation of α₂-adrenergic receptors (α₂-AR) under prolonged stress. α₂-AR activation triggers negative feedback that inhibits further NE release [66]. Following splenectomy and chemogenetic inhibition of the PBN-CeA pathway, splenic NE levels significantly increased, likely due to the blockade of α₂-AR-mediated suppression. Consistent with this mechanism, the α₂-AR antagonist yohimbine significantly enhances NE signaling in LPS-stimulated splenic tissue while reducing TNF-α, IL-6, and IL-10 levels [67]. This further explains the additional decline in IL-10 observed after splenectomy and PBN-CeA pathway inhibition. This restoration of NE homeostasis confirms the PBN-CeA-spleen axis's pivotal role in coordinating sympathetic-immune interactions during sepsis.

Flow cytometric analysis demonstrated that sepsis induced a profound shift in splenic lymphocyte composition, marked by reductions in NK and T cells and an increase in B cells. Both interventions (CDL and GiSL) restored T and NK cell proportions and normalized B cell frequency. Subtype analysis focusing on the most altered cell populations revealed that splenic denervation significantly reduced germinal center numbers in septic mice, whereas PBN-CeA inhibition specifically improved memory B cell levels. Additionally, sepsis impaired NK cell maturation in the spleen, evidenced by an increased proportion of transitional CD27⁺CD11b⁺ NK cells and a reduction in the proportion of terminally differentiated CD27⁻CD11b⁺ NK cells. This skewed maturation pattern suggests an arrest in the transition to the cytotoxic effector phenotype. Such disruptions are consistent with previous reports that NE dose-dependently reduces NK cell cytotoxicity, perforin and granzyme B expressions, and IFN-γ production—effects that are reversible with β₂-adrenergic antagonists [68]. Both CDL and GiSL interventions could further increase the proportion of transitional CD27⁺CD11b⁺ NK cells and restore the proportion of terminally differentiated CD27⁻CD11b⁺ NK cells by locally attenuating the splenic sympathetic signaling, which led to elevated levels of IFN-γ in the spleen of mice. Previous studies demonstrate that NK cell dysfunction during sepsis progression correlates with increased secondary infection risk and mortality [69–71], modulating this pathway may provide a novel therapeutic target for sepsis-induced immune dysregulation.

All experiments used adult male C57BL/6 mice (8–10 weeks). We employed intraperitoneal LPS administration (10 mg/kg) rather than CLP to reduce the impact of CLP surgery on behavior tests. LPS administration is also commonly used in other similar studies [72, 73]. The dual immunomodulatory role of splenic nerves in sepsis pathogenesis remains mechanistically elusive and paradoxically unresolved despite its critical therapeutic relevance. This investigation systematically characterizes the PBN-CeA-spleen axis's neural-immune network in sepsis, but several limitations should be noted. First, the exclusive focus on early-phase sepsis limits our understanding of long-term neuroimmune adaptations. Second, the molecular underpinnings of cytokine and immune cell modulation downstream of this circuit were not elucidated. Future investigations should integrate single-cell transcriptomic profiling to decode splenic immune cell heterogeneity under neural modulation. Integration of optogenetic and pharmacogenetic techniques will further enable spatiotemporally dissection of the PBN-CeA circuit.

Conclusions

This study identifies the PBN-CeA-spleen axis as a neuroimmune circuit driving the pathophysiology of sepsis-associated delirium. By orchestrating both splanchnic sympathetic output and peripheral immune responses, this pathway mediates the dynamic interplay between central neuroinflammation and systemic immunosuppression. Chemogenetic and surgical disruption of the axis attenuates maladaptive cytokine release, restores immune cell homeostasis, and improves sepsis-induced behavioral deficits. These findings delineate a novel mechanistic framework for developing precision neuromodulation strategies against sepsis-induced immune paralysis and associated neurological complications.

Abbreviations

SAD

Sepsis-associated delirium

PBN

Parabrachial nucleus

CeA

Central amygdala

LPS

Lipopolysaccharide

ELISA

Enzyme-linked immunosorbent assay

TNF-α

Tumor necrosis factor-alpha

IFN-γ

Interferon-gamma

NE

Norepinephrine

NK

Natural killer

IL

Interleukin

MODS

Multiorgan dysfunction syndrome

BBB

Blood–brain barrier

CRH

Corticotropin-releasing hormone

α9nAChR

α9 Nicotinic acetylcholine receptor

DMV

Dorsal motor nucleus of the vagus

GABA

γ-Aminobutyric acid

CGRP

Calcitonin gene-related peptide

BNST

Bed nucleus of the stria terminalis

CLP

Cecal ligation and puncture

CNO

Clozapine-N-oxide

DMSO

Dimethyl sulfoxide

6-OHDA

6-Hydroxydopamine hydrobromide

rAAVs

Adeno-associated viruses

AP

Anteroposterior

ML

Mediolateral

DV

Dorsoventral

PFA

Paraformaldehyde

BSA

Bovine serum albumin

VGAT

Anti-vesicular GABA transporter

VGLUT2

Anti-vesicular glutamate transporter 2

DAPI

4′,6-Diamidino-2-phenylindole

MAD

Median absolute deviation

ANOVA

Two-way repeated measures analysis of variance

OFT

Open field test

BFST

Buried food-seeking test

CNS

Central nervous system

DMF

Dimethyl fumarate

α₂-AR

α₂-Adrenergic receptors

Authors’ contributions

MLD performed the experiments and wrote the manuscript. QYP designed and supervised the research. YZ and ZWY collected the data. WYZ contributed to the revision of the manuscript. YHA and LNZ reviewed the manuscript. All the authors reviewed and approved the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82372199), the Natural Science Foundation of Hunan Province (2025JJ40070), the Natural Science Foundation of Hunan Province (2024JJ5559), China.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All animal procedures were approved by the Medical Ethics Committee of Xiangya Hospital of Central South University (No. 2023020009).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.