Stress-induced vagal activity influences anxiety-relevant prefrontal and amygdala neuronal oscillations in male mice

The vagus nerve (VN) conveys interoceptive signals from peripheral organs to the brain and is strongly implicated in emotional states and psychiatric disorders. Altered vagal signaling (e.g., via gut microbiota changes or vagotomy) increases anxiety- and depression-like behaviors, whereas vagus nerve stimulation (VNS) ameliorates treatment-resistant depression in humans and produces anxiolytic and antidepressant effects in rodents. Despite this, how VN activity dynamically relates to moment-to-moment anxiety states and how this relationship is pathologically altered in stress-related mental disorders remain unclear. Anxiety-related processing in the prefrontal cortex (PFC) and amygdala (AMY), including interregional oscillatory coordination, suggests that VN activity could support these brain activity patterns. This study simultaneously recorded VN activity and local field potentials (LFPs) from PFC and AMY in naïve, stress-resilient, stress-susceptible, and vagotomized male mice during rest and elevated plus maze (EPM) behavior, and further tested whether chronic VNS can restore anxiety-related behavior and oscillations in stress-susceptible mice.

Prior work has established that afferent VN fibers (the majority of cervical VN) transmit interoceptive information to the CNS and that manipulating vagal signaling influences emotion. Clinical VNS improves treatment-resistant depression, and rodent studies report anxiolytic and antidepressant effects of VNS. Noninvasive transcutaneous auricular VNS has also been shown to modulate human physiology and brain rhythms. Anxiety involves PFC and AMY circuits, with oscillatory coordination (notably in theta and beta bands) modulating anxiogenic behavior. However, the specific dynamic coupling between VN activity and PFC-AMY oscillations during anxiety behavior, and its disruption under stress susceptibility, had not been delineated. The literature also notes variability in EPM open-arm time after social defeat (SD) stress, suggesting that finer-grained behavioral states (e.g., move vs stop within closed arms) may better capture anxiety-related physiology.

Approvals: All procedures were approved by animal ethics committees at the University of Tokyo (P29-14) and Tohoku University (2022 PhA-004) and followed NIH guidelines. Subjects: Male C57BL/6J mice (8–10 weeks, 29–35 g) underwent social defeat (SD) stress, behavioral testing, and electrophysiology. Male CD-1 mice (>13 weeks, 40–50 g) served as aggressors. Social defeat stress: CD-1 residents were screened over three days for aggressiveness (attack frequency/latency criteria). For 10 consecutive days, an intruder C57BL/6J mouse was placed with a screened aggressor for 5–10 min (terminated upon wounding), then housed across a perforated partition for 24 h with sensory (no physical) contact. A different aggressor was used each day. Phenotyping: Social interaction (SI) test comprised two 150-s sessions (no-target and target CD-1 in a wire-mesh cage), separated by 30 s. The SI ratio was time in the interaction zone (14.5 × 24 cm around cage) during target divided by no-target. Mice with SI<1 were classified as stress-susceptible; SI>1 as stress-resilient. Stability of phenotype was verified by re-testing ~3 weeks later. Behavior: Elevated plus maze (EPM) consisted of a central square (7.6 × 7.6 cm) with two open and two closed arms (28 × 7.6 cm; 15 cm walls on closed), elevated 30 cm. Mice explored for 10–20 min; the first 10 min were analyzed. Behavioral states were segmented as open vs closed arms and, within closed arms, move (>1 cm/s) vs stop (<1 cm/s). Position tracked at 15 Hz (downsampled to 3 Hz). Surgery and electrodes: Under isoflurane anesthesia, a custom cuff electrode (inner diameter 0.3 mm; electrode area 0.15 mm²; 2.0 mm cathode–anode spacing; 4.0 mm length) was implanted on the left cervical VN with a reference electrode on the adjacent salivary gland. For vagotomy experiments, the left cervical VN was transected. An electrode assembly with up to seven tetrodes targeted left PFC (AP +2.0 mm, ML 0.0 mm, DV −1.8 mm) and left AMY (AP −1.5 mm, ML +3.5 mm, DV −4.5 mm). Tetrodes were 17 μm Pt-Ir, tips plated to 200–250 kΩ. An EMG electrode was sutured to dorsal neck muscle; ground on cerebellum. Electrophysiology: Signals were acquired via Blackrock Cereplex systems. LFP/EMG sampled at 2 kHz (low-pass 500 Hz). Unit activity: 750 Hz–6 kHz bandpass, threshold at 50 μV, 30 kHz sampling. VN signals sampled at 30 kHz via cuff and reference. On recording days: pre-rest (≥30 min), EPM (10–20 min), post-rest (≥30 min). VN signal processing: To enhance SNR, VN electrode signal was subtracted from salivary gland reference. During movement/EPM, compound VN activity precluded unit isolation; signals were band-pass filtered at 300–1000 Hz and RMS computed in 1/3 s bins as VN spike power. During quiescent rest (≥5 min segments defined by EMG), spike-like events were detected at ~5 SD below mean of filtered VN baseline; bursts within 5 ms counted as single spikes. Heartbeat-locked artifacts (110–130 ms intervals) were excluded. EMG processing: EMG band-pass 20–200 Hz; RMS in 1-s bins. Movement thresholds were set per mouse (~8 SD above baseline) by visual inspection. LFP analysis: Wavelet-based power spectra (1–100 Hz). Instantaneous LFP power was z-scored per frequency band per mouse. Correlations between instantaneous VN spike power and PFC/AMY LFP power were computed across frequencies. Spike–LFP phase locking of PFC single units quantified by mean vector length (MVL); significance determined by shuffling (P<0.001). Vagus nerve stimulation (VNS): In separate cohorts, mice received electrical VNS through the cuff: 0.8 mA, 0.1 ms pulse width, 20 Hz, 600 pulses (30 s) every 1 min for 3 h. For c-Fos mapping, VNS was delivered every minute for 3 h; brains collected 1.5 h later for immunostaining (c-Fos) in NTS, PFC, and AMY. For therapeutic experiments, stress-susceptible mice received daily VNS for 1–2 weeks, then underwent SI, EPM, and electrophysiology. Histology and immunostaining: Electrode placements confirmed by cresyl violet. c-Fos immunostaining used anti-c-Fos primary and Alexa 488 secondary; imaging with fluorescence microscopy. Statistical analyses included paired/unpaired t-tests with Bonferroni correction, one-way ANOVA with Tukey’s post hoc tests, and Mann–Whitney U tests; data reported with n, t/F/Z, and P values.

- SD stress phenotyping: Among 36 SD-exposed mice, 20 were stress-susceptible (SI<1) and 16 stress-resilient (SI>1). SI ratios were stable over ~3 weeks (R=0.66, P=0.011). - Reduced basal VN activity in susceptibility: During quiescent rest, VN spike power and spike rates were lower in susceptible vs naïve/resilient mice (n=9 naïve, 12 resilient, 12 susceptible). Spike power: ANOVA F2,32=7.40, P=0.0024; naïve vs susceptible P=0.012; resilient vs susceptible P=0.0044; naïve vs resilient P=0.99. Spike rate: ANOVA F2,32=11.01, P=3.0×10−4; naïve vs susceptible P=3.9×10−11; resilient vs susceptible P=0.0040; naïve vs resilient P=0.50. - EPM behavior: Open-arm time did not differ across groups (F2,53=0.60, P=0.55). Susceptible mice showed increased proportion of move states within closed arms (F2,53=9.07, P=0.00040; naïve vs susceptible P=0.0012; resilient vs susceptible P=0.0029). - Behavior-dependent VN power: In naïve and resilient mice, VN spike power was higher in open vs closed arms and in closed move vs stop (naïve n=11: open vs closed move t10=3.10, P=0.034; open vs closed stop t10=4.16, P=0.0058; closed move vs closed stop t10=2.98, P=0.041. Resilient n=12: t11=3.85, P=0.0081; t11=5.74, P=3.9×10−4; t11=3.87, P=0.0078). Susceptible mice showed no significant VN power differences across states (n=14; all P>0.16). Across groups, open-arm VN power: ANOVA F2,36=3.95, P=0.029; naïve>susceptible (P=0.035). - VN–LFP coupling in naïve mice: VN spike power correlated negatively with 2–4 Hz and positively with 20–30 Hz LFP power in PFC and AMY. PFC: 2–4 Hz negative in 6/9 mice; 20–30 Hz positive in 7/9. AMY: 2–4 Hz negative in 4/9; 20–30 Hz positive in 9/9. - PFC single-unit entrainment: Of 47 PFC neurons (6 mice), 6 (12.8%) showed significant phase-locking to 2–4 Hz and 2 (4.3%) to 20–30 Hz oscillations (MVL, P<0.001 by shuffling). - Behavior-linked LFP dynamics (naïve): PFC 2–4 Hz power was lower and 20–30 Hz higher in open vs closed arms (n=16; 2–4 Hz: t15=2.87, P=0.035; t15=3.56, P=0.0086. 20–30 Hz: t15=3.89, P=0.0044; t15=6.24, P=4.7×10−5). In closed arms, PFC 20–30 Hz was higher in move vs stop (t15=6.07, P=6.4×10−5) but 2–4 Hz was not (t15=1.36, P=0.58). AMY showed similar 20–30 Hz differences (n=9: open vs closed move t8=3.73, P=0.017; open vs closed stop t8=6.32, P=6.8×10−5; closed move vs stop t8=4.13, P=0.0099) but not 2–4 Hz. - Resilient vs susceptible LFP patterns: Resilient mice retained significant 20–30 Hz PFC/AMY differences across behavioral states; 2–4 Hz differences were not significant. Susceptible mice lacked significant differences for both bands in PFC and AMY (P>0.09). Across-group PFC 20–30 Hz: open F2,40=5.40, P=0.0086 (naïve vs susceptible P=0.048; resilient vs susceptible P=0.011); closed stop F2,40=5.17, P=0.010 (naïve vs susceptible P=0.016; resilient vs susceptible P=0.042). - Vagotomy phenocopies susceptibility: Vagotomized mice had increased closed-arm move states vs naïve (n=18 naïve, 11 vagotomy; t27=3.20, P=0.0035) with no open-arm time difference (t27=1.68, P=0.10). They showed no significant PFC/AMY 2–4 or 20–30 Hz power differences across behavioral states (all P≥0.061), indicating VN activity is necessary for dynamic PFC-AMY oscillatory modulation. - VNS restores behavior and oscillations in susceptible mice: Daily VNS (0.8 mA, 0.1 ms, 20 Hz, 30 s every 1 min for 3 h, 1–2 weeks) increased SI ratios (Z=4.18, P=3.0×10−3), reduced closed-arm move proportion vs susceptible controls (n=20 vs 18; t37=3.04, P=0.0044), and restored VN spike power differences (n=7: open vs closed stop t6=5.88, P=0.0032; closed move vs stop t6=7.72, P=7.4×10−4; open-arm VN power higher than susceptible t19=4.19, P=5.0×10−4). PFC and AMY 20–30 Hz power differences across behavioral states were restored (PFC n=11: open vs closed move t10=5.12, P=0.0013; open vs closed stop t10=6.09, P=3.5×10−5; closed move vs stop t10=4.76, P=0.0023. AMY n=10: open vs closed stop t9=7.87, P=7.6×10−5; closed move vs stop t9=5.26, P=0.0016). Across-group, PFC 20–30 Hz open t24=3.47, P=0.020; closed stop t24=2.18, P=0.039; AMY open t18=3.08, P=0.0065.

The study shows that physiological VN activity dynamically tracks anxiety-relevant behavioral states and is tightly coupled to specific PFC-AMY oscillations. In healthy and stress-resilient mice, VN spike power increases in more anxiogenic contexts (open arms; movement in closed arms) and correlates with decreased 2–4 Hz and increased 20–30 Hz power in PFC and AMY. A subset of PFC neurons phase-lock to these rhythms, indicating local circuit entrainment. Stress-susceptible mice exhibit reduced basal VN activity and a breakdown of behavior-dependent modulation of both VN signals and PFC-AMY oscillations. Vagotomy mimics these behavioral and electrophysiological deficits, supporting a causal role for vagal inputs in shaping anxiety-related cortical-limbic dynamics. Conversely, chronic VNS restores VN excitability, rescues 20–30 Hz oscillatory modulation in PFC/AMY, and normalizes anxiety-relevant behavior in susceptible mice. These findings indicate that vagal-brain communication is a key substrate underlying anxiety and may be a mechanistic target for neuromodulatory therapies.

This work establishes that stress-induced vulnerability is associated with attenuated vagal signaling and disrupted anxiety-relevant oscillations in PFC-AMY circuits. In naïve/resilient mice, VN activity correlates negatively with 2–4 Hz and positively with 20–30 Hz PFC/AMY power, and these rhythms track behavioral states in the EPM. Stress-susceptible and vagotomized mice lack these dynamic patterns, whereas chronic VNS restores VN activity, 20–30 Hz cortical-limbic oscillations, and anxiety-related behavior. The study highlights vagal-brain communication as a physiological foundation for anxiety and mood disorders and supports VNS-based interventions. Future research should dissect organ-specific vagal pathways and neuronal subtypes, and optimize stimulation parameters to refine mechanistic understanding and therapeutic efficacy.

- VN recordings during movement yielded compound activity, preventing isolation of single-fiber units; analyses relied on RMS power in these epochs. - The study did not identify the specific peripheral organ sources or VN branch subtypes driving the observed effects; more refined recordings/manipulations of VN branches and genetic identification of VN subtypes are needed. - PFC single-unit analyses did not determine neuronal cell types; phase-locked units were inferred but not classified. - VNS parameters (duration, interval) were effective but not optimized; further work is needed to define optimal stimulation protocols. - Experiments were conducted in male mice; sex differences were not addressed.