Auricular vagus nerve stimulation drives analgesia via an auricle–brain axis in a mouse model of neuropathic pain

Abstract

Despite its long-recognized effects in relieving pain, the neural substrates of auricular stimulation remain elusive. Here, we show that trans-auricular vagus nerve stimulation (taVNS), i.e., electrical stimulation of the auricular concha, effectively induces analgesia in a mouse model of neuropathic pain. Viral tracing, microendoscopic calcium imaging, and multi-electrode recordings reveal that auricular vagal signals travel to the jugular-nodose ganglia (JNG), which in turn connect to pro-opiomelanocortinergic neurons in nucleus tractus solitarius (NTS), subsequently activating glutamatergic neurons in ventrolateral periaqueductal gray (vlPAG). Optogenetic stimulation of central vagus terminals, JNG-derived auricular peripheral fibers, or vlPAG-projecting NTS neurons mimics taVNS-induced analgesia, whereas chemogenetic silencing of central vagus terminals or NTS neurons abolishes this effect. This study identifies an auricle–to–brain circuit underlying taVNS-driven analgesia in mice, with potential for facilitating taVNS optimization for pain management and other neurological conditions.

The neuroanatomical basis of auricular vagus nerve stimulation-induced analgesia remains unclear. Here, the authors identify an auricle-to-brain axis from the auricular concha to the ventrolateral periaqueductal gray, which engages the descending antinociceptive pathway.

Introduction

The Food and Drug Administration (FDA) has approved surgical vagus nerve stimulation (VNS) for treatment of drug-resistant epilepsy and depression^1–4, although trans-auricular VNS (taVNS) is increasingly used as a less invasive option to replace surgical VNS^5–7. Furthermore, taVNS has shown promise in alleviating a variety of neurological diseases, such as major depressive disorders⁸. Despite its long history and promising results in clinic, the mechanism(s) through which taVNS induces analgesia remain ambiguous, suggesting that unraveling the neuroanatomy and physiology underlying the peripheral and brain responses to this treatment could advance its application, potentially benefitting treatment selection for a great number of patients.

Functional magnetic resonance imaging (fMRI) studies in human subjects have shown that stimulation of the auricular vagus nerve by taVNS can activate multiple brain regions^5,9. These regions innervated by the auricular vagus nerve include nucleus tractus solitarius (NTS), the locus coeruleus (LC), and the ventrolateral periaqueductal gray (vlPAG)^10,11, all of which are known to play critical roles in regulating pain perception^12–14. Although the vagus nerve does not equally innervate all regions of the auricle, and taVNS efficacy can depend on stimulation of the concha rather than the margin, insufficient understanding of the anatomical and functional connections between the auricle and different brain regions involved in pain processing has hindered improvement of the taVNS responder rate¹. Previous studies have shown that VNS can reverse auditory or motor cortical plasticity to alleviate tinnitus-like symptoms¹⁵, enhance perceptual learning¹⁶, or promote rehabilitation of limb movement¹⁷, which has been proposed to occur through changes in plasticity of pain-associated brain areas. However, precise cell-type-specific mapping of this neural circuitry is necessary to investigate this possible mechanism¹⁸.

In this study, we investigated the functional organization of an auricle–jugular/nodose ganglion (JNG)–NTS–vlPAG circuit and examined changes in its activity following taVNS in spared nerve injury (SNI) model of neuropathic pain in mice through viral tracing, microendoscopic calcium imaging, and multi-electrode recordings. Optogenetic activation or chemogenetic inhibition manipulations were used to determine whether each nucleus or brain region was indeed necessary for the analgesic effects of taVNS. Our findings revealed that taVNS-induced analgesia relies on an auricle-to-brain axis via the vlPAG-dependent descending pain inhibitory system.

Results

taVNS induces analgesia in mice with neuropathic pain

To assess the impact of taVNS on pain perception, we established neuropathic pain models by applying SNI, a clinically relevant paradigm for studying chronic neuropathic pain, in wild-type C57 mice (Fig. 1)¹⁹. Following taVNS administration in an anesthetized state, after regaining consciousness, the awake mice exhibited significantly increased paw withdrawal thresholds in von Frey tests. Consistent with previously documented antinociceptive effects^20–22, electrical stimulation at the left concha (taVNS, 0.2 ms pulse width, alternating frequencies between 3 and 15 Hz, 30 min) (Fig. 1a) led to an obvious elevation in mechanical pain threshold, with the maximum effect observed at 0.5 h post-taVNS. This acute antinociception dramatically diminished by 1.5 h post-taVNS. Considerable chronic analgesia was achieved after 9 consecutive days of taVNS treatment, and mice exhibited significantly elevated nociceptive thresholds that persisted for 4 days before sensitization recurred (Fig. 1b, c), suggesting substantial analgesia of the persistent pain. Additionally, taVNS could alleviate cold allodynia in 5 °C cold plate tests (Fig. 1d, e), and induced preference for the taVNS-paired chamber in conditional place preference (CPP) tests in SNI mice (Fig. 1f–h). taVNS-treated SNI mice also displayed significantly shorter duration of hindpaw lifting in 3 °C cold plate tests (Fig. 1i, j, and Supplementary Movie 1), further supporting that taVNS could reduce coping behaviors associated with cold allodynia. Similar acute and chronic antinociceptive effects were observed in mice treated with complete Freund’s adjuvant (CFA), thus demonstrating taVNS efficacy across different pain etiologies (Supplementary Fig. 1a, b). In contrast, control mice treated with electrical stimulation at the left auricle margin (i.e., trans-auricular non-vagus nerve stimulation; tnVNS) showed no such effects on sensory or emotional components of pain throughout the experimental observation period (Fig. 1c, e, g, h, j). Notably, taVNS also resulted in acute analgesia in awake, head-fixed SNI mice, indicating that anesthesia was dispensable for the effects of taVNS (Supplementary Fig. 1c–e).

Fig. 1. Analgesic effects of taVNS in SNI mice.

a Schematic for taVNS in isoflurane-anesthetized mice. b Timeline for inducing neuropathic pain by SNI, taVNS treatment, and nociceptive threshold measurements. c Mechanical nociceptive threshold in taVNS- or tnVNS-treated SNI mice and sham controls (n = 10 mice/group; F_{(96, 1152)} = 32.34, P < 0.0001). d Timeline for inducing neuropathic pain by SNI, taVNS treatment, and cold allodyina measurements. e Withdrawal latency to cold stimulation in taVNS- or tnVNS-treated SNI mice and sham controls (n = 10 mice/group; F_{(105, 560)} = 18.83, P < 0.0001). f Schematic for the CPP tests paradigm. Heatmaps (g) of time spent in the stimulation chambers during CPP test, and quantification of the preference ratio (h, n = 9 mice/group; F_{(3, 18.83)} = 12.64, P =  0.014). Schematic (i) and quantification (j, n = 6 mice/group; F_{(18, 120)} = 20.30, P < 0.0001) of coping behaviors in taVNS- or tnVNS-treated SNI mice and sham controls. Two-way ANOVA with Bonferroni post hoc analysis for c, e, j. Brown–Forsythe ANOVA with two-sided Dunnett’s T3 post hoc analysis for h. All data are reported as mean ± SEM. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001. See also Supplementary Table 1. Source data are provided as a Source Data file. CPP conditioned place preference, SNI spared nerve injury, taVNS trans-auricular vagus nerve stimulation, tnVNS trans-auricular non-vagus nerve stimulation.

Given that the taVNS is believed to act similarly to direct cervical VNS²³, we investigated changes in the vagus nerve activity during the treatment process. Electrophysiological recordings in the nerve trunk revealed that spontaneous firing activity significantly increased in vagal nerve fibers during and after taVNS (Supplementary Fig. 2a–e). Furthermore, we observed altering stimulation parameters concomitantly affected the strength of response, with our taVNS protocol resulting in effective activation of vagus nerve (Supplementary Fig. 2f–h). In Fos^TRAP2::Ai9 mice, we observed that taVNS induced significant c-Fos expression in the vagal sensory ganglion (i.e., JNG), while no such c-Fos activation was detected in untreated control mice (Supplementary Fig. 2i–k); immunofluorescence staining showed that these c-Fos⁺ cells were glutamatergic (Supplementary Fig. 2l, m). Collectively, these results suggested that taVNS could elicit analgesia in mice with neuropathic pain, which we hypothesized might be due to activation of auricular vagal afferent fibers and auricle-innervating subsets of vagal sensory neurons.

Anatomical characterization of the auricle-to-brain circuit

As the analgesic effects of taVNS might involve the central projection of the auricular vagus nerve²³, we therefore investigated whether this brain circuit was indeed involved in taVNS-mediated antinociception. We directly injected an AAV2/9-CAG-EGFP virus into the JNG (Fig. 2a), and 3 weeks later observed profuse vagus fibers at the concha, but relatively sparse fibers near the auricle margin (Fig. 2b, c, and Supplementary Movie 2). These findings aligned well with human studies that reported a high density of vagus nerves at the concha²⁴. This tracing strategy also enabled visualization of brain-wide regions innervated by the vagus nerve in which we subsequently observed prominent EGFP⁺ vagal central fibers at the NTS and, to a lesser extent, at the spinal trigeminal nucleus (SP5; Fig. 2d). To trace the central fibers and cell bodies of the JNG neurons involved in the auricular branch of the vagus nerve, we introduced a retrograde scAAV-Retro-EGFP virus at the auricle to label auricular JNG neurons (aJNG) (Fig. 2e). After 3 weeks of virus expression in mice given injections at the concha, we observed abundant EGFP⁺ cells at the JNG and EGFP⁺ fibers at the NTS (Fig. 2f, g), whereas injection at the auricular margin resulted in no detectable labeling (Supplementary Fig. 3a–d). In situ hybridization for Prdm12 or Phox2b revealed that auricle-specific neurons within the jugular ganglia predominantly expressed Prdm12, while those in the nodose ganglia exhibited strong Phox2b expression (Fig. 2h–j). Despite both populations sharing a glutamatergic phenotype (Supplementary Fig. 3e–g), this molecular distinction suggested that auricular vagal afferents comprise a molecularly heterogeneous sensory ensemble derived from both the neural crest and placodal origins^25,26.

Fig. 2. Anatomical characterization of the auricle-to-brain circuit.

a Schematic for anterograde tracing by virus injection into JNG in naïve C57 mice. Typical images of cell bodies (b), peripheral afferent fibers at the auricle (c), and central projection fibers (d) of virus-infected JNG neurons from 6 replicates. e Schematic for retrograde tracing by viral injection into the auricular concha. Typical immunofluorescence images of labeled JNG cell bodies (f) and axon terminals (g) in the NTS from 8 replicates. Typical fluorescence in situ hybridization (FISH) images of Prdm12 (h) or Phox2b (i) co-localization with retrogradely traced JNG neurons from the auricular concha, with quantitative analysis (j, n = 6 ganglia slices from 3 mice). k Schematic for double retrograde tracing by simultaneous virus injection at the concha and NTS. l Typical images of JNG neurons retrogradely labeled from the auricle (EGFP, green), or NTS (mCherry, red), and overlapping populations (yellow) innervating both regions from 5 replicates. All data are reported as mean ± SEM. AP area postrema, DMV dorsal motor nucleus of vagus, L-JNG left jugular-nodose ganglion, NTS nucleus tractus solitarius, SP5 spinal trigeminal tract. Source data are provided as a Source Data file.

Furthermore, delivery of retrograde scAAV-Retro-EGFP and scAAV-Retro-mCherry viruses at the concha and NTS, respectively, resulted in extensive co-localization of EGFP-expressing auricle-projecting JNG neurons with those mCherry-expressing NTS-projecting JNG neurons (Fig. 2 k, l). This circuit was also confirmed using conventional retrograde tracers, with Fluorogold (FG) applied to the auricle and wheat germ agglutinin (WGA) injected into the NTS (Supplementary Fig. 3h, i). The number of FG-labeled auricle-projecting nodose neurons was lower than that of WGA-labeled neurons, likely due to differences in uptake efficiency between these two tracers in different tissues. Given that the auriculotemporal nerve constitutes the auricular branch of the trigeminal nerve²⁴, we injected the same anterograde viral reporter into the trigeminal ganglion (TG). This labeling revealed EGFP⁺ fibers localized to the tragus, spatially distinct from auricular vagus nerve projections (Supplementary Fig. 3j–m). These findings confirm that taVNS specifically activates the NTS via JNG. Together, these results strongly supported the presence of an Auricle →JNG^Glu → NTS circuit.

An Auricle → JNG → NTS circuit mediates taVNS-induced analgesia

In order to monitor changes in the activity of NTS neurons, we then conducted in vivo multi-electrode recordings in freely moving mice. We found that spontaneous neuronal firings significantly increased in the NTS upon taVNS, but not tnVNS, treatment in awake mice (Supplementary Fig. 4a–c). To determine whether auricular vagus afferents synapsing on the NTS were required to induce analgesia, we injected an scAAV-Retro-Cre virus at the concha and a Cre-dependent AAV2/9-DIO-ChR2-mCherry virus at the JNG to express excitatory light-sensitive channelrhodopsin-2 (ChR2) in aJNG neurons (Fig. 3a, b), which enabled optogenetic manipulation of auricular vagal afferents. Consistent with the above auricular retrograde tracing with scAAV-Retro-EGFP, anatomical assessment confirmed that clusters of ChR2-mCherry⁺ vagal fibers were centrally present in the NTS and peripherally within the concha (Fig. 3c, d). Moreover, we implanted an optrode comprising eight tetrodes and an optical fiber into the NTS and observed that spontaneous NTS neuronal firings were significantly enhanced in freely-moving mice upon optical activation of auricular vagal central terminals, which was absent in mice injected with mCherry control virus (Fig. 3e–g). Mechanical threshold tests indicated that optical excitation of the central aJNG fiber at the NTS (30 min, frequencies alternating between 3 and 15 Hz) could mimic both acute and chronic antinociceptive effects of taVNS in SNI mice (Fig. 3h). Optogenetic activation of aJNG→NTS circuit in SNI mice induced significant CPP (Fig. 3i, j) and attenuated cold allodynia-related coping behavior (Fig. 3k). These effects were strictly dependent on ChR2-mediated activation, as no behavioral changes were observed in control groups, including SNI mice expressing the mCherry-only virus control, and sham-operated mice expressing either ChR2 or mCherry.

Fig. 3. Auricle → JNG → NTS circuit in taVNS-induced analgesia.

a Schematic of viral strategy for optogenetic labeling. Typical images of cell bodies (b), central projection (c), and peripheral fibers (d) of virus-labelled auricular JNG (aJNG) neurons from 4 replicates. e Schematic for virus injection, optogenetics, and optrode recording. Raster plot and voltage traces of spontaneous firings recorded in the NTS before and during light stimulation (f), with quantification of firing rates (g, mCherry: W = −127, P = 0.4164, n = 41 units from 7 mice; ChR2: W = −985, P < 0.0001, n = 45 units from 7 mice). h Mechanical nociceptive thresholds following optical excitation of aJNG central terminals (n = 5 mice/group; F_{(90, 480)} = 14.97, P < 0.0001). Heatmaps (i) of CPP, and quantification of the preference ratio (j, n = 7 mice/group; F_{(1, 24)} = 50.10, P < 0.0001). k Quantification of cold allodynia-related coping behaviors (n = 6 mice/group; F_{(15, 100)} = 10.69, P < 0.0001). l Schematic for virus injection, auricular illumination in mice under anesthesia, and in vivo recordings performed post-recovery. Raster plots and voltage traces of spontaneous firings recorded in the NTS before and following auricular illumination (m), with quantification of firing rates (n, mCherry: W = −14, P = 0.8797, n = 28 units from five mice; ChR2: W = −323, P < 0.0001, n = 26 units from five mice). o Mechanical nociceptive thresholds following auricular illumination (n = 5 mice/group; F_{(90, 480)} = 11.77, P < 0.0001). Heatmaps of optogenetically induced CPP (p) and quantification of preference ratio (q; sham, n = 7; SNI mCherry, n = 7; SNI ChR2, n = 8; F_{(1, 25)} = 14.97, P < 0.0001). r Quantification of cold allodynia-related behaviors (n = 6 mice/group; F_{(15, 100)} = 8.245, P < 0.0001). Two-sided Wilcoxon matched-paired signed rank test for g and n. Two-way ANOVA with Bonferroni post hoc analysis for h, j, k, o, q and r. All data are reported as mean ± SEM. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001; n.s., not significant. See also Supplementary Table 1. Source data are provided as a Source Data file.

Applying the same viral strategy to optically stimulate ChR2-expressing auricular vagal peripheral fibers showed that blue light delivery at the concha resulted in enhanced spontaneous firing of the vagus nerve and NTS in ChR2-expressing naïve mice, but not in mCherry-expressing control animals, compared to firing rates before stimulation (Fig. 3l–n). We also noted that optical stimulation at the concha, which harbored ChR2-expressing vagus fibers, but not at the auricular margin, elicited an obvious increase in pupil size of head-fixed mice (Supplementary Fig. 4d–g, and Supplementary Movie 3), thereby confirming that the auricular vagus nerve was successfully activated. In line with taVNS stimulation of the auricular vagus nerve, auricular optogenetic manipulations resulted in elevated mechanical thresholds (Fig. 3o), and induced CPP (Fig. 3p, q) while reducing pain-related coping behaviors (Fig. 3r). These effects were exclusively observed in ChR2-expressing SNI mice, whereas no such behavioral changes were recorded in the SNI-mCherry, Sham-ChR2, or Sham-mCherry control groups (Fig. 3o–r). Collectively, these data demonstrated that optogenetic activation of auricular vagal afferents could faithfully recapitulate the pain-relieving effects of taVNS.

Finally, to suppress central projections of aJNG neurons, we delivered an scAAV-Retro-Cre virus into the concha and a chemogenetic inhibitory AAV2/9-DIO-hM4Di-mCherry virus into the JNG. After 3 weeks, local injection of clozapine N-oxide (CNO, 1 mg/kg) into the NTS abolished the analgesic effects of taVNS in hM4Di-expressing SNI mice (Supplementary Fig. 4h–r). In contrast, no behavioral changes occurred in the SNI-mCherry, Sham-hM4Di, or Sham-mCherry control mice, thus confirming that the auricular vagal afferent circuit was necessary for taVNS-induced analgesia. Together, these results demonstrated that this specific vagal pathway was sufficient and necessary for the observed analgesia.

Auricle → JNG → NTS^POMC circuit activation in taVNS

We then investigated the specific neuronal subpopulation in the NTS that receives inputs from the auricular vagus nerve by administering an anterograde AAV2/1-hSyn-FLP virus into the concha, where it is taken up by peripheral fibers of sensory ganglion²⁷. At the same time, we delivered an AAV2/9-fDIO-mCherry virus into the NTS together with AAV2/9-DIO-EGFP targeting Glu, POMC, or GABA neurons in respective Cre-transgenic mice or cell-type-specific promoter-driven EGFP targeting tyrosine hydroxylase (TH⁺), cholecystokinin (CCK⁺), or neuropeptide Y(NPY⁺) cells in wild-type mice to selectively label these respective neuronal subtypes in NTS. Subsequent co-localization analysis revealed that aNTS neurons were primarily proopiomelanocortinergic (60.9%, NTS^POMC) and glutamatergic (78.5%, NTS^Glu) (Fig. 4a–c, and Supplementary Fig. 5a–e).

Fig. 4. Auricle → JNG → NTS^POMC circuit activation in taVNS.

a Schematic for virus injection to selectively label NTS neurons receiving inputs from auricular JNG neurons. b Typical images of mCherry⁺ cells receiving inputs from aJNG neurons and EGFP-expressing glutamatergic, POMCergic, or GABAergic neurons in the NTS of Vglut2-Cre, POMC-Cre or Vgat-Cre mice, respectively, from 6 replicates. c Percentage of cells traced from auricle co-localizing with virally labeled NTS cells expressing the indicated neuronal cell type markers (n = 6 slices from 3 mice/group). d Schematic of the viral strategy for microendoscopic calcium imaging in NTS of POMC-Cre mice. e Typical GCaMP6m fluorescence image of NTS^POMC neurons from 4 replicates. Example calcium transient traces in NTS^POMC neurons before and after taVNS or tnVNS (f), with quantitative analysis (g; tnVNS, n = 20 neurons from 4 mice, t₍₁₉₎ = 1.274, P  = 0.2180; taVNS, n = 20 neurons from 4 mice, t₍₁₉₎ = 6.99, P < 0.0001). h Schematic of the viral strategy for optogenetic activation combined with calcium imaging in NTS^POMC neurons. Example calcium transient traces in NTS^POMC neurons before and after optogenetic activation of auricular vagal fibers (i), with quantitative analysis (j; mCherry: n = 20 neurons from 5 mice, t₍₁₉₎ = 0.304, P = 0.7642; ChR2: n = 20 neurons from 5 mice, t₍₁₉₎ = 6.272, P  <  0.0001). k Schematic for viral expression of ChR2 in auricle-specific NTS (aNTS) neurons. l Mechanical nociceptive thresholds after optogenetic activation of aNTS neurons (n = 5 mice/group; F_{(90, 480)} = 28.6, P < 0.0001). Heatmaps (m) of optogenetically-induced CPP and quantification of preference ratio (n, sham, n = 7; SNI mCherry, n = 7; SNI ChR2, n = 8; F_{(1, 25)} = 11.74, P = 0.0002). o Quantification for cold allodynia-related coping behaviors (n = 6 mice/group; F_{(15, 100)} = 11.67, P < 0.0001). Two-tailed paired Student’s t-test for g and j. Two-way ANOVA with Bonferroni post hoc analysis for l, n and o. All data are reported as mean ± SEM. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001; n.s., not significant. CPP conditioned place preference, L-JNG left jugular-nodose ganglion, NTS nucleus tractus solitarius. See also Supplementary Table 1. Source data are provided as a Source Data file.

To directly visualize the calcium activity in various NTS neuronal subsets, we injected an AAV2/9-DIO-GCaMP6m virus into the NTS of POMC-Cre, Vglut2-Cre, and Vgat-Cre mice, followed by implantation of a gradient-index (GRIN) lens. Microendoscopic calcium imaging revealed that taVNS, but not tnVNS, led to a robust increase in calcium transients in NTS^POMC and NTS^Glu neurons, while NTS^GABA neurons exhibited a smaller, but statistically significant, response (Fig. 4d–g, Supplementary Fig. 5f–l, and Supplementary Movies 4 and 5). To selectively activate peripheral auricular vagal fibers, we induced ChR2 expression in JNG neurons via AAV2/9-CAG-ChR2-mCherry injection. Optical stimulation of the auricular concha replicated the subtype-specific calcium dynamics observed during taVNS (Fig. 4h–j, Supplementary Fig. 5m–o, and Supplementary Movie 6). Behaviorally, optogenetic activation of ChR2-expressing aNTS neurons ameliorated mechanical hypersensitivity, increased CPP, and attenuated coping behaviors in ChR2-expressing SNI mice, but not in sham or EGFP-only control groups (Fig. 4k–o). These cumulative results indicated that taVNS likely mediated its analgesic effects through selective activation of an NTS subpopulation.

The NTS^POMC→vlPAG circuit is required for taVNS-induced analgesia

To visualize the projection pattern of NTS neurons receiving auricular vagal afferent inputs (aNTS), we employed the combinational viral strategy described above in which AAV2/1-hSyn-Cre was injected in the concha and AAV2/9-DIO-EGFP was injected into the NTS (Fig. 5a). Three weeks later, EGFP⁺ cells were observed in the NTS (Fig. 5b). Closer interrogation of EGFP⁺ fibers revealed that those EGFP⁺ aNTS projections predominantly extended to the vlPAG, as well as to the LC, dorsal motor nucleus of the vagus nerve (DMV) and parabrachial nucleus (PBN) (Fig. 5c, and Supplementary Fig. 6a, b). Further whole-brain c-Fos mapping indicated that multiple brain areas were activated in response to taVNS, including the NTS, vlPAG, and LC (Supplementary Fig. 6c–e), regions known to play a crucial role in pain modulation, and consistent with recent descriptions of genetic labeling and projection fiber tracing of taVNS-responsive NTS neurons¹¹. Given that the most abundant fibers and c-Fos⁺ cells were located in the vlPAG, we further mapped the entire circuit from the auricle to vlPAG in wild-type mice by injecting a trans-synaptic anterograde AAV2/1-hSyn-FLP virus into the concha, a retrograde scAAV-Retro-EGFP virus into the vlPAG, and an AAV2/9-fDIO-mCherry virus into the NTS (Fig. 5d). Fluorescence microscopy showed that the NTS was indeed innervated by the auricular vagus nerve and projected to the vlPAG (Fig. 5e).

Fig. 5. NTS^POMC→vlPAG^Glu circuit in taVNS-induced analgesia.

a Schematic for anterograde viral tracing strategy to identify downstream brain areas of NTS receiving aJNG inputs. Typical image of NTS cell bodies receiving auricular vagal afferents (b) and their projection fibers at the vlPAG (c) from 3 replicates. d Schematic for viral strategy to selectively label NTS neurons receiving inputs from aJNG and projecting to vlPAG. e Typical images of mCherry-labeled NTS cells receiving auricular vagal inputs and EGFP-labeled NTS cells projecting to vlPAG, and enlarged view showing co-localization from 3 replicates. f Schematic for anterograde virus tracing of vlPAG neurons innervated by NTS^POMC neurons. Typical images of the vlPAG co-localizing with immunofluorescence signals for glutamate (Glu, g), tyrosine hydroxylase (TH, h), GABA (i), and PLCβ4 (j). k Quantitative analysis of co-localization (n = 5 slices from 5 mice). Typical traces of light-evoked postsynaptic currents (EPSCs) in vlPAG^Glu in response to optogenetic activation of NTS terminals (l), with EPSC amplitude quantification (m, n = 8 cells from 6 mice; t₍₇₎ = 13.77, P < 0.0001). n Schematic of the viral strategy for selective chemogenetic labeling of NTS^POMC neurons projecting to vlPAG. Mechanical nociceptive thresholds (o, n = 5 mice/group; F_{(15, 80)} = 61.36, P < 0.0001) and cold allodynia-related coping behaviors (p, n = 6 mice/group; F_{(12, 80)} = 4.920, P < 0.0001) in SNI mice following chemogenetic inactivation of vlPAG-projecting NTS^POMC neurons. Two-tailed paired Student’s t-test for m. Two-way ANOVA with Bonferroni post hoc analysis for o and p. All data are reported as mean ± SEM. ^**P < 0.01; ^***P < 0.001. dlPAG dorsolateral periaqueductal gray, DMV dorsal motor nucleus of vagus, lPAG lateral periaqueductal gray, L-JNG left jugular-nodose ganglion, NTS nucleus tractus solitarius, PAG periaqueductal gray, vlPAG ventrolateral periaqueductal gray. See also Supplementary Table 1. Source data are provided as a Source Data file.

Given previous reports of analgesic effects induced by stimulating the NTS^28,29 or vlPAG³⁰, together with our above observation that NTS^POMC neurons served as the primary target of auricular vagal inputs, we next explored the anatomical and physiological aspects of the NTS^POMC→vlPAG pathway in the context of taVNS-induced analgesia. To this end, we employed a series of trans-synaptic tracing techniques in which naïve wild-type mice were first injected with an AAV2/1-POMC-Cre virus in the NTS and an AAV2/9-DIO-EGFP virus in the vlPAG. This tracing strategy revealed direct connectivity between these regions (Fig. 5f). Subsequent immunofluorescence microscopy confirmed that the majority of vlPAG neurons receiving inputs from NTS^POMC were glutamatergic and PLCβ4-negative (Fig. 5g–k). Furthermore, injection of the NTS with an AAV2/9-DIO-mCherry virus in POMC-Cre mice revealed widespread projections, including to the vlPAG (Supplementary Fig. 6f, g).

To further verify these innervations, we employed retrograde tracing with a scAAV-Retro-mCherry at the vlPAG, together with immunofluorescence staining for glutamate, POMC, GABA, or TH. The results confirm that vlPAG-projecting NTS neurons were predominantly positive for glutamate and POMC (Supplementary Fig. 7a–c). Moreover, we observed that neurons projecting from NTS to vlPAG further extended to the rostral ventromedial medulla (RVM) (Supplementary Fig. 7d–f), supporting an NTS→vlPAG→RVM circuit. Subsequent functional investigations through whole-cell electrophysiological recordings in brain slices from Vglut2-Cre mice given vlPAG injection with AAV2/9-DIO-mCherry and AAV2/9-hSyn-ChR2-EYFP in the NTS revealed DNQX-sensitive EPSCs in mCherry⁺ vlPAG^Glu neurons upon blue light stimulation (Fig. 5l, m). Taken together, these results consistently demonstrated that NTS projections activate vlPAG^Glu neurons.

Next, we sought to determine the role of NTS^POMC→vlPAG circuit in mediating the analgesic effects of taVNS. Specifically, we employed a trans-synaptic retrograde strategy in POMC-Cre mice by administering an AAV-Retro-DIO-FLP virus to the vlPAG and an AAV2/9-fDIO-hM4Di-mCherry virus to the NTS in order to specifically inhibit vlPAG-projecting NTS^POMC neurons. After 3 weeks of virus expression, CNO administration abolished both taVNS-induced mechanical antinociception and improvement to cold-coping behaviors (Fig. 5n–p). A similar outcome was observed when the same viral strategy was applied in Vglut2-Cre mice (Supplementary Fig. 7g–i). Next, comparable modulation was observed upon chemogenetic manipulation of taVNS-activated vlPAG neurons in Fos^TRAP2::Ai9 mice (Supplementary Fig. 7j–n). Furthermore, chemogenetic inactivation or optogenetic activation of vlPAG neurons receiving inputs from NTS^POMC in C57 SNI mice given NTS injection with AAV2/1-POMC-Cre and vlPAG injection with either AAV2/9-DIO-hM4Di-EGFP or AAV2/9-DIO-ChR2-EGFP abolished or recapitulated the antinociceptive effects of taVNS, respectively (Supplementary Fig. 7o–u). Finally, chronic multi-electrode recordings revealed temporally coupled neuronal activity of NTS and vlPAG neurons that paralleled changes in mechanical pain thresholds (Supplementary Fig. 8a–d). These results collectively established that the NTS→vlPAG pathway was required for taVNS-induced analgesia.

Given the projection patterns of aNTS neurons and taVNS-induced c-Fos expression in the DMV, we tested whether anti-inflammatory pathways mediated by the DMV (the motor nucleus of vagal efferents) contribute to taVNS-induced analgesia. taVNS remained effective even following subdiaphragmatic vagotomy or chemogenetic inactivation of taVNS-responsive vagal motor neurons in SNI mice (Supplementary Fig. 8e, j, k). Additionally, the temporal dissociation between analgesia and reductions in serum TNF-α in SNI mice (Supplementary Fig. 8e, f) indicated that anti-inflammation³¹ is not likely a primary driver for taVNS-induced analgesia. While subdiaphragmatic vagotomy abolished the taVNS-induced rise in serum catecholamines, it did not diminish the analgesic effect (Supplementary Fig. 8e–i), further indicating that this peripheral pathway was not essential for antinociception under our experimental paradigm.

The Auricle → JNG → NTS→vlPAG circuit mediates taVNS-induced analgesia

To better understand the role of the auricle to vlPAG neural circuit in taVNS-induced analgesia, we aimed to visualize this pathway by administering a retrograde scAAV-Retro-Cre virus into the vlPAG to label neurons projecting from the NTS, along with sequential injection of Cre-dependent AAV helper viruses followed by NTS injection with a retrograde RV-EnVA-EGFP virus. We observed mCherry⁺EGFP⁺ starter cells in the NTS, and EGFP⁺ neurons in JNG, while also finding EGFP⁺ nerve fibers in the auricular concha (Fig. 6a–d). This simultaneous observation of retrogradely labeled neurons thus indicated the presence of anatomical connections comprising an Auricle→JNG → NTS→vlPAG circuit that forms a descending antinociceptive pathway targeting spinal sensory neurons³². The connectivity between auricular and spinal circuits was further validated using multi-synaptic anterograde herpes simplex virus (HSV)-based viral tracing (Supplementary Fig. 9a, b), thus supporting the existence of polysynaptic pathways mediating taVNS-induced analgesia.

Fig. 6. Auricle → JNG → NTS→vlPAG circuit in taVNS-induced analgesia.

a Schematic for retrograde virus labeling to verify anatomical connections in the Auricle→JNG → NTS→vlPAG circuit. Typical images of starter cells in the NTS (b), retrogradely traced JNG neurons (c), and their auricular fibers (d) from 3 replicates. e Schematic for virus injection in optogenetic labeling of vlPAG-projecting NTS neurons that receive auricular vagal afferents. f Mechanical nociceptive thresholds in mice with optogenetic excitation of vlPAG-projecting NTS neurons that receive auricular vagal afferents (n = 5 mice/group, F_{(87, 464)} = 18.71, P < 0.0001). Heatmaps (g) of optogenetically-induced CPP and quantification of preference ratio (h, sham, n = 7; SNI mCherry, n = 7; SNI ChR2, n = 9; F_{(1, 26)} =  13.60, P < 0.0001). i Quantitative analysis of cold allodynia-related coping behaviors in mice following optogenetic activation of vlPAG-projecting NTS neurons that receive aJNG inputs (n = 6 mice/group; F_{(15, 100)} = 4.383, P < 0.0001). j Schematic for retrograde virus labeling and chemogenetic inhibition of vlPAG-projecting NTS neurons that also receive aJNG afferents. k Typical images of virus-labeled NTS neurons from 5 replicates. l Mechanical nociceptive thresholds in mice with chemogenetic inactivation of vlPAG-projecting NTS neurons that also receive aJNG inputs (n = 5 mice/group; F_(15,80) = 48.43, P  < 0.0001). Heatmaps (m) of CPP following chemogenetic inactivation and quantification of preference ratio (n, sham, n = 7; SNI EYFP, n = 7; SNI hM4Di, n = 8; F_{(1, 25)} = 81.24, P < 0.0001). o Quantitative analysis of cold allodynia-related coping behaviors following chemogenetic inactivation of vlPAG-projecting NTS neurons that receive aJNG inputs (n = 6 mice/group; F_{(12, 80)} =  10.54, P  < 0.0001). Two-way ANOVA with Bonferroni post hoc analysis for f, h, i, l, n, and o. All data are reported as mean ± SEM. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001; n.s., not significant. See also Supplementary Table 1. Source data are provided as a Source Data file. AP area postrema, DMV dorsal motor nucleus of vagus, L-JNG left jugular-nodose ganglion, NTS nucleus tractus solitarius, vlPAG ventrolateral periaqueductal gray.

To manipulate the vlPAG-projecting aNTS neurons, we employed a dual Cre-lox/FLP-FRT strategy in which (1) an AAV2/1-hSyn-FLP virus was delivered into the concha to induce FLP expression in aNTS neurons, (2) an scAAV-Retro-Cre virus was injected into the vlPAG to express Cre-recombinase in vlPAG-projecting NTS cells, and (3) an AAV2/9-hSyn-Con Fon-ChR2-EYFP virus was administered to the NTS in order to express ChR2 in FLP- and Cre-expressing NTS neurons. Subsequent optogenetic activation of these ChR2-expressing aNTS neurons projecting to vlPAG led to acute and sustained mechanical antinociception even after discontinuing optical stimulation, together with improved CPP and cold-allodynia coping behaviors, in SNI mice (Fig. 6e–i, and Supplementary Movie 7). In addition, chemogenetic silencing of hM4Di-expressing aNTS neurons by systemic i.p. injection of CNO completely abolished the analgesic effects of taVNS in SNI mice, reversing antinociception, CPP, and coping behaviors (Fig. 6j–o, and Supplementary Movie 8). No such effects were observed in sham controls or SNI mice expressing the EYFP vector SNI mice. These findings collectively demonstrated that this Auricle→JNG → NTS→vlPAG circuit was functionally required for the analgesic effects of taVNS.

Discussion

Although several brain regions have been previously identified to participate in taVNS-induced analgesia, the detailed cell-type-specific circuit mechanisms underlying these effects have long remained elusive, lacking definitive evidence. Experiments in the present study demonstrate that electrical stimulation at the concha produced analgesia in mice with neuropathic pain by engaging the descending pain inhibitory system via an Auricle→JNG^Glu → NTS^POMC →vlPAG^Glu circuit.

Unlike the widespread brain activation resulting from direct VNS^33,34, we hypothesized that taVNS might activate a more selective set of brain regions, as the auricular vagus nerves show relatively limited innervation of JNG sensory neurons. Previous fMRI studies in mice have identified NTS and vlPAG as the primary brain regions activated by taVNS^5,9, while brain regions involved in diffuse modulatory systems, such as the LC³⁵ and basal forebrain¹⁷, may serve as downstream targets of the NTS. Other studies have also shown that NTS neurons are anatomically connected to the PAG^36–38, which in turn projects to the RVM to inhibit nociceptive inputs^13,39,40. Here, we observed monosynaptic connectivity between NTS^POMC and vlPAG^Glu neurons and established that the NTS→vlPAG pathway is required for taVNS-induced analgesia in neuropathic pain mice. The parallel between NTS/vlPAG neuronal firing change and nociceptive thresholds suggests that taVNS-induced analgesia arises from synaptic potentiation within the auricular vagal afferent circuit. This scenario is consistent with evidence that high-frequency stimulation can induce long-term potentiation, with effect duration dependent on stimulation parameters^41,42. We propose that taVNS activates NTS^POMC neurons to excite vlPAG^Glu neurons, thereby engaging descending pain modulatory circuits. However, taVNS engages neural circuits beyond those characterized here. For example, the activation of the LC we observed here likely contributes to pupil dilation through sympathetic excitation⁴³.

Although taVNS has demonstrated behavioral antinociceptive effects in multiple studies^20–22, the neural mechanisms driving these outcomes remain unresolved. Several key brain regions, such as NTS and vlPAG are activated by taVNS^5,9, and their connectivity is documented^36–38. However, their causal relationships and cell-type-specific circuits remain poorly understood. Moreover, existing neuroanatomical models of vagal afferent pathways largely focus on fibers originating from internal organs^44,45 or rely on genetic labeling of taVNS-responsive NTS neurons¹¹, rather than tracing projections directly from the auricle. This gap in tracing auricular-specific pathways has left the anatomical basis of taVNS incompletely defined. Building on these prior studies, our research establishes a neuroanatomical basis for the analgesic effects of taVNS via an auricle-to-brain axis (Supplementary Fig. 10). Peripherally, we resolve the neuroanatomical organization of the murine auricular vagus nerve, demonstrating dual jugular-nodose ganglion origins of its fibers—a finding that overturns the single-jugular origin dogma⁴⁶. Centrally, our unprecedented cell-type-specific functional and anatomical dissection mechanistically connects the auricular vagal circuit to central analgesic pathways, providing a neuroanatomical framework for vagus-mediated sensory processing. The FDA has approved VNS to treat refractory epilepsy and depression, supporting its potential for clinical pain management^47–49. In addition, the SPARC (Stimulating Peripheral Activity to Relieve Conditions) program, sponsored by the National Institutes of Health, is focused on advancing our understanding of vagus nerve mapping and physiology in order to define the most effective strategies to stimulate specific nerve fibers for targeted therapeutic benefits⁵⁰. taVNS has also recently gained traction over other treatment options as a non-invasive pain management strategy among the general public. Precise cell-type-specific circuit mapping can facilitate further optimization of electrical stimulation parameters for taVNS (e.g., site selection and intensity) to effectively activate descending pain modulatory pathways. Our current study can also guide mechanistic investigations of other neural pathways relevant to various neurological diseases involving the vagus nerve⁵¹, and sheds light on the anatomical and physiological connections between the auricle and internal organs, especially reflex auriculo-cardiac responses⁵².

Methods

Animals

All experiments were conducted using male 8- to 10-week-old C57BL/6 J, Fos^TRAP2, Ai9(RCL-tdT), Ai6(RCL-ZsGreen), Vglut2-Cre, POMC-Cre or Vgat-Cre mice sourced from Charles River or Jackson Laboratories. Mice were maintained in colonies of 3 to 5 mice per cage in a controlled environment, with temperatures ranging from 23 to 25 °C. Mice had access to standard mouse chow and water ad libitum, and were kept under a 12-h light/dark cycle, with lights switched on from 7:00 am to 7:00 pm. To reduce potential bias, we randomly assigned mice to different experimental groups. The Animal Care Committee of the University of Science and Technology of China approved all experimental procedures.

Establishment of murine pain models

Neuropathic pain

Neuropathic pain was surgically induced through SNI. Surgeries were conducted under anesthesia and maintained with isoflurane, using a 3% induction dose and 1.5% maintenance level. The procedure began with a surgical incision on the right thigh, followed by meticulous dissection to expose the sciatic nerve bundle, composed of sural, common peroneal, and tibial nerves. The common peroneal and tibial nerves were carefully ligated with nonabsorbable 4-0 chromic gut sutures and then transected at their distal ends. In contrast, the sural nerve was left intact. After surgery, the wound was closed with sutures and disinfected using iodophor. Sham surgeries were performed on mice without any nerve damage, ensuring a valid control group for comparison.

Inflammatory pain

To induce inflammatory pain, we administered a controlled dose of 20 µL of complete Freund’s adjuvant (CFA, Sigma-Aldrich) to the plantar surface of the right hind paw of each mouse, using an insulin syringe (BD Corporation, USA) and brief isoflurane anesthesia. Mice in the control group received an equivalent volume of saline (0.9% NaCl) administered intradermally.

Transcutaneous auricular vagus nerve stimulation

In administering trans-auricular nerve stimulation (taVNS or tnVNS), custom-made bipolar steel electrodes were magnetically attached to the auricle to ensure secure placement. For taVNS, electrode positioning was guided by anatomical landmarks targeting the auricular branch of the vagus nerve, as previously reported⁵³. The cathode was strategically placed at the intersection of the concha cavity and the postero-inferior wall of the external acoustic meatus, while the anode was positioned at the cymba concha. tnVNS involved placing electrodes outside the vagus nerve distribution on the exterior margin of the auricle (see Fig. 1 for specific locations), maintaining a distance of at least 2 mm between electrodes. Electrode leads were connected to a stimulator (HuaTuo Electronic Acupuncture, China) to deliver the electrical currents at alternating frequencies, with a ratio of 3:15 Hz. During stimulation, mice were anesthetized using isoflurane (0.5–1.0%), and body temperature was maintained at 36 °C. In general, each stimulation session was conducted for 30 min, followed by a brief recovery period on a heating pad to regain consciousness. Subsequently, mice were transferred to the testing chambers for mechanical threshold measurements. Throughout the process, mice were closely monitored for any signs of discomfort or distress, with adjustments to stimulation intensity or termination of the experiment if necessary. Of note, since stimulation of the right vagus nerve can reportedly adversely affect heart rate due to preferential innervation of the sinoatrial node compared with the left vagus nerve, all trans-auricular stimulations in the current study were performed on the left auricle. For taVNS administration in the awake state, head-fixed mice were habituated to the setup for over 3 days. During stimulation, the mice were able to move freely on a turntable⁵⁴.

Jugular-nodose ganglion manipulations

Mice were fully anesthetized with isoflurane in oxygen (3% for induction, 1.5% for maintenance) and secured in a stereotaxic device (RWD, Shenzhen, China). Their body temperature was maintained at a constant 36 °C throughout the experiment. Ophthalmic ointment was applied to the eyes to prevent dehydration. The mice were placed in a supine position on the surgical table, exposing their neck region for easy access. The surgical procedure was performed with extreme care, systematically peeling away layers of skin and muscle to expose the vagus nerve and its branches. Utmost care was taken to maintain a clear and sterile surgical field, safeguarding against accidental damage to surrounding blood vessels and nerves. Under a stereomicroscope, the jugular-nodose ganglion (JNG), discernible as a distinct round or oval structure, was located along the main trunk or branches of the vagus nerve. Accurate localization was achieved by cross-referencing anatomical atlases with real-time observations, ensuring precision targeting. The vagus nerve was then gently separated from the common carotid artery to facilitate subsequent virus injection, and surrounding connective tissue was carefully trimmed to minimize potential obstruction⁵⁵.

To inject the virus into the JNG, a 10-μL Hamilton microsyringe (Hamilton, USA) was affixed to a handheld micro 4 injection pump (WPI, USA), controlled remotely with a foot pedal. A total of 300–400 nL of virus suspension was delivered into the JNG at multiple target sites while ensuring the needle remained stability to avoid ganglion damage or tissue injury. After injection, the needle was retained in place for at least 1 min before withdrawal. Once the surgery was completed, the mice were carefully placed on a heating mat for recovery and then returned to their cages.

Stereotaxic surgery

Mice were anesthetized using isoflurane (3% for induction and 1.5% for maintenance) and secured in a stereotaxic frame (RWD, Shenzhen, China). A glass micropipette with a 10–15 μm tip diameter, connected to a 10 μL syringe (Hamilton, USA), was used for virus delivery. A craniotomy was conducted with the dura removed. A micro-infusion pump (micro-4, WPI, USA) was used to precisely control the injection procedure. Upon completing the injection, the pipette was maintained in place for several minutes to prevent backflow, then slowly withdrawn. The coordinates for the injections were as follows: NTS (AP, −7.64 mm from bregma; ML, ±0.40 mm; DV, −3.75 mm), vlPAG (AP, −4.72 mm from bregma; ML, ±0.50 mm; DV, −1.80 mm), DMV (AP, −7.76 mm from bregma; ML, ±0.37 mm; DV, −3.89 mm) and Amb (AP, −7.48 mm from bregma; ML, ±1.20 mm; DV, −4.70 mm). Following surgery, the mice were placed on a heating mat before being returned to their cages to recover.

Mapping c-Fos expression

Fos^TRAP2::Ai9 mice were bred by crossing Fos^TRAP2 with Ai9 mice to investigate c-Fos expression patterns in response to taVNS. A stock solution of 4-hydroxytamoxifen (4-OHT, 50 mg/mL in DMSO, stored at −20 °C) was diluted in corn oil to a 5 mg/mL concentration. Mice received a single intraperitoneal dosage of 4-OHT (40 mg/kg) 1 h before the taVNS procedure. As a control group, Fos^TRAP2::Ai9 mice underwent the same 4-OHT treatment but without undergoing taVNS. Notably, the inducible Cre-recombinase expression in the Fos^TRAP system takes several days to fully develop following 4-OHT injection; therefore, animals were allowed to rest for 6–7 days before being perfused.

Virus injections

Characterizing Auricle→JNG→NTS circuit

To trace connections between the auricle, JNG, trigeminal ganglion (TG), and other brain regions in C57 mice, rAAV2/9-CAG-EGFP-WPRE-hGH (AAV2/9-CAG-EGFP, 3.66 × 10¹² vg/mL, BrainVTA) was administered to the JNG or TG for anterograde tracing. To visualize NTS neurons specifically innervated by auricular vagal afferents (aNTS) and their projections, C57 mice received scAAV2/1-hSyn-Cre-WPREs (AAV2/1-hSyn-Cre, 1.08 × 10¹³ vg/mL, BrainVTA) in the concha, followed by Cre-dependent AAV2/9-hEF1α-DIO-EGFP-WPRE-pA (AAV2/9-DIO-EGFP, 2.0 × 10¹² vg/mL, BrainVTA) injection into the NTS ipsilaterally. To further identify the cell subtype of these aNTS neurons, a rAAV2/1-hSyn-FLP-WPRE-hGH (AAV2/1-hSyn-FLP, 1.03 × 10¹³ vg/mL, BrainVTA) was injected into the concha, followed by ipsilateral injection of FLP-dependent rAAV2/9-EF1α-fDIO-mCherry-mCherry-WPRE-hGH (AAV2/9-fDIO-mCherry, 5.05 × 10¹² vg/mL, BrainVTA) and Cre-dependent AAV2/9-DIO-EGFP viruses in the NTS of Vglut2-Cre, POMC-Cre, or Vgat-Cre mice. In a separate experimental group, C57 mice received an injection of AAV2/1-hSyn-FLP into the concha, followed by an ipsilateral injection of AAV2/9-fDIO-mCherry into the NTS, along with subtype-specific reporter viruses, including rAAV2/9-TH-EGFP-WPRE-PA (AAV2/9-TH-EGFP, 5.00 × 10¹² vg/mL, BrainVTA), rAAV2/9-CCK-EGFP-WPRE-hGH polyA (AAV2/9-CCK-EGFP, 5.0 × 10¹² vg/mL, BrainVTA), or rAAV2/9-NPY-EGFP-WPRE-bGH polyA (AAV2/9-NPY-EGFP, 5.27 × 10¹² vg/mL, BrainVTA).

Characterizing NTS→vlPAG circuit

Monosynaptic tracing from the NTS to the vlPAG was performed by injecting scAAV2/1-hSyn-Cre into the NTS and ipsilateral injection of a Cre-dependent AAV2/9-EF1α-DIO-mCherry-WPRE-hGH virus (AAV2/9-DIO-mCherry, 6.76 × 10¹² vg/mL, Brain Case) in the vlPAG of C57 mice. To highlight the extensive projections originating in NTS proopiomelanocortinergic neurons (NTS^POMC), POMC-Cre mice received a direct injection of AAV2/9-DIO-mCherry into the NTS. To label vlPAG neurons innervated by NTS^POMC neurons, we injected an anterograde AAV2/1-POMC-Cre virus into the NTS of C57 mice, and administered a Cre-dependent AAV2/9-DIO-EGFP virus to the ipsilateral vlPAG. After 3 weeks, mice were euthanized, and brain slices containing the NTS or vlPAG were stained with antibodies targeting GABA, glutamate, TH or PLCβ4.

Characterizing Auricle→JNG→NTS→vlPAG circuit

To further map the whole circuit of the Auricle→JNG→NTS→vlPAG pathway, an anterograde rAAV2/1-hSyn-FLP-WPRE-hGH (AAV2/1-hSyn-FLP, 1.03 × 10¹³ vg/mL, BrainVTA) virus was administered to the concha, an FLP-dependent AAV2/9-fDIO-mCherry virus was injected into the NTS, and a retrograde scAAV2/R-hSyn-EGFP-WPREs (scAAV-Retro-EGFP, 5.36 × 10¹² vg/mL, BrainVTA) virus was injected into the ipsilateral vlPAG of C57 mice.

To visualize retrogradely labeled aJNG neurons originating from the auricle, C57 mice received scAAV-Retro-EGFP, which retrogradely transports to the cell body after injection into the concha. To further trace the Auricle→JNG→NTS circuit, scAAV-Retro-EGFP viruses were injected into the concha, while a scAAV2/R-hSyn-mCherry-WPREs (scAAV-Retro-mCherry, 5.17 × 10¹² vg/mL, BrainVTA) virus was concurrently administered to the NTS of C57 mice, which allowed visualization of JNG neurons with both auricle and NTS projections. To visualize vlPAG-projecting NTS neurons, we injected scAAV-Retro-mCherry into the vlPAG of C57 mice. After 3 weeks, mice were euthanized, and brain slices containing the NTS were stained with GABA-, glutamate-, TH-, or POMC-targeting immunofluorescent antibodies. In another set of experiments, C57 mice underwent injection of rAAV2/R-hSyn-Cre-WPRE-hGH (AAV-Retro-Cre, 1.03 × 10¹³ vg/mL, BrainVTA) into the vlPAG, with subsequent rabies virus-based retrograde tracing by injection of rAAV2/8-EF1α-DIO-mCherry-F2A-TVA-WPRE-hGH (AAV2/8-DIO-TVA-mCherry, 5.59  × 10¹² vg/mL, BrainVTA) and rAAV2/8-EF1α-DIO-oRVG-WPRE-hGH (AAV2/8-DIO-RVG, 5.08 × 10¹² vg/mL, BrainVTA) into the contralateral NTS with complementary administration of RV-CVS-EnVA-N2C(△G)-EGFP (RV-EnVA-EGFP, 2.00 × 10⁸ IFU/mL, BrainVTA) at the same NTS site to trace the Auricle→JNG→NTS→vlPAG circuit. The anterograde trans-multisynaptic H129-hUbC-HBEGFP virus (HSV-EGFP; 1.00 × 10⁹ PFU/mL, Brain Case) was injected into the auricular concha to map neural connectivity between the auricular vagal circuit and the spinal circuit.

Genetically labeling Auricle→JNG→NTS circuit

To optogenetically stimulate auricular vagal afferents, C57 mice were delivered an AAV-Retro-Cre virus in the concha, followed by administration of a Cre-dependent AAV2/9-CAG-DIO-ChR2-mCherry virus (AAV2/9-DIO-ChR2-mCherry, 5.31 × 10¹² vg/mL, BrainVTA) to the JNG. An optical fiber was implanted towards the NTS and anchored to the skull using dental cement. For chemogenetic inhibition, scAAV-Retro-Cre was first injected into the concha, followed by Cre-dependent AAV2/9-CAG-DIO-hM4Di-mCherry (AAV2/9-DIO-hM4Di-mCherry, 5.44 × 10¹² vg/mL, BrainVTA) injection into the JNG. Bilateral NTS cannulation (250 μm diameter, RWD) was performed, with the cannula cemented to the skull. For optogenetic stimulation of aNTS neurons, the left concha of C57 mice was injected with an AAV2/1-hSyn-Cre virus followed by administration of a Cre-dependent AAV2/9-DIO-ChR2-EGFP in the NTS. Vector control mice underwent the same procedure as the experimental groups, receiving an equivalent volume of scAAV-Retro-Cre at the concha, but instead receiving a Cre-dependent AAV2/9-DIO-mCherry or AAV2/9-DIO-EGFP (5.57 × 10¹² vg/mL, BrainVTA) virus at the JNG or NTS.

Genetically labeling NTS→vlPAG pathway

To chemogenetically inhibit the vlPAG-projecting NTS^Glu or NTS^POMC neurons, Vglut2-Cre or POMC-Cre mice were administrated with scAAV2/R-CMV-DIO-FLP-WPREs (AAV-Retro-DIO-FLP, 5.37 × 10¹² vg/mL, BrainVTA) bilaterally in the vlPAG, while rAAV2/9-hSyn-fDIO-hM4D(Gi)-mCherry-WPRE-hGH (AAV2/9-fDIO-hM4Di-mCherry, 5.27 × 10¹² vg/mL, BrainVTA) was injected into the NTS. Mice given an FLP-dependent AAV2/9-fDIO-mCherry served as the vector control group. To chemogenetically inhibit vlPAG neurons receiving inputs from NTS^POMC neurons, C57 mice received bilateral injections of rAAV2/1-POMC-Cre-WPRE-PA (AAV2/1-POMC-Cre, 5.18 × 10¹² vg/mL, BrainVTA) into the NTS and rAAV2/9-hSyn-DIO-hM4D(Gi)-P2A-EGFP (AAV2/9-DIO-hM4Di-EGFP, 2.00 × 10¹² vg/mL, Brain Case) into the vlPAG. For optogenetic excitation of the same circuit, a separate cohort of mice received AAV2/1-POMC-Cre in the NTS and rAAV-hSyn-DIO-hChR2(H134R)-P2A-EGFP (AAV2/9-DIO-ChR2-EGFP, 2.00 × 10¹² vg/mL, Brain Case) in the vlPAG. Mice given a Cre-dependent AAV2/9-DIO-EGFP served as the vector control group.

Genetically labeling Auricle→JNG→NTS→vlPAG circuit

To optogenetically activate the Auricle→JNG→NTS→vlPAG circuit, C57 mice were administered AAV2/1-hSyn-FLP into the concha, followed by AAV-Retro-Cre injection into the vlPAG. Subsequently, a rAAV2/9-hSyn-Con Fon-hChR2(H134R)-EYFP-WPRE-hGH polyA (AAV2/9-hSyn-Con Fon-ChR2-EYFP, 8.13 × 10¹² vg/mL, BrainVTA) virus was introduced into the NTS. For chemogenetic inhibition, the same procedure was employed: AAV2/1-hSyn-FLP was delivered to concha and AAV-Retro-Cre was injected in the vlPAG, but a FLP/Cre-dependent rAAV-hSyn-Con Fon-hM4D(Gi)-EYFP-WPRE-hGH polyA (AAV-hSyn-Con Fon-hM4Di-EYFP, 4.94 × 10¹² vg/mL, BrainVTA) virus was injected bilaterally into the NTS of mice. Control mice were administered an equivalent volume of rAAV2/9-hSyn-Con Fon-EYFP-WPRE-hGH polyA (AAV2/9-hSyn-Con Fon-EYFP, 5.57 × 10¹² vg/mL, BrainVTA) without the optogenetic or chemogenetic components.

Chemogenetically labeling taVNS-activated neurons in specific brain regions

To chemogenetically label taVNS-activated neurons in targeted brain regions, Fos^TRAP2::Ai9 mice were bilaterally injected with AAV2/9-DIO-hM4Di-EGFP into the vlPAG and PBN, rAAV2/9-ChAT-DIO-hM4D(Gi)-EGFP-WPREs (AAV2/9-ChAT-DIO-hM4Di-EGFP, 3.02 × 10¹² vg/mL, BrainVTA) into the DMV and Amb; rAAV2/9-TH-DIO-hM4D(Gi)-EGFP-WPREs (AAV2/9-TH-DIO-hM4Di-EGFP, 5.59 × 10¹² vg/mL, BrainVTA) into the LC. 9 days post injection, 4-OHT was administered intraperitoneally to induce Cre-dependent recombination, followed by taVNS stimulation. After a recovery period of 13 days to allow for viral expression, mice underwent behavioral testing with CNO to inhibit hM4Di-expressing neurons. Control mice received bilateral injections of AAV2/9-DIO-EGFP and underwent identical taVNS, 4-OHT, and CNO protocols.

Optogenetic and chemogenetic manipulations

Before behavioral experiments, mice underwent a 3-day acclimation period in the testing chamber, after which they were transferred to the testing room for a 4-h acclimation period on the experimental day. For optogenetic manipulation, mice were anesthetized with isoflurane, and their chronically implanted 200-μm diameter fibers (Newdoon, China) were connected to a blue laser generator (Fiblaser, Shanghai) via optical fiber sleeves. Following connection to the laser generator, the mice were transferred to a mechanical threshold testing chamber, and allowed to acclimate for at least 30 min. Blue light (473 nm, 5–8 mW, 15-ms pulses, 3/15 Hz) was triggered using a Master-8 pulse stimulator (A.M.P.I., Israel) to activate auricular peripheral or central vagus fibers, or the JNG→NTS→vlPAG circuit.

For chemogenetic manipulation, CNO (1 mg/kg, Sigma-Aldrich) was administered intraperitoneally or locally under isoflurane anesthesia, followed by a 30-min waiting period before behavioral testing. Control animals underwent the same procedures. After all tests, mice were euthanized to verify delivery sites, ensuring the accuracy of virus injection and fiber placement. Any mice with mistargeted brain regions were excluded from data analysis. Brain slice schematics were created using Adobe Illustrator, referencing Paxino’s brain atlas for anatomical accuracy.

Subdiaphragmatic vagotomy

Mice were anesthetized with isoflurane (1.5–2.0%) and positioned in a sterile surgical field. A midline laparotomy was performed to expose the abdominal cavity. The diaphragm was gently retracted to reveal the vagus nerve, which was carefully identified and bilaterally sectioned below the diaphragm using fine surgical scissors. The severed nerve ends were isolated to prevent regeneration. The abdominal muscles were sutured, and the skin incision was closed with surgical staples⁵⁶. A 2–3 days recovery period was allowed before any experimental procedures were conducted.

Serum TNF-α and catecholamine measurements

Serum TNF-α levels were measured using a double-antibody sandwich ELISA kit (E-HSEL-M0009, Elabscience, Wuhan, China) according to the manufacturer’s protocol. Briefly, microplates pre-coated with anti-mouse TNF-α antibodies captured TNF-α from serum samples or standards, followed by sequential addition of biotinylated anti-mouse TNF-α antibodies and horseradish peroxidase (HRP)-labeled avidin. After adding the chromogenic substrate 3,3’,5,5’-tetramethylbenzidine (TMB), HRP catalyzed the oxidation of TMB to a blue product, which turned yellow upon the addition of a stop solution. The optical density (OD) was measured at 450 nm using a microplate reader. TNF-α concentrations were calculated by interpolating the sample OD450 values against a standard curve.

Norepinephrine (NE), epinephrine (E), and dopamine (DA) levels were quantified using competitive ELISA kits (NE: EU2565; E: EU2563; DA: EU0392; FineTest, Wuhan, China) according to the manufacturer’s instructions. Pre-coated plates contained immobilized NE, E, or DA. Sample or standard NE/E/DA competed with the pre-coated ligands for binding to a fixed amount of biotinylated anti-NE/E/DA antibodies. Unbound components were washed away, followed by the addition of HRP-streptavidin (SABC). TMB substrate was added, and the reaction proceeded as described above. OD450 values were measured and converted to concentrations using curve-fitting software to compare against a standard curve.

In vivo multi-tetrode recordings

A custom-designed, movable 8-tetrode array, each comprising four twisted, 12.5 μm diameter platinum/iridium wires from California Fine Wire (Grover Beach, CA), was implanted in the left NTS or vlPAG of mice for long-term extracellular recordings. The electrode array was housed in a screw-driven microdrive scaffold securely anchored to the skull using dental cement. Mice were allowed to recover for at least 3 days prior to starting the recordings. Electrodes were lowered in 70  μm increments to access different neuronal populations vertically.

An optrode was assembled to combine electrophysiology and optogenetics by embedding an optical fiber with the tetrodes, positioning the fiber tip about 200  μm above the electrode tips. Neuronal signals were acquired using a 32-channel headstage and Neurostudio amplifier (Greathink Medical Technology, China). The raw data were filtered between 300 and 5000 Hz to extract spikes. The T-Dis E-M algorithm built in Offline Sorter 4 (Plexon, USA) was employed for spike sorting, and Neuroexplorer 5 (Nex Technologies, USA) was used to quantify firing rates. After completing all recordings, tetrode placement was verified in all mice, and data from mice with incorrect targeting were excluded from further analysis.

Vagus nerve recording

Mice were induced and maintained under deep anesthesia with isoflurane (3% for induction and 1.5% for maintenance) while lying supine. The neck region was shaved and sterilized prior to surgery. A 2 cm-long incision was made to expose the left vagus nerve trunk, and a cuff electrode (KedouBC, China) was secured around the upper part of the nerve, while a ground electrode was placed between the salivary gland and the skin⁵⁷. Neuronal signals from the recording electrodes were amplified and stored using a Neurostudio amplifier and data acquisition software (Greathink Medical Technology, China).

Microendoscopic calcium imaging

Microendoscopic calcium imaging was utilized to monitor neuronal activity in vivo. A GRIN lens was implanted in the target brain by cranial surgery. The GRIN lens was coupled to a microendoscope (Inscopix, USA), enabling real-time imaging of neuronal populations expressing rAAV2/9-EF1α-DIO-GCaMP6m-WPRE-pA virus (AAV2/9-DIO-GCaMP6m, 5.22 × 10¹² vg/mL, Brain Case) injected the target region of Vglut2-Cre, POMC-Cre, or Vgat-Cre mice. GCaMP6m signal detection allowed quantitative analysis of calcium transients associated with neuronal activity. A low-power 488 nm laser was utilized to excite the GCaMP fluorescence, and the emitted light was captured by a high-speed camera. Neuronal activity was continuously monitored, with real-time fluorescence data (ΔF/F) recorded at a frequency of 30 Hz.

The raw fluorescence data were processed using the Mosaic software (Inscopix) to correct for motion, subtract background noise, and segment individual neuron activity. Regions of interest corresponding to individual neurons were identified based on fluorescence intensity and spatial patterns. Calcium transients were detected by identifying significant fluctuations in fluorescence (ΔF/F) over time. The frequency of calcium events was analyzed to assess neuronal activity patterns in response to taVNS or optogenetic stimulation.

Physiological recordings of vital signs

Electrocardiogram (ECG) recordings were conducted on isoflurane-anesthetized mice using two needle electrodes placed on the forelimbs and hind limbs. Respiratory activity was measured in head-fixed, awake mice with an airflow sensor (AMW3300V, Honeywell, USA) positioned adjacent to the nose⁵⁸. The voltage signals were filtered (30–300 Hz), amplified using a DAM50 amplifier (WPI, USA), digitized with a MiniDigi (Molecular Devices, USA), and recorded with Axonscope 11 software (Molecular Devices, USA).

Brain slice patch clamp recordings

Brain slice preparation

After anesthesia, mice underwent cardiac perfusion with 20 mL of ice-cold NMDG-acetylated sodium guanidinium chloride (NMDG-ACSF) solution, which contained the following components (in mM): 93 NMDG, 1.2 NaH₂PO₄, 2.5 KCl, 20 HEPES, 30 NaHCO₃, 5 Na-ascorbate, 2 thiourea, 25 glucose, 0.5 CaCl₂, 10 MgSO₄, 3 Na-pyruvate, and 3 GSH. The mice were then decapitated, and their brains were rapidly mounted on a vibrating microtome (VT1200s, Leica). Coronal slices containing the NTS or vlPAG region, 270 μm thick, were cut at a speed of 0.12 mm/s in an ice-cold, oxygenated NMDG-ACSF. After slicing, the brain slices were incubated for 12 min at 32 °C in NMDG-ACSF, followed by a 1-h incubation in HEPES-ACSF, which included (in mM): 92 NaCl, 2.5 KCl, 20 HEPES, 30 NaHCO₃, 1.2 NaH₂PO₄, 25 glucose, 5 Na-ascorbate, 3 Na-pyruvate, 2 MgSO₄, 2 CaCl₂, 2 thiourea, and 3 GSH. The slices were maintained at 28 °C for at least an hour before being transferred to a recording chamber (RC-27, Warner Instruments).

Whole-cell patch clamp recordings

Whole-cell patch clamp recordings were then performed on brain slices continuously perfused with oxygenated standard ACSF, which had a composition of 129 NaCl, 3 KCl, 3 HEPES, 2.4 CaCl₂, 20 NaHCO₃, 10 glucose, 1.3 MgSO₄, and 1.2 KH₂PO₄. All ACSF solutions were maintained at a pH of 7.3-7.4, an osmolarity of 300–310 mOsm/kg, and carbogen with a 95% O₂/5% CO₂. During the recording, an infrared-differential interference contrast microscope (Olympus BX51Wl) was used to visualize neurons in the target areas. Recording electrodes, with tip resistances ranging between 5 and 9 MΩ, were pulled from 1.5 mm borosilicate glass capillaries (VitalSense Scientific Instruments Co., Ltd) using a horizontal electrode puller (P1000, Sutter Instruments) and filled with a potassium gluconate-based internal solution containing 130 K-gluconate, 5 KCl, 2 MgCl₂, 10 HEPES, 0.6 EGTA, 0.3 Na-GTP, and 2 Mg-ATP, adjusted to a pH of 7.2–7.3 and an osmolarity of 285-290 mOsm/kg. Neurons were voltage-clamped at −70 mV for membrane current recordings and in current-clamp mode at 0 pA for membrane potential recordings. Light-evoked postsynaptic currents were recorded using a 200-μm-diameter optical fiber positioned 0.2 mm above the target area. To isolate the monosynaptic light-evoked response, we added tetrodotoxin (1 μM) and 4-aminopyridine (2 mM) to the standard ACSF to block sodium and potassium channels. All data were acquired using a Multiclamp 700B amplifier, an Axon 1550B digitizer, and pCLAMP 10.7.0 data acquisition software (Molecular Devices, USA), with a low-pass filter set at 2.8 kHz and a 10 kHz sampling rate. Data were offline analyzed using Clampfit 11 software (Molecular Devices, USA).

Morphology and imaging

Mice were initially anesthetized with pentobarbital (20 mg/kg, intraperitoneal injection) before undergoing transcardial perfusion with 0.9% saline, followed by fixation with 4% paraformaldehyde in 0.1 M phosphate buffer. The brain and vagal nerve ganglia were extracted and post-fixed for at least 24 h at 4 °C. Cryoprotection was achieved by immersion in a 30% sucrose solution, and 40 μm coronal brain sections or 12 μm vagal nerve ganglia slices were prepared using a Leica CM1860 cryostat (Leica, Germany). These slices were stored in a cryoprotectant solution containing 30% glycerol, 20% ethylene glycol, and PBS at −20 °C for future staining or imaging.

We carefully removed hair to prepare auricle samples, and samples were fixed for 24 h in 4% PFA. Prior to fluorescence microscopy, auricles were thoroughly rinsed with PBS (three times, 10 min per wash), and the connective and adipose tissues were removed to expose the epidermal layer.

For immunostaining, slices underwent a 30-min incubation in 0.3% Triton X-100 to facilitate permeabilization. A blocking step was then performed using 10% donkey serum for 1 h at room temperature to prevent non-specific reactions. Primary antibodies, including anti-glutamate (1:500, rabbit, Sigma, G6642) and anti-GABA (1:500, rabbit, Sigma, A2052), anti-PLCβ4 (1:200, mouse, SantaCruz, sc-166131), anti-POMC (1:100, rabbit, Cell Signaling, #23499), anti-tyrosine hydroxylase (1:500, rabbit, Cell Signaling, #58844), anti-choline acetyltransferase (1:500, goat, Sigma, SAB2500236), were diluted in a blocking solution containing 0.3% Triton X-100 and 10% donkey serum in PBS, and incubated at 4 °C for 24 h. After washing thoroughly with PBS, slices were incubated with Alexa 594-Donkey anti-goat IgG cross-adsorbed secondary antibody (1:500, Thermo Fisher, A-11058), Alexa 594-Donkey anti-Rabbit IgG Highly cross-adsorbed secondary antibody (1:500, Thermo Fisher, A-21207), Alexa 594-Donkey anti-Mouse IgG Highly cross-adsorbed secondary antibody (1:500, Thermo Fisher, A-21203), Alexa 488-Donkey anti-Rabbit IgG Highly cross-adsorbed secondary antibody (1:500, Thermo Fisher, A-21206) or Alexa 488-Donkey anti-goat IgG cross-adsorbed secondary antibody (1:500, Thermo Fisher, A-11055) for 2 h at room temperature. DAPI (1:2000, Sigma) was added for 5 min, followed by three washes with PBS and mounting. Fluorescence imaging was conducted using a Zeiss LSM980 or an Olympus FV3000 microscope. Quantitative image analysis, including cell counts and signal co-localization, was performed using ImageJ software (Fiji edition, NIH, USA).

Fluorescence In situ hybridization

The complete JNG was harvested from adult mice at 3 weeks after unilateral auricular injection of retrograde scAAV-Retro-EGFP. Tissues were cryoprotected, embedded, and sectioned at 12 μm on a cryostat. Probes for fluorescence in situ hybridization (FISH) were provided by Spatial FISH Company, and sections were processed using a π-FISH kit (Spatial FISH, RFR001-30) according to the manufacturer’s instructions. Single-fluorophore detection was conducted using commercial Alexa647-conjugated probes targeting Phox2b (Spatial FISH, RFP00211T3) and Prdm12 (Spatial FISH, RFP00212T3) transcripts. Finally, sections were counterstained with DAPI and imaged using an LSM980 confocal microscope.

Behavioral tests

The individuals conducting behavioral tests were distinct from those researchers administering the electrical or optogenetic stimulations of auricular vagus nerve to ensure that behavioral assessments were blinded.

von Frey and cold plate tests

To minimize stress before testing, mice underwent a 3-day acclimation period in a designated testing room. Each mouse was individually housed in a transparent plastic chamber and securely placed on a wire mesh grid for 30 min daily. The mechanical withdrawal threshold was measured using a series of precisely calibrated von Frey filaments. These filaments were applied perpendicularly to the plantar surfaces of hind paws, exerting just enough force to cause a subtle bend. The mechanical response threshold was the lowest force that provoked a rapid paw withdrawal, flinching, or licking response. If no reaction was observed, a filament with greater force was used, and the process was repeated three times for each paw to ensure consistency and accuracy of the results. Cold sensitivity was determined by observing the latency of hindpaw withdrawal from a cold plate (BIO-CHP, BiosebLab, France) set to 5 °C, with a 30-s cutoff to prevent paw injury.

Pupil recordings

Mice were anesthetized with isoflurane and positioned in a stereotaxic frame. A custom-designed headpost was securely affixed to their skull using dental cement. This headpost was integrated with a holder for stable head positioning. To ensure comfort and minimize stress, the mice underwent training to familiarize themselves with the fixation setup. Pupil dilation was monitored using a web camera equipped with a 16-mm lens. Facemap software⁵⁹ was employed to precisely extract pupil size changes, which were then normalized to a Z-score using the formula Z = (x− m)/d⁶⁰.

Conditioned place preference tests

To minimize stress before experiments, mice underwent a 3-day acclimation period in a standardized behavioral environment. A specialized setup featuring two interconnected chambers (40 × 20 cm) connected by a “neck” structure was employed to study changes in preference potentially resulting from optogenetic manipulation or taVNS. On the first day, mice were given a 15-min free exploration session without any stimulation, while the chamber with less initial exploration time was designated as the stimulation side. Over the following 5 days, conditioning took place in two sessions per day, one in the morning and one in the afternoon, each lasting 30 min with a 4-h interval between sessions. During the morning session, mice received no stimulation and were placed in the chamber; they spent more time during the baseline period. In the afternoon session, mice received taVNS or optogenetic stimulation while placed in the chamber, they spent less time in initially. On the seventh day, a preference test was conducted without stimulation wherein mice were allowed a 15-min exploration period. Video recordings of their movements were analyzed offline using EthoVision XT software (Noldus, Netherlands) to determine the time spent in the stimulation-paired chamber (Supplementary Fig. 9c–f). The preference ratio was calculated by comparing the time spent in the stimulation chamber on the seventh day (Post) to the time on the initial day (Pre).

Coping behaviors to cold stimulus

Mouse coping behaviors were assessed through video recordings to capture responses to cold exposure. Three cameras were strategically positioned around the testing chambers with their lenses position level to and parallel with the chamber walls. Cameras were set to record at 30 frames per second to monitor spontaneous behaviors throughout the test.

Mice were placed on a cold plate (BIO-CHP, BiosebLab, France) set at 3 °C for a 3-min duration. To minimize external disturbances, the chamber lid was kept closed. The recorded videos were analyzed offline to determine the duration of hindpaw lifting as a measure of cold allodynia-related coping behaviors.

Statistical analysis

Statistical tests were selected based on data assessment, with normal distribution verified by Shapiro–Wilk test. Non-parametric tests, like the Mann–Whitney U or Wilcoxon matched-paired signed rank tests, were employed for non-normal datasets. Both paired and unpaired two-tailed Student’s t-tests were conducted to compare two groups. Two-way analysis of variance (ANOVA) was employed to analyze data across multiple groups, and post-hoc comparisons were made using the Bonferroni test. All data are reported as mean ± standard error of the mean (SEM). Statistical significance was determined at the following levels: * for P < 0.05, ** for P < 0.01, and *** for P < 0.001. All statistical analyses and graphical representations were conducted using GraphPad Prism 8.0.2 software (GraphPad Software, USA). All illustrations in the manuscript were created by the authors using Adobe Illustrator (Adobe Systems Incorporated, USA), except Supplementary Fig. 10, which was created in Procreate (Savage Interactive Pty Ltd, Australia).

Reporting summary

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

Author contributions

Conceptualization: Haitao W., and Z.Z. Methodology: X.P., Yehao L., Yi L., Y.M, Q.X., A.L. Investigation: X.P., Yehao L., Yi L., Y.M., Q.X., A.L., Q.D., Y.T., B.L., and Hao W. Visualization: X.P., Haitao W., and Z.Z. Funding acquisition: Hao W., F.K., Haitao W., and Z.Z. Project administration: Haitao W., and Z.Z. Supervision: Haitao W., and Z.Z. Writing – original draft: Haitao W., and Z.Z. Writing – review & editing: Y.W., F.K., Haitao W., and Z.Z.

Data availability

All data necessary to understand and assess the conclusions of this study are available in the main text or the supplementary materials. There are no restrictions on data availability in the manuscript. Source data are provided with this paper.

Code availability

Facemap was used to analyze pupil size changes as described in ref. ⁵⁹, with its code and GUI publicly available at https://github.com/mouseland/facemap.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72214-0.