Exploring brain-gut interaction mechanisms in Transcutaneous auricular Vagus Nerve stimulation for Major Depressive Disorder

Abstract

Background

The gut microbiota is intricately implicated in the pathogenesis of Major Depressive Disorder (MDD), with the vagus nerve serving as a key regulatory bridge. Transcutaneous Auricular Vagus Nerve Stimulation (taVNS) has emerged as a promising non-invasive therapeutic strategy for MDD by modulating the gut-brain axis, yet the precise brain-gut interaction mechanisms underlying its antidepressant effects remain poorly characterized. This study is a registered clinical trial (ChiCTR2200059591; Registered 4 May 2022; https://www.chictr.org.cn).

Objective/hypothesis

This study aimed to verify the clinical efficacy of taVNS for MDD and elucidate the underlying brain-gut crosstalk mechanisms, by integrating comprehensive clinical assessments, resting-state functional magnetic resonance imaging (rs-fMRI) neuroimaging data and gut metagenomic profiling.

Methods

Ninety-five patients diagnosed with MDD were randomly allocated at a 1:1 ratio to either the active taVNS group (auricular concha stimulation) or the sham taVNS group (superior concha of mid-helix stimulation). Eighty patients (40 per group) completed the entire intervention course and were included in the final statistical analysis. All participants underwent 30-minute stimulation twice daily (4/20 Hz, 3–8 mA) for 8 consecutive weeks (5 days per week). Standardized clinical assessments were administered at baseline and post-intervention, including the 17-item Hamilton Depression Rating Scale (HAMD-17), 14-item Hamilton Anxiety Rating Scale (HAMA-14), and Gastrointestinal Symptom Rating Scale (GSRS). Rs-fMRI was performed to quantify core neural activity metrics, including amplitude of low-frequency fluctuation (ALFF), fractional ALFF (fALFF), regional homogeneity (ReHo), and degree centrality (DC); fecal samples were collected for high-throughput metagenomic analysis. Spearman correlation analysis and mediation analysis were further conducted to dissect the interactive relationships between brain neural activity and gut microbiota.

Results

The active taVNS group achieved significantly superior clinical efficacy relative to the sham group, with a HAMD-17 response rate of 62.50% and remission rate of 35.00%, versus 30.00% and 2.50% in the sham group (all P < 0.05). Rs-fMRI analyses revealed significant group×time interaction effects on neural activity: decreased ALFF in the right calcarine sulcus; altered fALFF in the right inferior temporal gyrus, left cuneus, right superior frontal gyrus (SFG) and right angular gyrus; reduced ReHo in the right calcarine sulcus and bilateral insula; and increased DC in the right caudate nucleus and left anterior cingulate gyrus. Gut microbiota profiling identified anaerobic butyrate-producing bacteria and Faecalibacterium prausnitzii as potential biomarkers linked to taVNS therapeutic effects. HAMD-17 scores were negatively correlated with Faecalibacterium prausnitzii abundance (r=-0.566, P < 0.01) and positively correlated with anaerobic butyrate-producing bacteria abundance (r = 0.406, P < 0.01). Mediation analysis suggested that fALFF values in the right SFG may indirectly modulate depressive symptoms via regulating Faecalibacterium prausnitzii abundance (indirect effect 95% CI: 0.3039–2.4466), with a significant partial mediation effect observed, though future studies controlling for dietary and other confounding variables are needed to confirm this relationship.

Conclusion

taVNS effectively alleviates depressive symptoms in MDD patients via dual complementary pathways: directly modulating neural activity in the right SFG to regulate depression-related brain function, and indirectly maintaining gut microbiota homeostasis by enriching beneficial taxa such as Faecalibacterium prausnitzii. These findings provide novel mechanistic insights into the brain-gut interaction underlying the antidepressant effects of taVNS, laying a theoretical foundation for its clinical application in MDD management.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-08409-y.

Introduction

Major Depressive Disorder (MDD) is a highly prevalent psychiatric disorder characterized by persistent low mood, anhedonia, sleep disturbances, and cognitive impairments, affecting over 350 million individuals globally [1]. Despite the availability of various antidepressant medications, a substantial subset of patients exhibits inadequate responses or faces challenges such as delayed onset of action [2, 3]. Accumulating evidence highlights the critical role of gut microbiota dysbiosis in MDD pathogenesis via the microbiota-gut-brain axis (MGBA), which modulates neuroinflammation, neurotransmitter synthesis, and hypothalamic-pituitary-adrenal (HPA) axis activity [4–6]. The vagus nerve, a key component of the MGBA, serves as a bidirectional communication pathway between the gut and brain, thereby regulating emotional and cognitive processes [7, 8].

Transcutaneous auricular Vagus Nerve Stimulation (taVNS) is a non-invasive neuromodulation technique targeting the auricular branch of the vagus nerve, with demonstrated antidepressant effects by regulating central nervous system activity and autonomic functions [9–11]. Unlike invasive vagus nerve stimulation (iVNS), taVNS avoids surgical risks and offers cost-effectiveness, making it a viable alternative for MDD treatment [12, 13]. Previous studies have shown that taVNS alleviates depressive symptoms by modulating default mode network (DMN) connectivity and reducing aberrant neural activity in emotion-related brain regions [14, 15]. However, the mechanisms linking taVNS-induced gut microbiota alterations to brain function changes remain incompletely understood.

Currently, there is a paucity of research investigating the effects of taVNS on gut microbiota in MDD patients and whether its antidepressant mechanisms involve the regulation of brain-gut interactions. To address this knowledge gap, the present study aimed to: (1) evaluate the clinical efficacy of taVNS for MDD; (2) integrate rs-fMRI and gut microbiota data to clarify relationships among clinical outcomes, neuroimaging findings, and gut microbiota profiles; and (3) test the hypothesis that taVNS regulates the “vagus nerve-gut microbiota-emotional center” axis, thereby elucidating the brain-gut interaction mechanisms underlying its therapeutic effects.

Materials and methods

Study design

In this study, participants were required to complete assessments using the 17-item Hamilton Depression Rating Scale (HAMD-17) [16], the 14-item Hamilton Anxiety Rating Scale (HAMA-14), and the Gastrointestinal Symptom Rating Scale (GSRS) [17]. Individuals who met the established inclusion criteria proceeded to undergo resting-state functional magnetic resonance imaging (rs-fMRI) and gut microbiota data collection. Following this initial assessment, participants were randomly assigned to either the taVNS group or the sham taVNS group for an intervention period of 8 weeks.

Upon completion of the 8-week intervention, fMRI and gut microbiota data were re-collected, and participants were re-evaluated using the HAMD-17, HAMA-14, and GSRS scales. The gathered data were subsequently subjected to rigorous statistical analysis. The detailed workflow of the study is illustrated in Fig. 1.

Fig. 1.

Study design

Study population and recruitment

Outpatients diagnosed with MDD were recruited from Guang’anmen Hospital, China Academy of Chinese Medical Sciences, between May 2022 and March 2024. Diagnosis was confirmed using the Structured Clinical Interview for DSM-V criteria by two experienced psychiatrists, and depression severity was assessed via HAMD-17. We hereby confirm that the present study is a clinical trial and it has been officially registered with the Chinese Clinical Trial Registry (ChiCTR). The detailed registration information is as follows:

Inclusion criteria

(1) DSM-V-confirmed MDD; (2) age 18–70 years; (3) right-handed; (4) HAMD-17 score 8–24; (5) medication-naïve or antidepressant-free for ≥ 4 weeks.

Exclusion criteria

(1) History of alcohol/substance dependence; (2) comorbid psychiatric disorders; (3) MRI contraindications (e.g., pacemaker, metallic implants, claustrophobia); (4) pregnancy/lactation; (5) antibiotic use within 4 weeks prior to enrollment or during the study period; (6) incomplete clinical, fMRI or gut microbiota data (not fully collected or failed to meet quality standards).

A total of 95 patients were randomly assigned to two groups: 48 participants in the transcutaneous auricular vagus nerve stimulation (taVNS) group and 47 participants in the sham taVNS group. During the course of the study, four patients were lost to follow-up due to non-compliance, with two from each group. Additionally, four patients were classified as dropouts because of adverse reactions, consisting of three from the taVNS group and one from the sham taVNS group (two due to skin irritation at the stimulation site and two due to dizziness during MRI scans requesting withdrawal).

For the analysis, it was essential that all included patients had complete clinical, fMRI, and gut microbiota data. Consequently, four patients were excluded due to inadequate fMRI data (two from the taVNS group and two from the sham taVNS group), and three were excluded due to insufficient gut microbiota data (one from the taVNS group and two from the sham taVNS group). Ultimately, a total of 80 patients diagnosed with MDD were included in the final analysis, with 40 patients in each group. A flowchart illustrating the participant recruitment process is presented in Fig. 2.

Fig. 2.

Flowchart

All participants were all informed of the study’s protocol and provided voluntary consent to take part. The study protocol was approved by the Ethics Committee of Guang’anmen Hospital (Approval No. 2022-031-KY) and registered with the China Clinical Trials Registry (ChiCTR2000059591). All participants provided written informed consent.

Randomization and blinding

Eligible patients were randomized 1:1 using a computer-generated random number table with a block size of 4. Allocation sequences were sealed in opaque envelopes by an independent research assistant. Patients, outcome assessors, and statisticians remained blinded to treatment assignment throughout the study.

Intervention and control

The intervention protocol was based on a preliminary clinical trial [18], which utilized an electroacupuncture stimulator (Hwato, SDZ-II B, Suzhou, China) and special silica gel ear clips.

Participants were positioned either supine or seated for the procedure. Prior to treatment, the cymba conchae area of each participant’s ear was disinfected using a cotton ball soaked in 75% ethanol. The device’s two ear clips were then attached to designated areas corresponding to the vagus nerve auricular branch distribution, specifically the auricular point for the kidney (CO10) and the cavum conchae (auricular point for the heart, CO15), both of which point have a rich distribution of vagus nerve fibers [11]. The stimulation parameters were meticulously set as follows: alternating frequency waves, a pulse frequency of 4/20 Hz, a sparse wave frequency of 4 Hz, a dense wave frequency of 20 Hz, and a current intensity ranging from 3 to 8 mA (mean ± SD: 5.62 ± 1.38 mA). This intensity range was defined based on previous clinical trials [18] and participant tolerance. The output intensity was adjusted according to the participant’s tolerance, ensuring that the current sensation remained mild and did not elicit sharp pain. At the start of the intervention, the initial intensity was set at 3 mA for all participants. Participants could adjust the intensity by themselves until reaching a comfortable, tolerable level (maximum 8 mA). No participant experienced sharp pain or required discontinuation due to intensity-related issues.

For the sham taVNS group, stimulation was applied to a non-vagus nerve distribution area located at the superior concha of the mid-helix [19], with the same intensity range of 3–8 mA adjusted according to participant tolerance. All participants underwent the intervention for 30 min, twice daily, across five consecutive days each week, for a duration of 8 weeks. Participants were instructed to administer the stimulation once in the morning after waking and once in the evening before bedtime, with at least 8 h between sessions. Following the screening process, all patients received training on the application of either taVNS or sham taVNS. The illustration of taVNS procedure was described in eFigure 1 in Supplement materials.

All treatments were self-administered by the patients or assisted by family members in a home setting. Researchers provided comprehensive training on the operation of the device and the treatment procedures to ensure that each participant or their family member could independently and accurately utilize the device. Patients were instructed to maintain a daily treatment diary, documenting the exact treatment time, date, mood state, and any adverse events that were either directly related to or temporally associated with the treatment. Investigators reviewed these diaries biweekly through video calls on WeChat, an instant messaging application that facilitates text, emoji, image sharing, and voice and video communications, to promote adherence to the treatment protocol. The intensity adjustment process was also monitored during these biweekly reviews to ensure standardization across participants. Among the 80 participants who completed the study, all adhered to the prescribed stimulation schedule of 5 days/week for 8 weeks, as confirmed through diary reviews and biweekly video call assessments.

Clinical data outcome measurements

Primary outcome

HAMD-17 score reduction, with response defined as ≥ 50% reduction from baseline and remission as post-intervention HAMD-17 ≤ 7.

Secondary outcomes

Changes in HAMA-14 (anxiety) and GSRS (gastrointestinal symptoms) scores.

rs-fMRI data scan acquisition

All subjects underwent fMRI scans at Guang’anmen Hospital, China Academy of Chinese Medical Sciences and using a Magnetom Skyra 3.0 T scanner (Siemens, Erlangen, Germany). Before the scanning procedure, the subjects were instructed to remain awake and avoid active thinking. During the scanning process, the subjects were required to wear earplugs and noise-cancelling headphones, to use a hood to immobilize the head, and to lie flat on the examination bed. The scanning procedure involved a localizer scan, high-resolution three-dimensional T1-weighted imaging, and resting state BOLD-fMRI.

The scanning parameters were as follows: for three-dimensional T1-weighted imaging, time repetition/time echo = 2500/2.98 ms, flip angle = 7°, matrix = 64 × 64, field of view = 256 mm×256 mm, slice thickness = 1 mm, slice number = 48, slices = 192, scanning time = 6 min 3 s; for BOLD-fMRI, time repetition/time echo = 2000/30 ms, flip angle = 90°, matrix = 64 × 64, field of view = 240 mm×240 mm, slice number = 43, slice thickness/spacing = 3.0/1.0 mm, number of obtained volumes = 200, and scanning time = 6 min 40 s.

Gut microbiota data sampling

Participants were instructed to line the toilet bowl with a plastic bag to minimize contamination. Utilizing a sterile spoon, they collected 3–5 g of fecal matter from the central portion of the sample, ensuring that it had not come into contact with air or the toilet surface. This sample was then placed into a sterile collection tube, which was securely sealed.

To preserve the integrity of the samples and prevent repeated freeze-thaw cycles, they were stored at -80 °C. For transportation, the samples were maintained at low temperatures using dry ice and subsequently shipped to Shan Jun (Xi’an) Biopharmaceutical Technology Co, Ltd. for metagenomic sequencing. Fecal samples were collected at two distinct time points: prior to the intervention (week 0) and following the intervention (week 8).

Data processing

fMRI data preprocessing

DPABI (http://rfmri.org/DPABI) software, an SPM-based functional MRI pre-processing pipeline, was used for data pre-processing, according to the following process: The specific processes are as follows [20]: (1) DICOM raw data transfer to NIFTI format; (2) Removal of the first 10 time points of NIFTI data; (3) Slice timing; (4) Head motion correction, realignment of head motion and removal of patients with head movements > 2 mm in any direction and motor rotation > 2◦) (5) Normalization, the resulting aligned image time series for each subject were each co-registered with the corresponding 3D T1-weighted image and the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) tool was used to normalize the data for all subjects to Montreal Neurological Institute (MNI) space, which was performed using the MNI coordinate space with 3 mm × 3 mm × 3 mm; (6) Smoothening (a 6-mm full-width at half-maximum Gaussian kernel); (7)Remove linear tread, long-term physiological shifts, movement related noise remaining after realignment or instrumental instability may contribute to a systematic increase or decrease in the signal with time; (8) Regression of covariates, including brain white matter signal, cerebrospinal fluid signal, and head movement parameters); (9) Filtering (0.01–0.08 Hz) was applied to reduce the influence of low-frequency drift and physiologic high frequency respiratory and cardiac noise.

Amplitude of Low Frequency Fluctuation (ALFF) Analysis [21]: The calculation is performed prior to filtering. Initially, a 6 mm×6 mm×6 mm smoothing kernel is applied for spatial smoothing of the preprocessed images. Then, ALFF is calculated and analyzed using DPARSF with a band-pass filter of 0.01 to 0.08 Hz to remove noise generated by physiological signals such as respiration and heartbeat. A Fourier transform is conducted on the magnetic resonance signals within this frequency band. The square root of each power spectrum is calculated over frequency, and the corresponding square root and normalization of the frequency power spectrum are performed to obtain the ALFF. Subsequently, the ALFF values of individual voxels are compared to the whole-brain mean ALFF, followed by a Z-transformation to obtain the standardized zALFF values.

Fraction Amplitude of Low Frequency Fluctuation (fALFF) Analysis [22]: The calculation is conducted prior to filtering. A 6 mm×6 mm×6 mm smoothing kernel is applied for spatial smoothing of the preprocessed images. The fALFF is obtained by comparing the root mean square of the low-frequency power spectrum within 0.01–0.08 Hz with the root mean square of the full frequency range (0–0.25 Hz) power spectrum. Each voxel’s fALFF value is then divided by the whole-brain mean fALFF value and subjected to a Z-transformation to obtain the standardized zfALFF values.

Regional Homogeneity (ReHo) Analysis [23]: The calculation is performed prior to smoothing. Initially, the preprocessed images undergo band-pass filtering (0.01–0.08 Hz). Then, the Kendall’s coefficient of concordance is calculated for the time series of each voxel and its 26 neighboring voxels, resulting in a ReHo map for the subjects in MNI standard space. A 6 mm×6 mm×6 mm Gaussian smoothing kernel is applied for smoothing. The resulting ReHo map is then standardized and subjected to a Z-transformation to obtain the standardized zReHo map.

Degree Centrality (DC) Analysis [24]: The calculation is performed prior to smoothing. The preprocessed images are band-pass filtered (0.01–0.08 Hz). A 6 mm×6 mm×6 mm Gaussian smoothing kernel is applied to analyze the preprocessed images. The resulting DC map is then standardized and subjected to a Z-transformation to obtain the standardized zDC map.

Gut microbiota data preprocessing

A specified volume of fecal sample was utilized for microbial DNA extraction, employing the MagBeads Fast DNA Kit for Feces (MP Biomedicals, Shanghai, China) in accordance with the manufacturer’s protocol. The concentration of the extracted DNA was quantified using a NanoDrop 2000 spectrophotometer, followed by assessment of DNA integrity via 1% agarose gel electrophoresis, utilizing KAPA HyperPlus PCR-free reagents (Illumina, CA, USA).

For fragment preparation, 200 ng of nucleic acid was sheared using a Covaris system to achieve a primary fragment size of approximately 420 bp. Fragment selection was conducted to isolate the main band within the range of 380 to 420 bp. Paired-end libraries were constructed using 50 ng of the fragment-selected DNA. Quantification of the DNA was performed with a Qubit 2.0 fluorometer (Thermo Fisher, Massachusetts, USA), and read quality was assessed using the Agilent 2100 Bioanalyzer (Agilent, Palo Alto, USA).

All sample libraries were pooled to obtain a total DNA quantity of 300 ng in a volume of 48 µl. Subsequent to pooling, the samples underwent heat denaturation, single-strand circularization, enzymatic digestion, and purification prior to sequencing on the MGISEQ-T7 metagenomic sequencing platform. To preprocess the raw sequencing data, fastp (v0.23.0) was employed with parameters set to (length_required = 50, n_base_limit = 2) to eliminate low-quality reads and adapter sequences. Host reads were filtered out using Bowtie2 software (v2.3.5.1), referencing the human genome (version 38).

Metagenomic sequencing data were annotated using MetaPhlAn4 (v4.0.3) to analyze microbial community composition and derive relative abundance metrics. Functional profiling at the species level was conducted with HUMAnN3 (v3.6), utilizing a tiered search strategy that aligns against pan-genomes. This approach facilitated the quantification of gene families and pathway abundances, enabling rapid access to diverse species and functional compositions within the microbial community.

Statistical analyses

Clinical data analysis

Clinical data were analyzed using SPSS 26.0 statistical software (IBM Corporation, Somers, NY, USA). For clinical efficacy analysis, repeated measures ANOVA was employed to assess group × time interaction effects on HAMD-17, HAMA-14, and GSRS scores, with group (taVNS vs. sham taVNS) as the between-subject factor and time (pre- vs. post-intervention) as the within-subject factor. Post-hoc pairwise comparisons were performed when significant interactions were observed. For continuous variables that followed a normal distribution, independent sample t-tests were employed to compare differences between two groups, while paired t-tests were used for pre- and post-intervention comparisons, with results reported as mean±standard deviation. For variables that did not meet the criteria for normality, the Mann-Whitney U test was utilized to assess group differences, and the Wilcoxon signed-rank test was applied for within-group comparisons, with results expressed as interquartile ranges. Categorical data were evaluated using the chi-square test. All analyses were conducted using a two-tailed approach, with a significance level set at P < 0.05 to denote statistical significance.

fMRI data analysis

In this study, Image data processing was conducted using the DPARSF 7.0 toolkit. A repeated measures ANCOVA was employed to assess treatment effects using the interaction effect of group (taVNS and sham taVNS groupsl)×time (pre-intervention, post-intervention). Covariates included gender, age, education level, and head motion displacement. Multiple comparisons were corrected using Gaussian random fields (GRF), with significance thresholds set at P < 0.001 at the voxel level and P < 0.05 at the cluster level, indicating statistical significance. Post-hoc analyses were performed by extracting ALFF, fALFF, ReHo, and DC values from the four groups of differential brain regions. Further, Spearman correlation analysis was conducted to examine the relationship between differential brain regions and clinical scales, controlling for gender, age, education level, and head motion displacement, with P < 0.05 considered statistically significant.

Gut microbiota data analysis

Statistical analyses were conducted using SPSS Statistics 26.0 and R (v4.3.1). For gut microbiota indices, repeated measures ANOVA was employed to assess group × time interaction effects, with group (taVNS vs. sham taVNS) as the between-subject factor and time (pre- vs. post-intervention) as the within-subject factor. Post-hoc pairwise comparisons were performed when significant interactions were observed. Multiple comparison correction was performed using the Benjamini-Hochberg FDR method. Additionally, the Wilcoxon rank-sum test was used for inter-group comparisons of dominant bacterial species at the species level. For non-normally distributed data, the Mann-Whitney U test was used for comparisons between two groups, and the Wilcoxon signed-rank test was used for pre- and post-intervention comparisons, with results expressed as interquartile ranges. Categorical data were analyzed using the chi-square test. Covariates such as age, gender, and years of education were controlled. A two-tailed test was applied, with P < 0.05 indicating statistical significance.

Furthermore, to explore the gut-brain mechanism of taVNS modulation in MDD, Spearman correlation analysis and mediation effect analysis were employed to investigate relationships among clinical, neuroimaging and gut microbiota data. Mediation effects were tested using the bias-corrected non-parametric percentile Bootstrap method (n = 5000 samples). A Bootstrap 95% confidence interval (CI) not containing zero indicated a significant mediation effect, with P < 0.05 considered statistically significant.

Results

Clinical efficacy

There were no statistically significant differences between the two groups of patients in terms of gender, age, education level, BMI, and severity of illness (P > 0.05), indicating that the groups were comparable, as detailed in Table 1. Severity of illness was classified based on HAMD-17 scores: mild depression was defined as HAMD-17 score 8–17, and moderate depression was defined as HAMD-17 score 18–24, according to standard HAMD-17 severity grading criteria. All enrolled participants were first-episode, medication-naïve MDD patients who had never received antidepressant treatment, ECT, rTMS, or other neuromodulation treatments. The mean duration of the current depressive episode was 4–6 weeks, which also represented the total illness duration given the first-episode nature of the sample. No participants met criteria for treatment-resistant depression.

Table 1.

Comparison of demographic data between the two groups

Variable	Group		X2/t	P
	taVNS(n = 40)	Sham taVNS(n = 40)
Age(years)	33.62 ± 10.25	33.65 ± 8.57	-0.012b	0.900
Gender(male/female)	17/23	16/24	0.052a	0.820
Education(years)	14.65 ± 2.28	15.12 ± 2.34	-0.918b	0.943
BMI (kg/m2)	22.04 ± 3.79	22.54 ± 3.98	-0.577b	0.856
Severity of illness (mild/moderate)	10/30	14/26	0.952a	0.329

^aThe P-values of sex distribution between the two groups were obtained using the chi-square test

^bP-value from one-way analysis of variance tests

At baseline, there were no statistically significant differences between the two groups in terms of HAMD-17, HAMA-14, and GSRS scores (P > 0.05), indicating comparability. Within-group comparisons revealed that after the intervention, both groups exhibited statistically significant differences in HAMD-17, HAMA-14, and GSRS scores compared to pre-intervention levels (P < 0.05). Furthermore, post-intervention between-group comparisons showed statistically significant differences in HAMD-17, and HAMA-14 scores (P < 0.05), as shown in eTable 1 in Supplement materials. Repeated measures ANOVA revealed significant group × time interaction effects for HAMD-17 (F(1, 78) = 12.51, P = 0.0007) and HAMA-14 (F(1, 78) = 6.62, P = 0.012) scores, while the interaction for GSRS did not reach statistical significance (F(1, 78) = 2.40, P = 0.125).

After 8 weeks of treatment, the response rate of HAMD-17 in the taVNS group was 62.50%, and the remission rate was 35.00%. In the sham taVNS group, the response rate of HAMD-17 was 30.00%, and the remission rate was 2.50%. The statistical results indicated that the differences in response rates and remission rates between the two groups were statistically significant (P = 0.004, P = 0.001 < 0.05). The clinical efficacy of the taVNS group was superior to that of the sham taVNS group. As shown in Fig. 3 and eTable 2 in Supplement materials.

Fig. 3.

Comparison of response rates and clinical remission rates between the two groups (***P<0.001, ****P<0.0001)

Changes in brain functional activities

Interaction effects of ALFF and fALFF pre and post intervention in two groups

Pre and post intervention in the two groups, a significant group × time interaction effect was observed in the ALFF of the right occipital inferior gyrus (Occipital_Inf_R). Additionally, significant group × time interaction effects were found in the fALFF of the right inferior temporal gyrus (Temporal_Inf_R), the left calcarine gyrus (Calcarine_L), the right superior frontal gyrus (Frontal_Sup_R), and the right angular gyrus (Angular_R). As shown in Fig. 4 and eTable 3 in Supplement materials.

Fig. 4.

Differences in ALFF and fALFF between the two groups pre and post intervention. Note: The color bar represents the F-value from the group × time interaction analysis, with yellow indicating higher F-values and red indicating lower F-values

Post-hoc analysis of ALFF and fALFF between pre and post intervention

Before the intervention, there was no statistically significant difference in the ALFF of the right Occipital Inferior Gyrus (Occipital_Inf_R) between the two groups (t = 1.331, P = 0.01872). After the intervention, comparison between the two groups revealed that the ALFF of the right Occipital Inferior Gyrus (Occipital_Inf_R) was significantly reduced in the taVNS group compared to the sham taVNS group (t = 3.344, P = 0.0013). Within the taVNS group, the ALFF of the right Occipital Inferior Gyrus (Occipital_Inf_R) was significantly lower after the intervention than before (t = 0.8744, P = 0.038), indicating a statistically significant difference.

Before the intervention, there were no statistically significant differences in the fALFF of the right Inferior Temporal Gyrus (Temporal_Inf_R) and the left Calcarine Gyrus (Calcarine_L) between the two groups (t = 2.213, P = 0.298; t = 3.156, P = 0.23). After the intervention, the fALFF of the right Inferior Temporal Gyrus (Temporal_Inf_R) and the left Calcarine Gyrus (Calcarine_L) was significantly reduced in the taVNS group compared to the sham taVNS group (t = 3.562, P = 0.0006; t = 2.282, P = 0.0252). Within the taVNS group, the fALFF of the right Inferior Temporal Gyrus (Temporal_Inf_R) and the left Calcarine Gyrus (Calcarine_L) was significantly lower after the intervention than before (t = 2.469, P = 0.0180; t = 1.620, P = 0.0133), demonstrating statistically significant differences.

Before the intervention, no statistically significant differences were found in the fALFF of the right Superior Frontal Gyrus (Frontal_Sup_R) and the right Angular Gyrus (Angular_R) between the two groups (t = 3.550, P = 0.23; t = 3.697, P = 0.23). After the intervention, the fALFF of the right Superior Frontal Gyrus (Frontal_Sup_R) and the right Angular Gyrus (Angular_R) was significantly increased in the taVNS group compared to the sham taVNS group (t = 1.666, P = 0.0252; t = 2.702, P = 0.0085). Within the taVNS group, the fALFF of the right Superior Frontal Gyrus (Frontal_Sup_R) and the right Angular Gyrus (Angular_R) was significantly higher after the intervention than before (t = 2.199, P = 0.0033; t = 2.199, P = 0.0033), indicating statistically significant differences. Show in Fig. 5.

Fig. 5.

Comparison of Post Hoc analysis of ALFF and fALFF between the two groups pre and post intervention (*P < 0.05, **P < 0.01)

Interaction effects of ReHo and DC between pre and post intervention

Pre and post interventions, significant group × time interaction effects in ReHo were observed in the right Occipital Inferior Gyrus (Occipital_Inf_R), right Insula (Insula_R), and left Insula (Insula_L). Similarly, significant group × time interaction effects in DC were found in the right Caudate (Caudate_R) and left Anterior Cingulate Cortex (ACC_L). Refer to eTable 4 in Supplement materials and Fig. 6 for details.

Fig. 6.

Differences in ReHo and DC between the two groups pre and post intervention. Note: The color bar represents the F-value from the group × time interaction analysis, with yellow indicating higher F-values and red indicating lower F-values

Post-hoc analysis of ReHo and DC between pre and post intervention

Before the intervention, there were no statistically significant differences in the ReHo values between the taVNS group and the sham taVNS group in the right Occipital Inferior Gyrus (Occipital_Inf_R), right Insula (Insula_R), and left Insula (Insula_L) (t = 1.778, P = 0.0794; t = 1.807, P = 0.0746; t = 2.752, P = 0.074, respectively). After the intervention, the ReHo values in the taVNS group were significantly lower compared to the sham taVNS group in the right Occipital Inferior Gyrus (Occipital_Inf_R), right Insula (Insula_R), and left Insula (Insula_L) (t = 3.516, P = 0.0007; t = 3.475, P = 0.0008; t = 2.374, P = 0.0201, respectively). Within the taVNS group, the ReHo values in the right Occipital Inferior Gyrus (Occipital_Inf_R), right Insula (Insula_R), and left Insula (Insula_L) were significantly reduced after the intervention compared to baseline (t = 2.661, P = 0.0133; t = 3.799, P = 0.0005; t = 3.368, P = 0.0017, respectively). The differences were statistically significant.

Before the intervention, there were no statistically significant differences in the DC values between the taVNS group and the sham taVNS group in the right Caudate Nucleus (Caudate_R) and left Anterior Cingulate Cortex (ACC_L) (t = 1.492, P = 0.1373; t = 0.807, P = 0.4219, respectively). After the intervention, the DC values in the taVNS group were significantly higher compared to the sham taVNS group in the right Caudate Nucleus (Caudate_R) and left Anterior Cingulate Cortex (ACC_L) (t = 3.571, P = 0.0157; t = 2.003, P = 0.0048, respectively). Within the taVNS group, the DC values in the right Caudate Nucleus (Caudate_R) and left Anterior Cingulate Cortex (ACC_L) were significantly increased after the intervention compared to baseline (t = 1.517, P = 0.0137; t = 0.8117, P = 0.0421, respectively). The differences were statistically significant. Show in Fig. 7.

Fig. 7.

Comparison of Post Hoc analysis of ReHo and DC between the two groups pre and post intervention (*P < 0.05, **P < 0.01)

Gut microbiota data analysis

Alpha and Beta diversity analysis

In this study, we analyzed the Alpha diversity indices (Shannon, Simpson, and InvSimpson) of gut microbiota at the species level. Comparisons of Alpha diversity indices pre- and post-intervention between the taVNS group and the sham taVNS group did not reveal any statistically significant differences (P > 0.05), as illustrated in eFigure 2 and eTable 5 in the Supplementary Materials. Repeated measures ANOVA further confirmed no significant group × time interaction effects for Shannon (F(1, 78) = 0.01, P = 0.921), Simpson (F(1, 78) = 0.95, P = 0.333), or InvSimpson (F(1, 78) = 0.30, P = 0.587) indices.

Furthermore, we examined Beta diversity at the species level using Principal Coordinates Analysis (PCoA). No significant differences in Beta diversity were observed between the two groups prior to intervention or between the sham taVNS group before and after the intervention (P = 0.323 and P = 0.584, respectively, both > 0.05). However, post-intervention comparisons between the two groups, as well as comparisons within the taVNS group pre- and post-intervention, indicated an increasing trend in sample distances. This change suggests significant alterations in gut microbiota structure, achieving statistical significance (P = 0.041 and P = 0.04, both < 0.05), as depicted in eFigure 3 in the Supplementary Materials.

Differential species analysis of dominant microbes

We conducted an inter-sample comparative analysis of dominant bacterial taxa between groups with significant Beta diversity differences. At the species level, we first used stacked bar plots to display the composition of the top 20 dominant bacterial species, as shown in Fig. 8(a-b).

Fig. 8.

Stacked bar chart of dominant species composition at the species level: (a) taVNS group pre and post intervention; (b) taVNS group and sham taVNS group after intervention. Note: The x-axis represents different groups, and the y-axis represents the relative abundance percentage of each species. Species not in the top 20 are all merged into ‘Others’

The Wilcoxon rank-sum test was applied to analyze inter-group differences in dominant bacterial species at the species level. The results demonstrated that, compared to the pre-intervention period, the post-intervention taVNS group showed a decrease in the abundance of Anaerobutyricum hallii and an increase in the abundance of Faecalibacterium prausnitzii. Post-intervention, compared to the sham taVNS group, the taVNS group exhibited a decrease in the abundance of Phocaeicola vulgatus and an increase in the abundance of Ruminococcus bromii and Fusicatenibacter saccharivorans. Multiple comparison correction was performed using the Benjamini-Hochberg FDR method for the top 20 dominant bacterial species. After FDR correction, Anaerobutyricum hallii (FDR-adjusted P = 0.011) and Faecalibacterium prausnitzii (FDR-adjusted P = 0.049) remained significant in the taVNS pre-vs-post comparison. These findings are detailed in Fig. 9 (a-b).

Fig. 9.

Boxplot of differences in dominant species between groups: (a) taVNS group pre and post intervention; (b) taVNS group and sham taVNS group after intervention. Note: Indicates a significant difference. (*P<0.05; **P<0.01; ***P<0.001)

Species LEfSe biomarker analysis

Biomarkers exhibit high sensitivity and specificity, reflecting the severity of disease and providing insights into pathophysiology, thereby facilitating the optimization of diagnostic and therapeutic approaches. Linear discriminant analysis (LDA) Effect Size (LEfSe) analysis employs LDA for dimensionality reduction of data and evaluates interspecies differences to determine their contribution to the overall impact (LDA score ≥ 2). This method can identify and interpret high-dimensional biomarkers, uncover genomic characteristics, pinpoint microbial communities with statistically significant differences between groups, and assess the influence of the relative abundance of each species on the observed differences.

From the LDA discriminant bar chart, this study identified the following microbial biomarkers at the species level: prior to intervention, the biomarker in the taVNS group was Anaerobutyricum_hallii, while after intervention, the biomarker was Faecalibacterium_prausnitzii. In the sham taVNS group, the biomarker after intervention was Phocaeicola_vulgatus, whereas in the post-intervention taVNS group, the biomarker was Clostridium leptum (Fig. 10).

Fig. 10.

LDA discriminant bar chart from LefSe analysis showing gut microbiota biomarkers with significant differences between groups. Note: The color represents the corresponding group, and the length of the bar represents the magnitude of the contribution of the differential species (i.e., the LDA Score)

In summary, through various analyses of intergroup differences in dominant species and LEfSe biomarker identification, this study clarified the microbial species with significant differences at the species level. The results demonstrate that the taVNS group intervention primarily impacted the relative abundance of Anaerobutyricum_hallii and Faecalibacterium_prausnitzii. These two species were identified as the differential microbial species of this study. The significant changes in the abundance of these microbial communities likely influence the overall function of taVNS treatment in patients with MDD.

Correlation analysis

The correlation analysis results demonstrated a significant relationship between clinical data and brain imaging metrics within the taVNS group. Specifically, changes in fALFF in the right Superior Frontal Gyrus (Frontal_Sup_R) were positively correlated with improvements in HAMD-17 scores (r = 0.362, P = 0.022). Conversely, this change in fALFF was negatively correlated with improvements in GSRS scores (r = -0.382, P = 0.015).

Additionally, the analysis revealed correlations between clinical data, brain imaging, and gut microbiota. Notably, the improvement in HAMD-17 scores was negatively correlated with changes in the relative abundance of Faecalibacterium prausnitzii (r = -0.566, P < 0.01) and positively correlated with changes in the relative abundance of Anaerobutyricum hallii (r = 0.406, P < 0.01). Furthermore, the change in fALFF values in the right Superior Frontal Gyrus (Frontal_Sup_R) was also negatively correlated with changes in the relative abundance of Faecalibacterium prausnitzii (r = -0.348, P < 0.05). These findings indicate complex interrelationships among clinical outcomes, brain activity, and gut microbiota composition. As illustrated in eFigure 4 in the Supplementary Materials.

To further elucidate the mechanisms underlying the antidepressant effect of taVNS and the gut-brain interaction, a mediation effect analysis was conducted. The fALFF value of the right Superior Frontal Gyrus (Frontal_Sup_R) was used as the independent variable (X), the relative abundance of Faecalibacterium_prausnitzii as the mediator (M), and the improvement in depressive symptoms, represented by the HAMD-17 score, as the outcome variable (Y).

The analysis demonstrated that the change in the fALFF value of the right Superior Frontal Gyrus (Frontal_Sup_R) had a significant positive effect on the improvement in depressive symptoms (effect size = 1.6096, P = 0.0444). In contrast, the change in the relative abundance of Faecalibacterium_prausnitzii had a significant negative effect on the improvement in depressive symptoms (effect size = -0.3008, P < 0.0001). Furthermore, the change in the fALFF value of the right Superior Frontal Gyrus (Frontal_Sup_R) significantly negatively influenced the change in the relative abundance of Faecalibacterium_prausnitzii (effect size = -4.2741, P = 0.0222), as shown in Fig. 11.

Fig. 11.

Mediation effect analysis results. Note: Indicates a significant correlation. (*P < 0.05;****P < 0.0001)

Bootstrap correction further indicated that the indirect effect through the change in the relative abundance of Faecalibacterium_prausnitzii had an effect size of 1.2857, with a standard error of 0.5518 and a 95% confidence interval (0.3039, 2.4466) that did not include 0, suggesting a significant indirect effect. Meanwhile, the 95% confidence interval for the direct effect (-0.2778, 3.1771) included 0, indicating that the mediation effect was a partial mediation effect.

Discussion

Clinical efficacy of taVNS for MDD

MDD is a common emotional disorder marked by persistent low mood, loss of interest, and anhedonia [25]. In 2005, the FDA approved (iVNS as an adjunctive treatment for Treatment-Resistant Depression (TRD) [26]. However, the need for surgical implantation, along with its associated risks and costs, has limited its widespread use. Our research team has identified the auricular concha, specifically the visceral area of the ear acupuncutre theory, as the primary site of the Auricular Branch of the Vagus Nerve (ABVN). Inspired by iVNS and traditional Chinese medicine, we developed taVNS, a non-invasive method aimed at treating brain disorders, including MDD and insomnia, with similar efficacy comparable to iVNS [10, 18].

In this study, we observed significant improvements in HAMD-17, HAMA-14, and GSRS scores in both the taVNS and sham taVNS groups after 8-weeks’intervention. The taVNS group exhibited a response rate of 62.50% and a remission rate of 35.00%, significantly higher than the sham taVNS group with a response rate of 30.00% and a remission rate of only 2.50%. These results indicate that taVNS offers superior clinical efficacy compared to sham treatment, suggesting the observed effects in the sham group likely stem from placebo responses [27]. The placebo responses observed in the sham group may be attributed to several factors: (1) patient expectations and beliefs regarding the therapeutic potential of the intervention; (2) the therapeutic ritual of receiving regular treatment and interactions with healthcare providers; (3) natural disease fluctuations and regression to the mean; and (4) nonspecific effects of attention and care provided during the study. Previous meta-analyses have shown that approximately 80% of the response to antidepressant medication can be duplicated by placebo [28], with large placebo effect sizes observed in depression trials [29], which is consistent with the magnitude of improvements observed in our sham group (30% response rate).

Effects of taVNS on brain activity MDD

Our findings demonstrated a significant reduction in spontaneous neuronal activity in the right Inferior Temporal Gyrus following taVNS treatment. The temporal lobe, while primarily recognized for its role in auditory processing, is also crucial for a variety of higher cognitive functions, including learning, memory consolidation, and the regulation of emotional behavior [30–32]. When the temporal lobe is functioning abnormally, it can contribute to a range of psychological issues, including anxiety, depression, and insomnia, which are commonly interconnected and can exacerbate one another [33–34]. This region is part of the Default Mode Network (DMN), which plays a critical role in emotional regulation and cognitive processing [35]. Abnormal function within the DMN is associated with various psychological disorders, making it a crucial target for therapeutic intervention. Prior studies have shown that taVNS can effectively reduce functional connectivity within the DMN, thereby alleviating excessive neuronal excitation and normalizing heightened arousal levels [36]. By modulating activity in these critical brain regions, taVNS may effectively mitigate anxiety and depression, which are often interrelated with chronic insomnia [37–39].

The Superior Frontal Gyrus (Frontal_Sup_R) is integral to advanced cognitive functions, including working memory, abstract reasoning, judgment, and emotional regulation [40]. Research by Huang et al. [41] indicated that patients with MDD exhibit decreased fALFF in the right Superior Frontal Gyrus. Consistent with these findings, our study demonstrated an enhancement in fALFF within the Superior Frontal Gyrus following taVNS treatment. Notably, this increase in fALFF was positively correlated with HAMD-17 scores and negatively correlated with GSRS scores. This suggests that reduced activity in the Superior Frontal Gyrus may contribute to deficits in emotional regulation observed in MDD patients. Additionally, research conducted by Fu-Jian Chen et al. [42] revealed dysfunction and diminished spontaneous activity in the Superior Frontal Gyrus among MDD patients. Following repetitive transcranial magnetic stimulation (rTMS) treatment, functional activity in this region improved, coinciding with a reduction in depressive symptoms, which aligns with our findings. Moreover, the functional activity of the Superior Frontal Gyrus has been implicated in the regulation of gastrointestinal function. Xiaoya Fu et al. [43] identified a significant correlation between gastrointestinal motility symptom scores and activity in the Superior Frontal Gyrus among patients experiencing gastrointestinal symptoms.

In conclusion, this study suggests that taVNS ameliorates depressive symptoms in MDD patients by modulating brain activity in regions associated with the DMN, specifically the right Inferior Temporal Gyrus and the Superior Frontal Gyrus. Although improvements in gastrointestinal symptom scores were observed in both groups, this study was not designed to evaluate the efficacy of taVNS for insomnia or gastrointestinal disorders. Future studies with appropriate primary endpoints are warranted to investigate these potential therapeutic effects. These findings underscore the potential of taVNS as a therapeutic intervention targeting both emotional and physiological aspects of MDD.

Influence of taVNS on gut microbiota diversity and its relationship with depressive symptoms in MDD

Alpha diversity reflects the richness and abundance of species within a single sample, without facilitating inter-sample comparisons. In contrast, Beta diversity focuses on analyzing differences in species composition among multiple samples. In this study, we observed no significant differences in Alpha diversity indices at the species level. However, the Beta diversity analysis indicated an increasing trend in inter-sample distances following the intervention within the taVNS group, as well as between the taVNS and sham taVNS groups post-intervention, suggesting an enhancement in microbial diversity. Previous research by Schaub AC et al. [44, 45] has indicated that depressive symptoms can lead to a reduction in gut microbiota diversity. This decrease in diversity may further exacerbate negative emotional states, likely due to a decline in specific beneficial bacteria. Such dysbiosis in the gut microbiota is believed to contribute to the onset and progression of MDD. Therefore, our findings on Beta diversity could imply that taVNS may play a role in enhancing gut microbiota diversity, which might have positive implications for emotional well-being in individuals with depression.

Through differential analysis, we identified Anaerostipes and Faecalibacterium prausnitzii as key biomarkers in the taVNS group. The increase in Faecalibacterium prausnitzii abundance and decrease in Anaerostipes were linked to improvements in HAMD-17 scores, suggesting that taVNS may promote a healthier gut microbiota balance, potentially mitigating the gut dysbiosis associated with MDD. Additionally, alterations in the abundance of these bacteria were correlated with changes in fALFF in the right Superior Frontal Gyrus, highlighting the role of gut microbiota in mediating the effects of taVNS. Anaerostipes is a genus of anaerobic bacteria that plays a crucial role in fatty acid metabolism, particularly in the production of butyrate. Increased butyrate production is generally associated with enhanced gut epithelial function and improved immunological characteristics [46]. Conversely, Faecalibacterium prausnitzii is recognized as a beneficial bacterium that contributes to gut health and the stability of the microbial community. It accomplishes this by fermenting dietary fibers and generating short-chain fatty acids, which exert anti-inflammatory and immunoregulatory effects, including the reduction of intestinal pro-inflammatory cytokine levels [47]. Research has indicated that patients with MDD typically present lower relative abundances of Faecalibacterium prausnitzii, a bacterium significantly associated with gut inflammation [48]. This microorganism is recognized for its beneficial effects in the context of MDD, as its reduced levels are linked to an imbalance in gut microbiota. Such an imbalance may disrupt neurotransmitter signaling and inflammatory pathways, ultimately affecting the mood and mental state of individuals with MDD.

Correlation analysis revealed that changes in the relative abundance of Anaerostipes were positively correlated with reductions in HAMD-17 scores (i.e., greater increases in Anaerostipes abundance were associated with greater decreases in HAMD-17 scores), while changes in the abundance of Faecalibacterium prausnitzii showed a negative correlation with HAMD-17 scores (i.e., greater increases in Faecalibacterium prausnitzii abundance were associated with smaller reductions in HAMD-17 scores). Additionally, alterations in the fALFF in the right Superior Frontal Gyrus were negatively correlated with changes in the relative abundance of Faecalibacterium prausnitzii. As a beneficial gut bacterium, Faecalibacterium prausnitzii plays a crucial role in regulating depressive symptoms by maintaining gut microbiota balance and exerting anti-inflammatory effects. The prefrontal cortex, including the right Superior Frontal Gyrus, is closely linked to emotional regulation and is involved in higher cognitive processes such as decision-making, attention control, and emotional expression. It is particularly critical for suppressing negative emotions and modulating stress responses [48]. Our analysis suggests that changes in the abundance of Faecalibacterium prausnitzii may partially mediate the effects of taVNS treatment, indicating that gut microbiota alterations play a significant role in the therapeutic process for MDD.

Limitations

Several limitations of the present study should be noted. First, this is a single-center trial with a small final sample size, which may limit the generalizability of the conclusions and reduce the statistical power of subgroup analyses. Second, the study only set an 8-week intervention without long-term follow-up, so the durability of taVNS antidepressant effects and the persistence of related brain function and gut microbiota alterations remain unclear. Third, this study only explored brain-gut interactions via correlation and mediation analyses without targeted interventional verification, failing to confirm the causal relationship; besides, single metagenomic analysis is insufficient to fully reveal the in-depth molecular mechanisms. Fourth, the home-based intervention design may limit complete assurance of protocol adherence despite monitoring measures. Fifth, formal blinding assessment was not conducted, which should be addressed in future studies. Sixth, individual stimulation intensities were not systematically recorded for correlation analysis with outcomes.

Conclusion

This study elucidates the brain-gut interaction mechanisms in taVNS alleviating symptoms of MDD. The clinical study showed taVNS significantly relieved the symptom of MDD, and the functional MRI revealed the modulation of the DMN, particularly in the right Inferior Temporal Gyrus and the Superior Frontal Gyrus. Furthermore, the gut microbiota analysis highlighted the significant changes in the relative abundances of Faecalibacterium prausnitzii and Anaerostipes, suggesting that taVNS influences in gut microbiota composition. The positive correlation between changes in Anaerostipes abundance and HAMD-17 improvements, coupled with the negative correlation with Faecalibacterium prausnitzii, reflects a complex interplay that mediates the therapeutic effects of taVNS. We propose that taVNS produces the antidepressant efficacy through two pathways: first, by directly modulating emotional regulation brain regions, and second, by indirectly influencing gut microbiota dynamics that affect mood. These insights underscore the potential of taVNS as a novel non-invasive intervention for MDD and provide a foundation for future research that integrates brain-gut interactions to enhance treatment efficacy.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Jiliang Fang conceived and designed the experiments and revised the manuscript. Yue Ma collected data, analyzed data, and wrote manuscripts. Jifei Sun collected, analyzed data and revised the manuscript. Jiudong Cao and Lei Zhang performed fMRI and collecting data. Feng Zhu, Xiong Yu, Lichang Yang involved in the collection of fecal samples and the analysis of gut microbiota data.

Funding

This research was supported by the National Natural Science Foundation of China (82474663) and High Level Chinese Medical Hospital Promotion Project (No. HLCMHPP2023072, HLCMHPP2023094).

Data availability

Research data for this experiment are available from the first author on request.

Declarations

Ethical approval

The experimental protocol was approved by the Ethics Committee of Guang’anmen Hospital, China Academy of Chinese Medical Sciences (No. 2022-031-KY). This trial was registered with the Chinese Clinical Trial Registry (No. ChiCTR2200059591) on 4 May 2022. All patients provided written informed consent prior to study enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.