South Medical University’s Experimental Animal Center, Guangzhou, China, supplied the animals (Certificate No. SCXK [Guangdong] 2021-0041). In the course of the research, twenty-eight male 4-week-old C57BL/6 mice (20 g) were housed in a controlled environment (24 $\pm$ 2 °C, 50–60% humidity, 12:12 hours light/dark cycle). During this period, good indoor ventilation was ensured. After three days of quarantine, all mice were kept in the new environment for at least one week before use.

After baseline screening (excluding the mice with less than 20% decrease in sucrose preference after the CUMS), 28 mice were randomly divided into the normal control group (n = 8), and CUMS model group (n = 20). The normal control group did not have any intervention. The model group underwent CUMS model establishment for 28 days. The successful establishment of our CUMS model was determined by significant differences in depressive-like behaviors acquired on an experimental day 28 when compared with those of the normal control group. After confirming the modeling was successful, they were randomly divided from the CUMS model group into CUMS model group (n = 8) and the EA intervention group (n = 8). EA intervention group underwent EA intervention for 14 days after 28 days of CUMS modeling. CUMS group also grasped fixed but did not do EA intervention.

According to the method of Zhang [5], model group animals were housed individually and repeatedly exposed to the following CUMS conditions: (1) water deprivation for 24 hours; (2) food deprivation for 24 hours; (3) immobilization for 12 hours; (4) horizontal shaking for 5 minutes; (5) strobe for 12 hours; (6) day and night reversal. One stressor was applied in random order each day, and the whole stress process lasted 28 days.

CUMS modeling was followed by a 14-day intervention. EA was performed on mice in the EA group at GV20 (the center of the parietal bone) and GV29 (the middle of both brow bones) [3]. The mice were lightly immobilized in a clear acrylic mouse holder (number of version: SMT-HC-GDS, Smart (Guangzhou) Biotechnology Co Ltd, Guangzhou, Guangdong, China) and disposable acupuncture needles (lot number: 210802A, 0.22 mm $\times$ 5 mm, SUZHOU ACUPUNCURE & MOXIBUSTION APPLIANCE CO LTE, Suzhou, Jiangsu, China) were inserted horizontally into the two points at a depth of 2–3 mm. After insertion, the needles were connected to the electrodes and a sparse wave electrical stimulation (current of 1 mA, frequency of 2 Hz) was performed, when the mice’s heads trembled slightly. The entire EA treatment was administered for 30 minutes at a time, once a day. During the EA treatment, mice in CUMS groups were also lightly immobilized in the clear acrylic mouse holder in the same way as those in EA.

Upon completion of 14 days of intervention (at day 42), the Sucrose Preference Test (SPT) and Open Field Test (OFT) were administered.

All statistical analyses were performed using SPSS 22.0 (IBM Corp., Chicago, IL, USA) and GraphPad Prism 9 software (GraphPad, San Diego, CA, USA). All data were presented as mean $\pm$ SD both complied with normal distribution. Differences between two groups were assessed by independent samples t-test, and differences between groups were assessed by multivariate ANOVA followed by LSD post hoc tests. Statistical significance was given by p 0.05, or LDA scores $>$2.

After 28 days of CUMS modeling. Firstly, excluding the modeling failure, In the remaining 18 mice, approximately 80% of them successfully modeled CUMS-induced depression based on SPT, weight, and OFT evaluations (Supplementary Fig. 1). They were then divided into CUMS and EA groups with 8 mice in each group. As the intervention progressed, CUMS models remained stable, with significantly different behavioral results from the control group. The EA group was accepted 14 days of EA intervention. Neither the model establishment nor the intervention period resulted in any mouse deaths. Furthermore, the EA group showed positive results compared with the CUMS group, indicating that the treatment of EA improved mice’s depression-like behavior (weight, NC: 26.9700 $\pm$ 1.66, CUMS: 21.64 $\pm$ 0.91, EA: 23.18 $\pm$ 0.82, p = 0.0001; SPT, NC: 0.91 $\pm$ 0.024, p = 0.0001, CUMS: 0.72 $\pm$ 0.059, EA: 0.76 $\pm$ 0.021, OFT-total distance traveled, NC: 1686.63 $\pm$ 234.04, CUMS: 1285.69 $\pm$ 303.86, EA: 1674.47 $\pm$ 375.81, p = 0.027; OFT-time in the center area, NC: 50.83 $\pm$ 22.98, CUMS: 7.20 $\pm$ 5.11, EA: 20.19 $\pm$ 12.74, p = 0.0004) (Fig. 1).

A total of 872,581 single-end reads were obtained by sequencing; the DADA2 plug-in in the QIIME2 process was used for noise reduction of the sequenced reads, and the noise reduction steps included filtering noise and correcting sequence errors, removing chimeras and single sequences, and de-duplicating sequences to obtain high-resolution ASVs for subsequent analysis; a total of 4,85,522 sequences were generated from 18 samples after DADA2 noise reduction, with 22,615–33,287 sequences per sample, and 4239 ASVs were obtained.

The variation of total species and core species with sample size was described using PAN/Core species analysis. In microbial diversity studies, the species abundance and the number of core species in the environment are widely used to determine whether the sequencing sample size is sufficient. The results of PAN/Core species analysis showed (Fig. 2A) that the number of species under the genus classification stabilized and the curve leveled off as the sample size increased. The number of species of gut microorganisms no longer increased with the increase in sample size, indicating that the sample size was sufficient.

Alpha community diversity analysis and beta community diversity analysis are commonly used indicators to analyze species diversity. The results of the Alpha diversity index to test the differences between groups proved that the Shannon index and shannoneven index (ASV level) were lower in the NC groups (Fig. 2C,D) than in the CUMS group (p = 0.036) (p = 0.022). There was also a difference in the Simpson, Chao and Ace index in the NC group compared to the CUMS and EA groups, but it was not statistically significant (Supplementary Fig. 2). The results of the beta-diversity result to test the differences between groups proved that the PCA analysis compared the differences in microbial communities between the NC, CUMS, and EA groups (Fig. 2E). The PCA analysis of results showed significant differences between the EA and CUMS groups (p = 0.01, R = 0.2222). The above results indicated that the composition of intestinal flora was different between NC, CUMS, and EA groups, and the main composition of the intestinal flora of mice after CUMS intervention was more different from that of the NC group, while the main composition of the intestinal flora of mice after receiving EA intervention was more similar to that of NC group (Fig. 2E).

A hierarchically clustered heatmap was further constructed to evaluate the relationships among the samples at the genus level based on the top 30 most abundant microbial communities and the abundance was distinguished by the shade of color. The result showed that the 24 samples could be separated into three groups (Fig. 2F). The most abundant genera were Dubosiella (20.77%) in NC, Lactobacillus (17.42%) in CUMS, and Lachnospiraceae (17.07%) in EA, respectively. In comparison to CUMS group, the bacterial taxonomic groups of Lactobacillus, Lachnospiraceae_NK4A136_group, Bacillus, Desulfovibrio, Candidatus_Saccharimonas, and Staphylococcus, were detected to have higher similarity in colony composition at the genus level in NC and EA group. The heatmap analyses (Fig. 2F) agree with the results of the PCA (Fig. 2E), with Both analyses showing a high degree of similarity between NC and EA mouse intestinal flora sample characteristics and microbial communities.

Analyzed at the genus level (Fig. 3A), the highest relative abundance in the three groups was found for Lactobacillus. The proportion of Lactobacillus in the NC group was 34.45%, 33.13% in the EA group, and 19 in the CUMS group. The relative abundance of Lactobacillus in the CUMS group was significantly lower than in the EA group, while other genera that showed more significant differences in the multi-group comparative analysis of the three groups, such as Staphylococcus, were not found in the NC and EA groups and 2.20% in the CUMS group. Ruminococcaceae was 0.20% in the NC group, 2.11% in the EA group, and 2.11% in the CUMS group. The proportion of Tyzzerella was 8.99% in the NC group, 6.33% in the EA group, and 13.23% in the CUMS group.

The results of the comparative analysis of multiple groups (Fig. 3B) showed that. Lactobacillus had the highest abundance levels and the differences between groups were statistically significant (p = 0.013). Also, there was a statistically significant difference between groups for Staphylococcus and Tyzzerella (The abundance of Staphylococcus in the EA groups and Tyzzerella in the CUMS and EA groups were much shorter than other groups and could not be shown in the histogram). And norank_f__Ruminococcaceae was seen to have a more pronounced trend, but the difference between groups was not statistically significant (p = 0.059).

LEfSe can be used to find the specific characteristics that best explain the differences between groups in both groups and the extent to which these characteristics influence the differences between groups. LDA discriminant histograms counted the microbial groups that played a significant role in many groups. The LDA discriminant analysis shows that species with LDA $>$2 have a greater effect on intergroup variation (Fig. 3D). The LDA of the three species with significant differences in multiple comparisons at the genus level were Lactobacillus LDA = 4.90, Staphylococcus LDA = 4.03, Ruminococcaceae LDA = 3.98, and Tyzzerella LDA = 4.73.

CUMS had been chosen as the animal model in this study. And certain kinds of behavioral testing were involved to assess the success of modeling and intervention effect. Anhedonia and decreased locomotor activity are two common depressive behaviors shown in the CUMS model, which is widely used as an animal model for depression. Depression-related anorexia nervosa is characterized by a lack of enjoyment of food in humans and animals. The reduction in food intake may then lead to weight loss.

Sucrose preference tests are classic tests for detecting depression in rodents, and a reduced preference for sucrose is a key indicator of depression. In addition, OFT measures motor activity, and a decrease in motor activity indicates anxiety-like behaviors associated with depression [18]. Therefore, we used SPT, weighing, and OFT to evaluate depressive behaviors in CUMS mice in the present study.

After modeling, CUMS mice exhibited significant depression-like symptoms such as reduced preference for sugar water decreased spontaneous motor function and weight loss. These results are consistent with the findings in previous reports and support the success of the modeling [19].

EA was chosen as the treatment in this study. EA is a commonly recommended treatment for depression [20]. Additionally, the 2014 “Evidence-based Guidelines of Clinical Practice with Acupuncture and Moxibustion: Depression” also recommended EA treatment and auricular acupuncture as treatments for depression [21]. And the behavioral tests described above were used to assess the effect of the intervention. Researchers have reported that EA at GV20 and GV29 acupoints alleviate depression-related symptoms in chronic unpredictable mild stress-induced rats [22].

This finding is supported by our findings, which show that EA enhances sucrose uptake in SPT and reduces immobility times in OFT, allowing the reversal of the typical manifestations of depression [23]. Notably, EA treatment resulted in a structure and abundance of gut microbes in CUMS mice that more closely resembled controls, which is consistent with other studies showing that gut flora functions through the brain-gut axis to influence depression [9]. In conclusion, as a result of these findings, EA has demonstrated its effectiveness in treating depression-like behaviors caused by CUMS.

Intestinal flora in the feces of mice was analyzed in this study. The species of intestinal flora analyzed with statistically significant differences in the three sets of multiple comparisons at the genus level were Lactobacillus and Staphylococcus, and both species had LDA $>$2 (considered statistically significant). Therefore, we focused on the possible mechanisms by which changes in the abundance of Lactobacillus and Staphylococcus improve depression-like behavior in mice. In addition to this, there are Ruminococcaceae and Tyzzerella with LDA scores also $>$2. It has been suggested that they might also be the flora associated with depression-like behavior, so we discussed them as well.

This study found that NC and EA groups had higher relative Lactobacillus abundances than CUMS groups. Recent studies have shown that Lactobacillus has some depression-relieving effects. Lactobacillus Plantarum DP189 reduced hippocampal neuronal apoptosis in mice, attenuated depression-like behavior, and prevented hippocampal nerve damage by modulating immune function. DP189 reduced serum IL-1$\beta$ and TNF-$\alpha$, decreased mitogen-activated protein kinase 7 and c-jun amino-terminal kinase 2 levels in the hippocampus, and down-regulated the immune content of the pro-apoptotic protein Bax. These results suggest that DP189 therapy may prevent and/or attenuate cortisol-induced depression-like behavior and hippocampal neurological damage [24].

Meanwhile, it has been shown that Lactobacillus cases may improve the protein expression of the monoamines dopamine, norepinephrine, and 5-hydroxytryptamine, brain-derived neurotrophic factor (BDNF) and its tyrosine kinase receptor B (TrkB) receptor, N-methyl-D-aspartate receptor 1, through the BDNF-TrkB pathway, thereby improving depressive behavior in mice [19]. Long-term application of Lactobacillus rhamnosus (JB-1) induced region-dependent alterations in GAB1b mRNA in the brain and decreased the expression of GABA receptors in the hippocampus, amygdala, and blue spot, thus exerting an antidepressant effect, as well as reducing cortisol and improving cortisol-induced anxious-depressive behavior. These two effects are mediated by the vagus nerve [19].

In the present experiment, after chronic unpredictable mild stress (CUMS) stimulation, the number of lactobacilli in mice decreased, and after EA treatment, the number of lactobacilli increased close to normal levels and was statistically different (p = 0.013), which is the same as the experimental results of other researchers. Therefore, we propose a hypothesis that CUMS stimulation can reduce the abundance of Lactobacillus and EA can exert antidepressant effects by upregulating the abundance of Lactobacillus. The specific mechanism of Lactobacillus for depression is still being explored and may be a potential target for depression treatment.

As compared to the other two groups, the CUMS group had a higher relative abundance of Staphylococcus. Staphylococcus is a gram-positive bacterium associated with infection and toxicity. Staphylococcus aureus produces staphylococcal enterotoxins and glutamine, which stimulate the vagus nerve and send signals to the brain, causing vomiting and nausea behavior [25], suggesting that staphylococci may damage the nervous system. Some studies have demonstrated the antimicrobial activity of antidepressants selective 5-hydroxytryptamine reuptake inhibitors (SSRIs), including sertraline, paroxetine, and fluoxetine, against Staphylococcus and reduced the abundance of Staphylococcus [11].

This is consistent with the results of this experiment, no Staphylococcus was detected in the NC group, and the abundance of Staphylococcus increased after CUMS modeling. the abundance of Staphylococcus decreased after EA treatment, with a statistically significant difference compared to the CUMS group (p = 0.0032). This suggests a potential target of Staphylococcus for the treatment of depression and shows that EA may have some anti-inflammatory effects.

Ruminococcaceaes were more abundant in CUMS and EA groups than in NC groups (p = 0.059). Ruminococcaceae is a pathogenic bacterium [26]. Several studies have shown that Ruminococcus can reduce the antidepressant effect of antidepressant drugs and that antidepressant drugs can reduce the abundance of Ruminococcus [27]. Researchers have found that patients with autism spectrum disorders have higher levels of C. tumefaciens in their bodies [28]. These results imply that rumen bacteria may be a deleterious factor in depression.

In terms of the mechanism of action of rumen bacteria on depression, several studies have shown that rumen cocci can secrete tryptophan decarboxylase and catalyze the formation of the neurotransmitter tryptamine. Tryptophan induces the release of 5-hydroxytryptamine, which regulates intestinal motility. Since tryptophan decarboxylase secreted by C. rumen consumes tryptophan in the intestine causing a decrease in plasma tryptophan levels, leading to a decrease in 5-hydroxytryptamine levels in the brain, this may be a direct mechanism by which C. rumen affects anxiety-depressive behavior [29].

In the present experiment, the number of Ruminococcaceae in mice increased after chronic unpredictable mild stress (CUMS) stimulation and was statistically different (p = 0.03), which is consistent with the experimental results of other researchers. However, intriguingly, the number of rumen cocci did not change after EA treatment. EA treatment did not seem to affect the abundance of this flora, the mechanism of which is unclear.

The relative abundance of Tyzzerella in the NC group was significantly higher than other two groups in this study. Recent studies on Tyzzerella have focused on the relationship between Tyzzerella and cardiovascular disease and obesity. Some studies have shown that Tyzzerella is associated with Obese [30]. Some studies have shown that Tyzzerella is associated with dietary patterns and has an elevated abundance in the intestinal flora of mice on a high-fat diet [31, 32]. The relative abundance and bodyweight of Tyzzerella were elevated in high-fat diet mice in these studies. This is consistent with our experiments, where Tyzzerella abundance was higher in the NC group compared to the EA group (p = 0.001975), while the CUMS group did not persist to Tyzzerella.

We speculate that this is related to the fact that the CUMS modeling intervention may affect the mice’s food intake. However, some studies of postpartum depressive disorder have shown that a correlation was observed between Tyzzerella levels and the severity of depressive symptoms [33]. But in our experiments, a higher abundance of Tyzzerella was found in the NC group than in the other group. Even after the EA intervention, Tyzzerella abundance levels did not change significantly. This raises the hypothesis that Tyzzerella may affect mouse behavior, but that the altered abundance of Tyzzerella is not affected by CUMS intervention and EA treatment, but is more related to sex hormones and dietary patterns.

This study showed that the flora with statistically different alterations in the three groups of multiple comparisons was Lactobacillus and Staphylococcus. The possible mechanisms linking these two bacteria to depression-like behavior are also discussed. Lactobacillus, which has been shown in earlier studies to be possibly related to the pathogenesis of depression, was shown to increase in abundance and improve behavior in mice treated with EA in this experiment. Lactobacillus may improve depression-like behavior by inhibiting inflammatory substances, decreasing hippocampal neuronal apoptosis, and altering BDNF protein expression by the BDNF-TrkB pathway.

Patients with sepsis are known to have an ecological dysbiosis of the gut microbiota due to systemic inflammation. In turn, changes in the intestinal flora can exacerbate the symptoms of sepsis [34]. In addition, it has been shown that electroacupuncture improves the signs and symptoms of ulcerative colitis and restores the balance of the intestinal flora by a mechanism that may be immunomodulation, especially T-cell regulation [35]. Considering our findings, we hypothesize that immunomodulation may be one of the most important parts of the mechanism of the effects of EA on adjusting the abundance of the microbes.

In contrast to Lactobacillus, Staphylococcus aureus may lead to depressive-like behavior by producing staphylococcal enterotoxins that damage the nervous system, and EA treatment resulted in decreased abundance of S. aureus and improved behavior in mice.

In addition to this, some bacteria with possible roles are discussed. For example, Staphylococcus tumefaciens may be secreting tryptophan decarboxylase, which depletes plasma tryptophan causing a decrease in brain 5-hydroxytryptamine levels and leading to depression-like behavior. Typhimurium was found to be associated with obesity and dietary patterns, suggesting that Typhimurium may influence host behavior [36]. These results should be translated into the human population to confirm preliminary animal data. Accordingly, further investigations and future research may be warranted.

Our study has some limitations. First, we have not yet explored the brain interaction intestinal mechanism, the next step to improve the metabolome and brain neural and other studies. Second, the research conclusions for mental disorders should be cautiously applied to human beings due to racial differences and mouse model limitations. Further research will be conducted to overcome these shortcomings.