Six-week-old female BALB/c mice were housed once per cage at a temperature of 24°C ± 2°C, 40%–60% humidity, and a 12-h light/dark cycle. All animal experimental procedures were carried out according to the governmental guidelines and approved by the Ethics Committee on Laboratory Animals of the Shanghai Institute for Biomedical and Pharmaceutical Technologies (protocol code 2022-43, approval date 11/30/2022).

The probiotics used for intervention were a combination of Bacillus coagulans CGMCC10182 (5 × 10⁸ CFU/g) and C. butyricum CGMCC1647 (5 × 10⁷ CFU/g).

After one week of adjustable feeding, all mice were subjected to a 4-day sucrose preference test, and the mice whose sucrose preference was greater than 60% were selected for further experiments (Figure 1A). Forty-two selected BALB/c mice were divided into six groups: the control group (CON), model group (CUMS), low-concentration complex probiotics group (LOW), middle-concentration complex probiotics group (MID), high-concentration complex probiotics group (HIGH), and fluoxetine group (FLX). In addition to the CON group, the other groups were subjected to CUMS for 5 weeks by exposing them to two out of eight treatments each day: clamping of the tail (3 min), swimming in 4°C water (5 min), restraining in 50 mL tubes (3 h), crowded feeding (24 h), water deprivation (18 h), food deprivation (20 h), empty cage (24 h), tilted cage (24 h), damp cage (24 h), and darkness (36 h). The two stimuli were randomly selected per day and not applied the next day, but all the mice in the five groups were subject to the same stimuli per day to avoid difference.

During CUMS treatment, the three probiotic treatment groups received complex probiotics at 13 mg/kg (LOW), 52 mg/kg (MID), and 204 mg/kg (HIGH) once a day, which was determined by referring to the usage dose of B. coagulans in clinical trials (Zhang et al., 2023) and C. butyricum across murine models (Stoeva et al., 2021). The model group (CUMS) received the same volume of PBS, and the FLX group received fluoxetine 30 mg/kg once a day (Taguchi et al., 2016).

The sucrose preference test (SPT) can be used to assess depression in mice. The degree to which the mice preferred sugar water was inversely proportional to the degree of depression. This test was designed based on the SPT protocol (Liu M.-Y. et al., 2018). All mice were housed in a single cage, acclimated for 24 h, and then fasted and water deprived for 24 h. Each mouse was given one bottle of 1% sucrose solution and one bottle of sterile water. The consumption of sucrose solution and sterile water was recorded for 24 h (changing the position after 12 h), after which the sucrose preference rate was calculated. The sugar preference experiment was performed at weeks 0 and 5. Initially, mice with a sucrose preference greater than 60% were selected for modeling, as a sucrose preference <60% is defined as no preference in normal mice (Liu M.-Y. et al., 2018). In the fifth week, only the sugar-water test was performed on all the model mice.

After 5 weeks of treatment, the mice were sacrificed, and blood samples collected from the heart were centrifuged at 3,000 rpm for 10 min. The supernatant was transferred to a new microcentrifuge to obtain the serum, which was stored at −80°C. Half of brain was fixed with 4% (w/v) paraformaldehyde for histopathological staining. And the prefrontal cortex (PFC) from the other part of the brain was isolated and stored at −80°C after flash freezing in liquid nitrogen for biochemical tests and transcriptome sequencing. Feces were collected in sterile tubes and then stored at −80°C.

The levels of inflammatory neurotransmitters and hormones in serum and PFC were measured by ELISA. The PFC were rinsed with precooled PBS (0.01 M, pH = 7.4) and cut into pieces after weighing. The trimmed PFC was homogenized in PBS (1:9 weight-to-volume ratio, and a protein inhibitor was added to PBS), the homogenate was centrifuged at 5,000 × g for 10 min, and the supernatant was stored at −80°C. Briefly, 100 μL of standard or sample was added to each well and incubated for 90 min at 37°C. Then, 100 μL of biotinylated detection antibody working solution was added to each well, and the plates were incubated for 60 min at 37°C. After the plates were aspirated, they were washed 3 times. A total of 100 μL of HRP conjugate working solution was added. The mixture was incubated for 30 min at 37°C. The plate was aspirated and washed 5 times. Subsequently, 90 μL of TMB substrate reagent was added, and the mixture was incubated for 15 min at 37°C. After the incubation, 50 μL of stop solution was added to each well to stop the reaction. Finally, the absorbance at 450 nm was measured immediately and calculated. The levels of adrenocorticotropic hormone (ACTH), corticosterone (CORT), and 5-hydroxytryptamine (5-HT) were determined using ELISA kits (Elabscience, China), and the level of brain-derived neurotrophic factor (BDNF) was determined using an ELISA kit (Boster, China).

The half brain was immersed in 4% paraformaldehyde solution for 24–48 h, stored, and transported at room temperature. The brain was removed from the fixative, and a scalpel was used to smooth the target tissue in the fume hood. The cut tissue and the corresponding label were placed in the embedding frame. Then, the dehydration box was placed in a dehydrator for dehydration with gradient alcohol. The following steps were used: 75% alcohol for 4 h, 85% alcohol for 2 h, 90% alcohol for hours, 95% alcohol for 1 h, anhydrous ethanol for 30 min, anhydrous ethanol for 30 min, alcohol for benzene for 5–10 min, xylene I for 5–10 min, 65°C melting paraffin I for 1 h, 65°C melting paraffin II for 1 h, and 65°C melting paraffin for 1 h. Next, the wax-soaked tissue was embedded in the embedding machine. First, the melted wax was placed into the embedding frame, and before the wax solidified, the tissue was removed from the dewatering box and put into the embedding frame according to the requirements of the embedding surface, and the corresponding label was affixed. After the wax solidified, the wax block was removed from the embedding frame and repaired. The trimmed wax blocks were placed into a paraffin slicer for slicing at a thickness of 4 μM. The tissue was flattened when the slice floated on the 40°C warm water of the spreading machine, and the tissue was picked up by glass slides and baked in an oven at 60°C. After the water-baked dried wax was melted, it was removed and stored at room temperature.

The paraffin sections were sequentially immersed in environmentally friendly dewaxing transparent liquid I for 20 min, environmentally friendly dewaxing transparent liquid II for 20 min, anhydrous ethanol I for 5 min, anhydrous ethanol II for 5 min, and 75% ethyl alcohol for 5 min and then rinsed with tap water. The tissue slices were treated with dye solution for 2–5 min and rinsed with tap water. Then, 0.1% glacial acetic acid was used for slight differentiation, the reaction was terminated by washing with running water, and the degree of differentiation was controlled under a microscope. The sample was washed with tap water and dried in an oven.

Fecal samples from each mouse were collected at the end of the 5-week treatment before sacrifice, and 42 samples were collected in total. A QIAamp DNA Stool Mini Kit (QIAGEN) was used for bacterial genomic DNA extraction. For the detection of 16S rRNA genes, the V3-V4 region was amplified with the primers 338F and 806R (Huse et al., 2007) using TransStart Fastpfu DNA Polymerase (TransGen). The thermocycling steps were as follows: 95°C for 5 min, 20 cycles at 95°C for 45 s, 55°C for 30 s, 72°C for 30 s, and a final extension step at 72°C for 10 min. Three repeat PCR amplifications of each sample were performed, and then all amplicons were purified with an AxyPrep DNA Gel Extraction Kit (AXYGEN) and pooled equivalent after being assessed by spectrophotometry (QuantiFluor-ST, Promega). Sequencing of 16S rRNA gene amplicons was performed on an Illumina NextSeq 2000 instrument (2 × 300 bp).

QIIME 2 (version 2023.9) (Bolyen et al., 2018) was used to analyze the microbiota data. DADA2 of the QIIME 2 plugin was used for quality control and ASV exaction with default parameters. The minimum sample size was the criteria for data normalization. Community richness, evenness, and diversity analysis (Shannon, Simpsonenven, ACE, Chao, and Good’s coverage) were performed using Mothur (Schloss et al., 2009). Taxonomy was assigned using the software RDP classifier (Wang et al., 2007) at the default parameter (80% threshold) based on the Ribosomal Database Project (Cole et al., 2009). Species assignment was performed by BLASTN against the SILVA reference database (version 138.1) (Quast et al., 2012) with the following parameters: the highest score, identity >97%, and alignment >97%. Differences between groups were assessed using ANOSIM and PerMANOVA in PAST (version 4.16c). Principal coordinate analysis (PCoA) was based on the Bray‒Curtis distance matrix. LEfSe (Segata et al., 2011) used the Kruskal‒Wallis test to detect differentially abundant taxa (p < 0.05) between the two groups and estimate the linear discriminant analysis effect size (LDA score >2.0). By normalizing the 16S rRNA gene copy numbers, PICRUSt2 (Douglas et al., 2020) was used to predict the microbiome functions based on the MetaCyc pathways. The difference in predicted functions was analyzed using STAMP (Parks et al., 2014) with default parameters.

The correlation analysis between the transcriptome and gut microbiota was performed through a nonparametric Spearman rank correlation algorithm, by calculating the correlation coefficient between the abundance of species and TPM of genes, and the parameters were set as a coefficient >0.68 or −0.68 and FDR<0.05, which were considered to represent high correlations.

Statistical analyses were conducted by using GraphPad Prism 9. All values are expressed as the means ± standard errors of the means. All the data were tested using one-way ANOVA. p < 0.05 was considered to indicate statistical significance.

To investigate the effect of treatment on CUMS-induced depression-like behaviors, the SPT, TST, FST, and OFT were used to detect depression and antidepression-like behaviors in mice.

After five weeks of treatment, sucrose preference did not show a significant difference among the six groups in SPT (Figure 1B), though the CUMS group showed a decreasing trend relative to CON and the HIGH group showed an increasing trend, which indicated that neither CUMS stimulation nor the probiotics or fluoxetine influenced anhedonia.

CUMS stimulation significantly increased immobility time in the FST compared with that in the CON group (Figure 1C, p < 0.05), and the high concentration of probiotics reduced immobility time to the CON level, showing a greater amelioration effect than did Fluoxetine (Figure 1C).

Moreover, as the TST showed, the immobility time of the CUMS group in TST was significantly greater than that of the CON group (Figure 1D, p < 0.0001). Treatment with low- and mid-concentration probiotics and Fluoxetine notably reduced immobility time (p < 0.01). The TST and FST showed that CUMS can lead to depression-like behavior in mice, and our treatment with different concentrations of the probiotics improved CUMS-induced despair-like behavior.

The OFT was used to evaluate the anxiety-like behavior of the mice, and the CUMS-treated mice exhibited significantly reduced activity in the central region compared with that in the CON group (Figures 1E, F, p < 0.05), suggesting that the CUMS-exposed mice exhibited anxiety-related behaviors. Although not significant, the central distance tended to improve in the LOW and MID groups compared with that in the CUMS group. These results suggested that low and middle concentrations of the probiotics may improve CUMS-induced anxiety.

CUMS significantly increased the serum ACTH concentration (Figure 2A, p < 0.05), and the probiotics tended to reduce the level of ACTH in serum. However, the levels of CORT in the serum (Figure 2B) and BDNF (Figure 2C) in the prefrontal cortex (PFC) showed no significant differences among the CON group, CUMS group, and probiotics treatment groups. Moreover, there was a noticeable decreasing trend in the 5-HT concentration in the PFC following CUMS, and various concentrations of the probiotics increased the 5-HT concentration to varying degrees compared to that in the CUMS group (Figure 2D).

Nissl staining was applied to track the morphological alterations of neurons in the cortex and hippocampus. As shown in Figure 2E, the cortex region of the CON group showed normal, numerous, and compactly arranged neurons. In contrast, the CUMS group exhibited a diminished quantity of neurons, along with irregular neuronal arrangement, nuclear shrinkage, and indistinct nuclear boundaries. The probiotic treatment reversed the effects of CUMS treatment. All concentrations of probiotics significantly reduced these signs of histopathological damage.

In addition, the neurons in the hippocampal CA1 region were sparsely distributed and exhibited a reduced population compared to those in the CON group (Figure 2F), while the number of Nissl bodies was notably increased in the CON group and the probiotics group.

To investigate the potential molecular mechanisms underlying the ameliorative effect of the probiotics on the CUMS model, transcriptomic analysis was carried out on the PFC of mice from the six groups. Compared with those in the CON group, 270 DEGs were identified in the CUMS group, with 87 upregulated genes and 183 downregulated genes (Figure 3A). The top 10 GO terms enriched for the downregulated genes in the CUMS group are shown in Figure 3D, most of which were associated with cellular components. The top 10 GO terms enriched for the upregulated genes in the CUMS group are shown in Figure 3E, such as anion transport (GO:0006820), neuron projection fasciculation (GO:0106030), and axonal fasciculation (GO:0007413). KEGG pathway enrichment analysis confirmed the impact of CUMS treatment on specific biological pathways. The downregulated genes after CUMS treatment were enriched in several pathways such as the Wnt signaling pathway and the TGF-beta signaling pathway (Figure 3F), while the upregulated genes were enriched in the cGMP-PKG signaling pathway, Neuroactive ligand-receptor interaction pathway, and Calcium signaling pathway, etc. (Figure 3G).

In addition, enrichment analysis of DEGs in each intervention group (LOW, MID, HIGH, and FLX) relative to CUMS was performed. The upregulated genes in both the LOW and HIGH groups were significantly enriched in the PI3K-Akt signaling pathway (Supplementary Figure S1).

Additionally, a comprehensive analysis was conducted to delineate the DEGs of CUMS group that were reversed by different concentrations of probiotics (Figures 3B, C; Tables 1, 2). We identified several genes related to the nervous system or depression, such as Sfrp5 and Angpt2. Sfrp5 was downregulated in the CUMS group and reversed by mid concentration of probiotics significantly (Table 1; Supplementary Table S1). Moreover, Angpt2 was downregulated in the CUMS group and reversed by high concentration of the probiotics and fluoxetine significantly.

For the DEGs in the CUMS group whose expression were reversed by probiotics treatment, the enrichment analysis revealed that they were mainly enriched in the GO terms such as sodium ion import across plasma membrane (GO:0098719), glutathione peroxidase activity (GO: GO:0004602), glutathione transferase activity (GO:0004364), etc. (Table 3).