Gut microbiota dysbiosis-mediated ceramides elevation contributes to corticosterone-induced depression by impairing mitochondrial function

Up to 322 million people (4.4% of the world’s population) are affected by depressive disorders, which were ranked as the 13th leading cause of the global disease burden in 2019¹. According to a recent article published in the Lancet, the COVID-19 pandemic has led to a remarkable increase in the prevalence of depressive disorders globally², thus contributing to a higher disease burden and a worsened economic situation. However, despite these costs, ~30% of individuals diagnosed with major depressive disorder (MDD) fail to exhibit significant improvement or merely display a partial reaction to conventional pharmacological and psychological interventions utilized for depression treatment. Commonly prescribed antidepressants have various side effects, with insomnia and anxiety being the most frequently observed symptoms. Hence, there is an urgent need to investigate the underlying pathogenic mechanisms and to develop innovative therapeutic interventions for depressive disorders.

Recently, the gut microbiota (GM), containing trillions of living microorganisms, is increasingly regarded as an indispensable “organ”, which serves an important role in the human body. Although the GM for a newborn appears haphazard, the dynamic microbial population develops rapidly until an adult-like composition is established by the age of 3³. Therefore, the GM plays a fundamental role in lifelong health. Numerous factors, including age, diet, nutrition, genetics, metabolism, medication exposure, and stress, can influence microbiota composition⁴. Furthermore, it is increasingly recognized that gut microbes greatly influence the functioning of body systems, including metabolism⁵, immunity⁶, and the central nervous system⁷. It is well documented that the gut and brain work through a bidirectional communication network called the gut-brain axis, linking the emotional brain system and peripheral intestinal functions⁸. As technology has progressed, scientists have increasingly focused on the involvement of GM in the communication network of the gut-brain axis.

Accumulating evidence suggests that GM dysbiosis is associated with depression pathogenesis. Several clinical studies have reported that patients diagnosed with depression show altered composition of GM^{9,10,11,12,13,14}, and fecal microbiota transplantation (FMT) from depressed patients induces depressive-like behaviors in germ-free recipient rodents^10,15. More importantly, the symptoms of patients with depression can be reversed by FMT from healthy people^16,17, and probiotic administration can attenuate abnormal behaviors in depressed mice¹⁸ and emotional disorders in patients with depression¹⁹. However, despite extensive research, the mechanism by which GM dysbiosis induces depression remains unclear.

In the present study, a typical depressive-like mouse model, chronic treatment with corticosterone (CORT), showed altered GM composition. FMT from CORT-treated mice transferred depression-like phenotypes and their dominant GM, especially Bifidobacterium, and Lactobacillus, to the recipient mice. Remarkably increased ceramide levels were detected in the feces of CORT-treated and recipient mice. Metagenomic sequencing and correlation analysis revealed the co-relationship between ceramides and increased abundance of Bifidobacterium and Lactobacillus. Colonization study exposed that oral gavage with Bifidobacterium pseudolongum (B. pseudolongum) and Lactobacillus reuteri (L. reuteri) could induce elevations of gut ceramides. To confirm the role of ceramides in the development of depression, a ceramide mixture was administered to mice. We found that treatment with ceramides was sufficient to recapitulate depressive-like phenotypes in mice. RNA-sequencing (RNA-seq) data revealed that the key differentially expressed genes (DEGs) between the hippocampus of ceramide-treated mice and the control group were enriched in the oxidative phosphorylation (OXPHOS) pathway, indicating that dysregulated mitochondrial energy metabolism or ATP deficits might be the underlying mechanism by which ceramides cause depression. Finally, mitochondrial function assay validated hippocampal mitochondrial dysfunction in CORT- or ceramides-treated mice. In this regard, we conclude that long-term exposure to CORT results in altered GM composition and consequent ceramide production, thus leading to dysregulated mitochondrial energy metabolism in the hippocampus, which may contribute to the development of depressive disorders.

CORT induces depression and altered composition of fecal microbiota in mice

Chronic exposure to CORT is a well-validated pharmacological stressor that induces depressive-like behaviors in rodents^20,21,22; therefore, C57BL/6 J mice were exposed to CORT for 36 days, and their behavior was evaluated (Fig. 1a). Compared to the control group, CORT-treated mice showed increased immobility time in the FST and TST (Fig. 1b, c), demonstrating a depressive-like state. In addition, CORT-treated mice developed anxiety-like behaviors (Supplementary Fig. 1). It is well documented that the reduction in adult hippocampal neurogenesis is involved in the pathogenesis of depression²³, thus, we examined whether chronic CORT exposure influenced neurogenesis in the dentate gyrus (DG) of the hippocampus. As expected, CORT-treated mice displayed a remarkable decrease in the number of EdU+ cells in the DG (Fig. 1d, e), indicating reduced hippocampal neurogenesis. To further elucidate whether CORT treatment could influence the activity of the brain areas critical for mood regulation, immunostaining of c-Fos, a well-documented marker of neuronal activity²⁴, was performed in hippocampus and medial prefrontal cortex (mPFC). Chronic exposure to CORT significantly elevated c-Fos expression in hippocampus without influencing c-Fos expression in the mPFC (Supplementary Fig. 2). Taken together, these results demonstrate that chronic exposure to CORT leads to depression-like phenotypes and hippocampal alterations in mice.

a Experimental design and the timeline. b Comparison of the immobility time in the TST for the two groups of mice. c Comparison of the immobility time in the FST for the two groups of mice. d Representative images of EdU staining within the DG of hippocampus. e Quantitative analysis of EdU+ cell within the DG of the hippocampus. f Beta-diversity-based principal component analysis (PCA) plots showing microbial compositional difference at genus level between the two groups. g Alpha diversity analysis of fecal microbial community diversity (Sobs index). h Bar graph showing the relative bacterial abundance at the order level. i Bar graph showing the relative bacterial abundance at the genus level. j Bar graph showing the relative bacterial abundance at the species level. For behavioral tests, n = 24 for each group, comparisons were performed using Student’s t test. For the EdU test, n = 5 for each group, comparisons were performed using nonparametric Mann–Whitney U test. For 16S rRNA sequencing, the Veh-treated group: n = 8; CORT-treated group: n = 12; comparisons were performed using Welch’s t test. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. CORT corticosterone, Veh vehicle, TST tail suspension test, FST forced swimming test, FC fecal collection, EdU-I EdU injection, DG dentate gyrus. Scale bar: 100 µm. Sobs: total number of species observed in a sample.

Full size image

A growing body of evidence indicates that GM dysbiosis is related to depressive disorders^{9,10,11,12,13}. To determine whether the composition of GM was altered in CORT-induced depressed mice, fecal microbiota composition was analyzed using 16S ribosomal RNA (rRNA)-based MiSeq sequencing. β-diversity-based principal component analysis (PCA) plots at the genus level showed two distinguishable clusters, indicating differences in fecal microbial composition between control and CORT-treated mice (Fig. 1f). Fecal microbial complexity was estimated based on alpha diversity (Sobs index), which showed that CORT exposure led to a reduction in fecal microbial diversity (Fig. 1g). Comparative analysis was then performed to identify the taxa with significantly altered relative abundance, which showed significant increases in the relative abundances of Lactobacillales (20.47% vs. 32%; p = 0.022) and Bifidobacteriales (1.204% vs. 3.015%; p = 0.035) at the order level in CORT-treated mice (Fig. 1h). At the genus level, Lactobacillus (20.42% vs. 31.56%, p = 0.027) was the most predominant bacterium, followed by Bacillus (1.47% vs. 4.86%, p = 0.0006) and Bifidobacterium (1.20% vs. 3.02%, p = 0.035) in CORT-treated mice (Fig. 1i). At the species level, the relative abundances of Lactobacillus vaginalis (1.474% vs. 4.856%, p = 0.0006) and unclassified Bifidobacterium (1.204% vs. 3.015%, p = 0.035) were significantly higher in CORT-treated mice (Fig. 1j). These data suggest that chronic exposure to CORT leads to depression-like behaviors and significant alterations in GM composition in mice.

FMT from depressed mice leads to depression and successful engraftment of dominant microbiota in GM-depleted mice

To determine whether GM composition plays an important role in the mental state, a broad-spectrum antibiotic mixture was administered to mice to induce GM depletion. The fecal microbiota of control or CORT-treated mice was then transferred to antibiotic-treated (Abx) mice. One week after FMT, behavioral tests were performed to evaluate the mental state of the recipient mice (Fig. 2a). Recipient mice transplanted with microbiota from CORT-treated mice (Abx-FMT/C) showed significantly increased immobility time in the FST (Fig. 2b), demonstrating a depression-like state, although there was no obvious change in immobility time in the TST (Fig. 2c). Abx-FMT/C mice also showed remarkably decreased proliferation in the DG (Fig. 2d, e). Additionally, Abx-FMT/C mice exhibited anxiety-like behavior (Supplementary Fig. 3). These data suggest that FMT from CORT-treated mice transferred a depression-like phenotype to recipient mice.

a Schematic for experimental procedures. b Comparison of the immobility time in the FST for the two groups of mice. c Comparison of the immobility time in the TST for the two groups of mice. d Representative images of EdU staining within the DG of hippocampus. e Quantitative analysis of EdU+ cell within the DG of the hippocampus. f Beta-diversity-based PCA plots showing microbial compositional difference at genus level between the two groups. g Taxonomic cladogram obtained using LEfSe analysis. Taxonomic levels are represented by the rings from phylum level (innermost ring) to genus level (outermost ring). LEfSe identified the taxa (highlighted by small circles and by shading) with significant difference in abundance value between the two groups. Colors of circle and shading indicate the microbial lineages that are enriched within the corresponding groups. h Bar graph showing the relative bacterial abundance at the order level. i Bar graph showing the relative bacterial abundance at the genus level. j Bar graph showing the relative bacterial abundance at the species level. For behavioral tests, n = 15 for each group, comparisons were performed using Student’s t test. For the EdU test, n = 5 for each group, comparisons were performed using nonparametric Mann–Whitney U test. For 16S rRNA sequencing, N = 7 for each group; comparisons were performed using Welch’s t test. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. CORT corticosterone, Veh vehicle, TST tail suspension test, FST forced swimming test, FC fecal collection, EdU-I EdU injection, DG dentate gyrus. Abx-FMT/V antibiotic treatment followed by FMT from vehicle-treated mice, Abx-FMT/C: antibiotic treatment followed by FMT from CORT-treated mice; Scale bar: 100 μm.

Full size image

Next, we investigated the composition of GM in the recipients to evaluate the efficiency of FMT. β-diversity-based PCA plots showed separate clusters for Abx-FMT/C and Abx-FMT/V mice at the genus level (Fig. 2f), indicating a significant difference in the microbiota composition between the two groups of mice. Linear discriminant analysis Effect Size (LEfSe) identified the taxa showing different abundance values between the two groups and showed that Lactobacillales and Bifidobacteriales were enriched in Abx-FM/C mice (Fig. 2g). Comparative analysis was performed to confirm the taxa with significantly different abundances between the two groups of mice. At the order level, the relative abundance values of Lactobacillales (3.488% vs. 8.971%, p = 0.003) and Bifidobacteriales (1.978% vs. 5.694%, p = 0.029) were significantly higher in the Abx-FM/C group than in the Abx-FMT/V mice (Fig. 2h). At the genus level, Abx-FMT/C mice also showed higher levels of abundance of Lactobacillus (3.384% vs. 8.359%, p = 0.005) and Bifidobacerium (1.978% vs. 5.694%, p = 0.029) than Abx-FMT/V mice (Fig. 2i). Furthermore, at the species level, the abundance of unclassified Lactobacillus (3.069% vs. 8.333%, p = 0.003) and unclassified Bifidobacterium (1.978% vs. 5.694%, p = 0.029) in the Abx-FMT/C group were significantly higher than those in the Abx-FMT/V group (Fig. 2j). These results suggest that GM, especially the key species enriched in CORT-treated mice, were successfully transferred to GM-depleted mice by FMT.

Fecal metabolic profiling exposes remarkably increased ceramides in CORT-treated and Abx-FMT/C mice

GM can interact with the host by producing an extremely diverse reservoir of metabolites from exogenous undigested dietary substrates or endogenous host compounds. Recent evidence also suggests that alterations in GM and microbiota-derived metabolites are associated with the occurrence of depression^25,26. Therefore, to ascertain the key metabolites that might contribute to the development of depressive-like behaviors in animal models, fecal metabolome analysis was performed. PCA analysis showed two characteristic clusters of the individual sample groups (Fig. 3a: Veh-treated group vs. CORT-treated group; Fig. 3b: Abx-FMT/V group vs. Abx-FMT/C group), indicating markedly different metabolic characteristics between the groups. The volcano plots exhibited differential metabolites that contributed to sample separation (Fig. 3c, d). Eighteen upregulated metabolites and fourteen downregulated metabolites (Fig. 3e), which were shared among the comparison groups, were selected to evaluate the fecal metabolic profiles of the depressed mice. According to the 32 (18 + 14) differential metabolites shown in the heatmap (Fig. 3f), ceramides, the key molecules in sphingolipid metabolism, were significantly increased in CORT-treated and Abx-FMT/C mice (Fig. 3g–i). This result indicated a substantial increase in the levels of ceramides in both CORT-treated mice and recipient mice that received FMT.

a PCA score plot for the metabolomics data acquired from Veh-treated mice and CORT-treated mice in positive ion mode. b PCA score plot for the metabolomics data acquired from Abx-FMT/V mice and Abx-FMT/C mice in positive ion mode. c Volcano plots showing the differential metabolites between Veh-treated mice and CORT-treated mice. The upregulated, downregulated, and unchanged metabolites are presented by red, blue, and gray dots, respectively. d Volcano plots showing the differential metabolites between Abx-FMT/V mice and Abx-FMT/C mice. e The Venn diagram showing differential metabolites unique or shared among the groups. Left: the upregulated metabolites; right: downregulated metabolites. f Heatmap showing the 32 differential metabolites among the groups. Red represents upregulated metabolites, and blue represents downregulated metabolites. g–i Relative abundance of 3 types of ceramides among the groups. N = 10 for each group. Unpaired t tests were used to compare the differences between the two groups. *p < 0.05, **p < 0.01, ***p < 0.001.

Full size image

Identification of GM contributing to the increased production of ceramides by hosts

Aberrant ceramide metabolism is associated with the pathogenesis of depressive disorders^27,28,29. Next, we sought to identify the source of ceramides, and metagenomic sequencing was conducted to describe the functional repertoire of the entire gut microbial community. The PCA plots showed that the GM composition changed dramatically between the groups (Fig. 4a). At the order level, the relative abundance of Lactobacillales (10.18% vs. 4.864%; p = 0.001) and Bifidobacteriales (3.601% vs. 0.942%; p = 0.01) was significantly increased in CORT-treated mice (Fig. 4b). At the genus level, Lactobacillus (9.513% vs. 4.347%, p = 0.002) was the most predominant bacterium, followed by Bifidobacterium (3.569% vs. 0.920%, p = 0.001), in CORT-treated mice (Fig. 4c). Although spearman correlation analysis revealed a positive correlation between ceramide and the relative abundances of Bifidobacterium pseudolongum and Lactobacillus reuteri (Supplementary Fig. 4a), we did not identify a significant correlation between the strains (Bifidobacterium pseudolongum and Lactobacillus reuteri) and dihydroceramide (DHCer, an immediate precursor to ceramide) or sphingomyelin (SM, downstream metabolite of ceramide) (Supplementary Fig. 4b). This result suggested that the strains might not be strongly correlated with upstream/downstream metabolites within the ceramide synthesis pathway. This finding raised an important question regarding whether ceramides are directly derived from bacteria or if bacteria influence the host to produce ceramides. At the species level, we observed a notable increase in the relative abundances of six strains within the genus Bifidobacterium (Bifidobacterium pseudolongum, Bifidobacterium longum, Bifidobacterium choerinum, Bifidobacterium cuniculi, Bifidobacterium animalis, unclassified Bifidobacterium) and five strains within the genus Lactobacillus (Lactobacillus vaginalis, Lactobacillus intestinalis, unclassified Lactobacillus, Lactobacillus sp. ASF360, Lactobacillus reuteri) in the CORT-treated group. (Fig. 4d). Next, we sought to determine the source of ceramides, whether they were derived from bacteria or the hosts. Serine palmitoyl transferase (SPT) catalyzes the initial conserved step in ceramide synthesis, and the presence of spt gene in bacterial genomes might be a reliable indication that the bacterial species may synthesize sphingolipids³⁰. In addition, a recent study reported that several genes are homologous to spt gene³¹. Through a comparison of gene segments in our metagenomic sequencing data with the reported bacterial spt homologous genes, we identified 293 similar sequences. Among these, 25 sequences exhibited significant differences between CORT and control groups (Supplementary Table 4, marked in red). Subsequently, we compared these 25 sequences with annotated sequences of Lactobacillus and Bifidobacterium (totally 14,247 sequences, Supplementary Table 5). Interestingly, two sequences (Supplementary Tables 4, 5, highlighted in yellow) were annotated into Bifidobacterium pseudolongum (belonging to the genus Bifidobacterium), showing a significantly increased abundance in the CORT-treated group. However, we observed that the abundance of these two gene segments was notably low compared to other gene segments. Given their low abundance, these two spt homologous gene segments might not significantly contribute to the production of ceramides. We were interested in B. pseudolongum and L. reuteri, for a profound enrichment in the feces content of CORT-treated mice and a remarkable increase of these species compared with those in control mice (Fig. 4e). Therefore, the supernatant of the two strains was collected for lipidomics analysis. The ceramide levels in the bacterial culture supernatant were similar to those in the blank control culture medium (Fig. 4f, g), indicating that the strains did not produce additional ceramides under these experimental conditions. Therefore, we suggest that ceramides in our study are not derived from bacterial sources. To gain more insights into the impact of specific strains on ceramides production by hosts, we employed mono-colonization studies by oral gavage of Abx mice with selected strains. Mice colonized with B. pseudolongum or L. reuteri showed remarkably increased production of multiple types of ceramides compared with Abx mice gavaged with PBS (Abx-PBS) (Fig. 4h, i). These data suggest that the significantly elevated levels of GM, specifically Bifidobacterium, and Lactobacillus, might influence the hosts to produce increased ceramides in the intestines.

a PCA plots showing microbial compositional difference derived from metagenomic data at genus level between the two groups. b Relative abundance of substantially changed bacteria taxa at the order level. c Relative abundance of substantially changed bacteria taxa at the genus level. d Taxonomic cladogram obtained using LEfSe analysis. Taxonomic levels are represented by the rings from phylum level (innermost ring) to species level (outermost ring). LEfSe identified the taxa (highlighted by small circles and by shading) with significant differences in abundance value between the two groups. The colors of the circle and shading indicate the microbial lineages that are enriched within the corresponding groups. The diameter of each circle is proportional to taxon abundance. e Bar graph showing the relative bacterial abundance at the species level. f Evaluation of levels of ceramides in supernatant of the culture medium of L. reuteri. g Evaluation of levels of ceramides in supernatant of the culture medium of B. pseudolongum. h Quantitative analysis of gut ceramides in Abx mice gavaged with B. pseudolongum. Measured concentrations (ng) were normalized to fecal weight (g). i Quantitative analysis of gut ceramides in Abx mice gavaged with L. reuteri. Measured concentrations (ng) were normalized to fecal weight (g). N = 6 for each group. Comparisons were performed using Welch’s t test (b, c, e) and unpaired t tests (f, g, h, i). *p < 0.05, **p < 0.01, ***p < 0.001.

Full size image

Ceramides induce depressive-like phenotypes and dysregulation of OXPHOS-associated DEGs

In order to explore the role of ceramides in the development of depressive disorders, the ceramides mixture (CerM) was administered to mice for 36 days. Quantification of gut ceramides was performed by using LC-MS/MS-based targeted lipidomics analysis, the ceramides-treated mice showed comparable levels of gut ceramides when compared to CORT-treated mice (Supplementary Fig. 5). This finding suggests that the concentration of ceramides used in our study is appropriate and reflective of similar pathological conditions under CORT treatment. In addition, we interestingly identified that several groups, including CORT-treated mice, mice colonized with B. pseudolongum or L. reuteri, and CerM-treated mice, showed a significant elevation of Cer (d18:1/14:0) (C14 ceramide) (Fig. 4h, i, see also Supplementary Fig. 5), indicating a potential role of C14 ceramide as a pathological mediator which might contribute to the alterations observed in the animals. Animal behaviors and hippocampal neurogenesis were evaluated following CerM treatment (Fig. 5a). Compared to the control group, CerM-treated mice showed significantly increased immobility time in the FST (Fig. 5b), whereas no significant difference was detected in the TST (Fig. 5c) and anxiety-associated behavioral tests (data not shown). Moreover, CerM-treated mice showed a remarkable decrease in proliferation markers EdU (Fig. 5d, e) and Ki67 (Fig. 5f). These results indicate that ceramide treatment is sufficient to induce depression-like phenotypes in mice.

a Schematic for animal experimental procedures. b Comparison of the immobility time in the FST for the two groups of mice. c Comparison of the immobility time in the TST for the two groups of mice. d Representative images of EdU staining within the DG of hippocampus. e Quantitative analysis of EdU+ cell within the DG of the hippocampus. f mRNA expression level of Ki67. g Volcano plots of the DEGs between the two groups. The upregulated genes are presented by red dots, and the downregulated genes are presented by blue dots. h GSEA enrichment score curves of the top 3 GO pathways. The top panel shows the enrichment scores of each pathway. The middle vertical bars display the position of each gene. The bottom panel indicates the ranking metric scores. i Heatmap of 17 DEGs enriched in oxidative phosphorylation KEGG pathway. j The schematic of the underlying mechanism by which chronic exposure to CORT induces depressive-like behaviors. For behavioral tests, n = 15 for each group, comparisons were performed using Student’s t test. For the EdU test, n = 5 for each group, comparisons were performed using nonparametric Mann–Whitney U test. For RNA-seq data, n = 3 for each group. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01. Veh/C: group of mice treated with vehicle of ceramides; CerM group of mice treated with ceramides mixture. OFT open field test, EPM elevated plus maze, TST tail suspension test, FST forced swimming test, EdU-I EdU injection, AHN adult hippocampal neurogenesis. Scale bar: 100 μm.

Full size image

Transcriptional profiles of the hippocampus were detected using RNA-seq to explore the possible underlying mechanisms by which ceramides cause depression. A total of 401 DEGs were identified, of which 171 were upregulated, and 230 were downregulated (Fig. 5g). Gene set enrichment analysis (GSEA) revealed that 401 DEGs were significantly enriched in four gene ontology (GO) pathways (“cellular respiration” pathway is completely overlapped with “aerobic respiration” pathway) (Fig. 5h), among which three were mitochondrial respiration-associated pathways. The KEGG enrichment analysis revealed that the 17 DEGs were highly enriched in the OXPHOS pathway, which was ranked as the top 1 KEGG pathway (Supplementary Fig. 6a, Supplementary Table 6. orange: OXPHOS-associated genes; green: proliferation-associated genes). We found that 16 of 17 OXPHOS-related DEGs were downregulated in the ceramide-treated group (Fig. 5i), indicating dysfunction of the mitochondrial respiratory chain in the hippocampus. Protein–protein interactions (PPI) network analysis revealed that five of the 17 DEGs were among the top 20 hub genes (Supplementary Fig. 6b, Supplementary Table 6). Furthermore, it highlighted the interactions between DEGs related to the OXPHOS pathway and those associated with cell proliferation (Supplementary Fig. 6b). This suggests that the abnormal downregulation of OXPHOS-related DEGs is a key factor contributing to hippocampal alterations induced by ceramide administration. qPCR exposed significant downregulation of several OXPHOS-related genes in ceramides treated group (Supplementary Fig. 7a), and downregulation of several OXPHOS-related genes were also identified in hippocampus of CORT-treated and Abx-FMT/V mice (Supplementary Fig. 7b, c). Indeed, the 16 downregulated DEGs enriched in the OXPHOS pathway were responsible for encoding crucial components of mitochondrial respiratory chain complexes. Damage to the mitochondrial respiratory chain can decrease mitochondrial membrane potential (ΔΨm) and induce dysregulated ATP production. Therefore, ΔΨm and ATP levels were investigated to evaluate mitochondrial function in hippocampus tissue. Consistent with the RNA-seq data, ceramides administration significantly decreased ΔΨm and ATP levels in hippocampus (Supplementary Fig. 8a, b), indicating a disrupted mitochondrial function in hippocampus of ceramides-treated mice. Moreover, Abx-FMT/C mice also showed remarkably decreased ΔΨm and ATP levels in hippocampus when compared to those of Abx-FMT/V mice (Supplementary Fig. 8c, d). Interestingly, in CORT-treated group, we observed significantly elevated ATP production despite a decrease in mitochondrial membrane potential (Supplementary Fig. 8e, f). This seemingly contradictory result prompted us to delve deeper into the literature. We propose that this increased ATP might be due to CORT-induced activation of glial cells, which are known to release ATP in response to stress³². This increased glial ATP release could potentially mask the reduced ATP production by neurons caused by ceramide accumulation, leading to the observed discrepancy. Subsequently, in order to prove that CORT-induced abnormal alterations are mediated by host ceramide production, ceramide synthesis was inhibited by myriocin. Myriocin notably attenuated depressive-like behavior in CORT-treated mice, as indicated by a significant reduction in immobility time observed in both the FST and TST (Supplementary Fig. 9a, b). Furthermore, myriocin treatment led to a significant increase in the number of EdU-positive cells in CORT-treated mice (Supplementary Fig. 9c, d), indicating enhanced neurogenesis. Additionally, an increase in hippocampal mitochondrial membrane potential was identified in CORT-treated mice following myriocin treatment (Supplementary Fig. 9e), suggesting improved mitochondrial function. These results indicated a potential mechanism whereby CORT-induced GM dysbiosis and subsequent gut ceramides production may lead to mitochondrial dysfunction and energy deficits, which subsequently causes reduced adult hippocampal neurogenesis (AHN) and the consequent development of depressive-like behaviors in mice (Fig. 5j). This hypothesis raised a question how ceramides influence brain function? We identified that CORT treatment or FMT from CORT-treated mice did not induce a significant rise in serum or hippocampal ceramides levels (Supplementary Fig. 10), indicating that ceramides might remotely influence mitochondrial function and hippocampal AHN in the local intestines without entering into circulation and passing through blood-brain barrier (BBB).

In this study, we found that: (1) chronic exposure to CORT led to depressive-like phenotypes and altered fecal microbial composition in mice; (2) FMT from CORT-treated mice transferred depressive-like phenotypes and their dominant microbiota to the recipient mice; (3) CORT-treated mice and Abx-FMT/C mice exhibited remarkably increased ceramides in fecal metabolites; (4) GM dysbiosis, specifically remarkable elevation of B. pseudolongum and L. reuteri, might influence hosts to produce ceramides; (5) ceramides treatment induced depressive-like phenotypes and dysregulation of DEGs responsible for mitochondrial function in the hippocampus. In conclusion, our study indicates that prolonged exposure to CORT leads to significant alterations in the microbial composition, which promotes the host to produce ceramides. This, in turn, hampers the critical mitochondrial function for hippocampal neural proliferation, ultimately leading to reduced AHN and consequent behavioral deficits (Fig. 5j).

It is well-documented that chronic stress exposure is a precipitating factor in the development of depression. Therefore, several animal models have been established to investigate the pathogenic link between chronic stress and depressive disorders, including chronic mild stress paradigms, inescapable stress, and social defeat stress. The hypothalamic pituitary adrenal (HPA) axis has been implicated in the pathogenesis of depression³³. Exposure to any type of stressor ultimately activates the stress-responsive HPA axis, thus leading to the subsequent increased production of cortisol. Accordingly, many research groups have employed long-term systemic CORT administration procedures to induce physiological and behavioral changes that recapitulate depression-like symptomatology^20,21,22. Consistent with previous literature, the present study demonstrated that chronic exposure to CORT led to depressive-like phenotypes, such as increased immobility time in the TST and FST (Fig. 1b, c) and reduced hippocampal neurogenesis (Fig. 1d, e).

According to the data derived from 16 s rRNA sequencing (Figs. 1 and 2) and metagenomic sequencing (Fig. 4), CORT-treated mice or Abx-FMT/C mice showed a loss of microbial diversity and altered microbial composition. Consistent with the present study, a previous study reported a modified composition of GM in mice following administration of CORT (subcutaneous injection, 20 mg/kg, 28 days)³⁴. Moreover, several clinical studies have shown severely altered GM composition in patients exposed to high levels of glucocorticoids (GCs), such as Cushing’s syndrome (endogenous GCs)³⁵ or when using large doses of exogenous GCs³⁶. We then identified an increased abundance of Lactobacillales and Bifidobacteriales in CORT-treated and Abx-FMT/C mice (Figs. 1, 2 and 4). This differs from a previous study, which reported that significantly increased abundances of Odoribacter and Ruminococcus occurred in response to long-term CORT treatment³⁴. This difference may be related to different CORT treatment procedures and different locations of samples collected for microbiota analysis (in the previous study, the ___location was cecal, while in this study, the ___location was the distal rectum), which may have affected the specific GM taxa. Several studies have reported that microbiota from cecal contents and feces are extraordinarily different in terms of composition and functional activity in humans and mice^37,38.

Recently, the role of the microbiota-gut-brain axis in the pathogenesis of depression has received widespread attention. Metabolites produced in the gut can be absorbed in the bloodstream, enter the circulatory system, and reach distant organs such as the brain. The production of metabolites with neuroactive and immunomodulatory properties by GM has been shown to affect brain function and mental health³⁹. Ceramides, the central hub of sphingolipid metabolism, were found to be upregulated in CORT-treated and Abx-FMT/C mice (Fig. 3f–i); and administration of a ceramide mixture via oral gavage-induced depressive-like behavior and reduced hippocampal neurogenesis in mice (Fig. 5a–f). Our study is consistent with previous animal studies: (1) In 2013, Gulbins et al. reported that chronic unpredictable stress (CUS) or administration of CORT led to increased levels of hippocampal ceramide and reduced hippocampal neurogenesis, and injection of C16 ceramide into the hippocampus induced depression-like symptoms in WT mice⁴⁰; (2) in 2022, it was reported that increased concentrations of ceramide-enriched exosomes in the blood plasma of CUS mice or CORT-treated mice and intravenous injection of ceramide-loaded exosomes recapitulated depressive-like behaviors in WT mice⁴¹. Although pharmacological evidence for the role of ceramides in depression is limited, several clinical evidences suggest that ceramides might contribute to depression pathophysiology: (1) Gracia-Garcia et al. reported that blood ceramides were elevated in Alzheimer’s disease (AD) patients with depressive disorders when compared to non-depressed AD patients²⁸; (2) Xing et al. also reported a positive correlation between plasma ceramides and depressive symptoms in AD patients⁴²; (3) in a study involving patients with Parkinson’s disease (PD), ceramides were reported to be positively associated with the severity of depressive symptoms as measured by the Geriatric Depression Scale (GDS) scores⁴³; (4) Schumacher et al. reported that the concentration of blood ceramides was positively correlated with the severity of depressive symptoms in MDD patients, as determined by the Hamilton Depression Rating Scale (HAM-D)⁴⁴.

The elevation of C14 ceramide observed across different experimental groups (Fig. 4h, i; see also Supplementary Fig. 5) suggests its potential role as a crucial pathological mediator contributing to the alterations of animals. In a previous study by Xing et al.⁴⁵, the upregulation of C14 ceramide level was significantly higher in patients with Parkinson’s disease who exhibited decreased memory function. Additionally, the C14 upregulation was associated with higher odds of multiple neuropsychiatric symptoms, including hallucinations, anxiety, and sleep behavior disturbances, compared to normal controls⁴⁵. Therefore, further study should be conducted to validate the role of C14 ceramide for the histological and behavioral alterations noted in the animal models. It is worth noting that the ceramide levels in the ceramides-treated group does not mimic the ceramide levels in CORT-treated group compared to their controls (Supplementary Fig. 5), the differences can be attributed to the mechanisms underlying ceramide production in these two treatment groups: CORT treatment leads to increased gut ceramides due to dysbiosis of the gut microbiota, which induces the host to produce more endogenous ceramides; while the increased ceramides levels in ceramides-treated group reflect the exogenous ceramides introduced into the intestines by oral gavage. Although both groups show increased levels of ceramides compared to their respective controls, the underlying mechanisms are different, leading to potentially distinct trends.

Although existing evidence suggests a correlation between plasma ceramide levels and the severity of depression^28,42,43,44, CORT treatment or FMT from CORT-treated mice did not increase ceramide levels in both the hippocampus and blood of the mice (Supplementary Fig. 10), indicating that gut ceramides in the local intestine may remotely influence brain function without directly entering circulation or passing through the BBB. Actually, there is substantial evidence supporting the communication between the gut and brain through various pathways that do not necessarily involve direct entry into circulation: (1) a recent study reported that administration of propionate in drinking water prevented motor impairments and dopaminergic neuronal loss in 6-OHDA-induced PD mice. Interestingly, the intervention with propionate resulted in a significant increase in fecal propionate levels, while serum propionate levels remained unchanged⁴⁶. The observation underscores the complex and intricate communication between the gut and the brain, even in instances where direct entering into circulation and penetration through the BBB may not occur. (2) Vagus nerve serves as a vital neural pathway connecting the gastrointestinal tract to the brain. Recent research has demonstrated that vagal afferents form synaptic connections with enteroendocrine cells in the gut, allowing for the communication of signals from various bacterial metabolites directly to the brain via glutamatergic neurotransmission⁴⁷. Moreover, vagal nerve fibers are equipped with receptors such as 5-HT3, Toll-like receptor 4 (TLR4), free fatty acid receptors, and gut peptide receptors. These receptors enable the transmission of signals from the gut lumen to the brain, highlighting the potential for local gut metabolites to influence brain function through direct synaptic connections or receptor activation within the vagus nerve in the intestines⁴⁸.

Finally, the possible underlying mechanism by which ceramides induce depression was explored. Hippocampal RNA-seq data revealed the interactions between OXPHOS-related genes and proliferation-related genes (Fig. 5h, i, see also Supplementary Fig. 6), indicating a correlation between aberrant mitochondrial function of the hippocampus and reduced hippocampal neurogenesis. OXPHOS provides the majority of ATP and supports the synthesis of molecular building blocks for cell proliferation^49,50,51. In addition, accumulated evidence indicated a correlation between aberrant OXPHOS pathway and depression: Zubenko et al. reported a dysregulated OXPHOS pathway identified in the hippocampus of a mutant mouse model of depression⁵²; several proteomic studies using different animal models of depression also identified altered levels of specific proteins involved in OXPHOS pathway⁵³; more importantly, a clinical study identified dysregulated levels of the OXPHOS complex in post-mortem brains from depressed patients⁵⁴. Mitochondrial OXPHOS is essential for cellular energy metabolism and is crucial for cell proliferation, survival, differentiation, and metabolism. Neurons display a high rate of ATP consumption compared to other cell types; therefore, they are very sensitive to energy limitations. Therefore, it is reasonable to speculate that ceramide treatment induces mitochondrial OXPHOS dysregulation, leading to decreased hippocampal neurogenesis and consequent depression-like behavior.

We found an upregulation of ATP in the hippocampal tissue of CORT-treated mice, whereas ATP levels were downregulated in the Abx-FMT/C and ceramide-treated mice (Supplementary Fig. 8). This discrepancy might be explained by the fact that CORT-induced upregulation of glial ATP release³² could potentially mask the reduced ATP production by neurons caused by ceramide accumulation, leading to the observed discrepancy. Although our data showed an overall increased hippocampus ATP production in CORT-treated mice, this does not negate the fact that specific neuronal ATP deficiency may still occur, particularly in response to ceramide accumulation. These deficits, even if not reflected in the overall tissue ATP levels, are sufficient to impair neuronal proliferation. However, the causal relationship between mitochondrial dysfunction and ATP production is complex and warrants further investigation. Identifying a hub gene critical for mitochondrial function and genetically manipulating this hub gene could indeed shed more light on the key mechanisms involved. However, pinpointing this hub gene is a significant challenge and represents a limitation of our work. We are committed to addressing this in our future research efforts, where we will aim to unravel the intricate mechanisms governing mitochondrial function and its impact on neuronal proliferation.

Housing of mice and CORT-treatment experiment

Male C57BL6/J mice (age: 6–8 weeks; weight: 23–25 g) purchased from Charles River Laboratories (Beijing, China) were maintained in a specific pathogen-free animal facility under standard conditions (room temperature: 24 °C; humidity: 55%) under a 12:12 light/dark cycle. All animals were provided with free access to water and food. All the interventions and animal care procedures were approved by the Animal Committee of the School of Life Sciences and Technology, Tongji University, Shanghai, China.

The CORT treatment experiment was performed as previously described⁵⁵. Briefly, two groups of weight-matched mice were administered CORT (40 mg/kg) or vehicle (23.07 mL/kg) per day via subcutaneous (s.c.) injection for 36 days.

Myriocin (1 mg) was dissolved in 2.5 ml DMSO as a storage solution, and then 22.5 ml PBS was added to obtain a working solution (0.04 mg/ml). Then, the mice received intraperitoneal injection of myriocin (0.4 mg/kg) every other day for 4 weeks⁵⁶.

Behavioral assessments

Anxiety and depression-like behaviors were evaluated as previously described⁵⁷. The forced swimming test and tail suspension test (TST) were used to assess depression levels; the open field test and elevated plus maze test were used to assess anxiety levels. All analyses were performed under blinded conditions.

The OFT was performed in an apparatus (40 × 40 × 30 cm) made of gray plastic. The test was conducted during the same period of day. Each mouse was placed in the center zone of the arena facing the wall and allowed to explore the entire arena freely. Smart software (Harvard, US) was utilized to analyze the total distance traveled in the whole arena and distance traveled/time spended/entries in the center zone for 10 min.

The elevated plus-maze apparatus comprised two open arms (30 × 5 × 16 cm) perpendicular to two enclosed arms (30 × 5 × 16 cm) with a center platform (5 × 5 cm). Each mouse was placed in the central area with its head directed toward an open arm and allowed to explore the apparatus freely for 10 min. The time spent and distance traveled in the open arms were analyzed using Smart software (Harvard, US).

For the TST, each mouse was suspended on the shelf 50 cm above the top of the table using a piece of tape placed 1 cm from the tip of the tail. The test lasted for 6 min, and the immobility time was recorded and analyzed using Smart software (Harvard, US).

For forced swimming test (FST), each mouse was individually placed in a cylindrical container (30 cm height × 15 cm diameter) filled with water (15 cm deep) at 25 °C. Mobility in the FST was defined as any movement other than that required to balance the body and keep the head above the water. The test length was 6 min, and only the last 4 min of the test was analyzed using Smart software (Harvard, US).

In vivo proliferation assay

The in vivo proliferation assay was performed as previously described²⁰ with some modifications. To visualize proliferative cells, mice were intraperitoneally injected with 5 mg/kg EdU dissolved in sterilized (phosphate-buffered saline) PBS per day for 7 days and then sacrificed 24 h later for histological analysis. Immediately following standard cardiac perfusion with 4% PFA, the brains were isolated and dehydrated with 30% sucrose. Free-floating brain sections were prepared as previously described⁵⁸. Brain sections were stained for EdU using the Click-iT Alexa Fluor 488 Imaging Kit (Thermo Fisher Scientific, C10337), and EdU-labeled cells in the DG of the hippocampus were counted using ImageJ software.

Depletion of recipient GM and FMT

Recipient mice were administered a combination of antibiotics (100 μg/ml neomycin, 50 μg/ml streptomycin, 100 U/ml penicillin, 50 μg/ml vancomycin, 100 μg/ml metronidazole, 1 μg/ml bacitracin, 125 μg/ml ciprofloxacin, 100 μg/ml ceftazidime, and 170 μg/ml gentamycin) in distilled drinking water for 4 weeks as previously described⁵⁹. On the day of FMT, fresh feces were collected directly from the rectum of donor mice, and fecal pellets (100 mg) were diluted in 1 ml PBS to obtain a fecal suspension. The suspension was then filtered and centrifuged to remove sizable particles and fibrous matter. The fecal suspension (200 μL) was then administered into the stomachs of mice via oral gastric gavage within 1 min. Recipient mice received two rounds of FMT on alternate days. The recipient mice were then left in donor mouse cages until behavioral testing.

16S rRNA sequencing-based microbial profiling

Fresh feces (50 mg/mouse) were collected and sent to Majorbio Bio-pharm Technology Co., Ltd (Shanghai, China) for DNA extraction and 16S rRNA sequencing. Briefly, Total microbial genomic DNA was extracted from fecal samples using the E.Z.N.A.® soil DNA Kit (Omega Bio-tek, Norcross, GA, U.S.) according to the manufacturer’s instructions. The quality and concentration of DNA were determined using 1.0% agarose gel electrophoresis and a NanoDrop® ND-2000 spectrophotometer (Thermo Scientific Inc., USA) and stored at -80 °C prior to use. The hypervariable region V3-V4 of the bacterial 16S rRNA gene was amplified with primer pairs 338 F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806 R (5′-GGACTACHVGGGTWTCTAAT-3′) using an ABI GeneAmp® 9700 PCR thermocycler (ABI, CA, USA). All samples were amplified in triplicates. The PCR product was extracted from a 2% agarose gel, purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer’s instructions, and quantified using a Quantus™ Fluorometer (Promega, USA).

Purified amplicons were pooled in equimolar amounts and paired-end sequenced on an Illumina MiSeq PE300 platform (Illumina, San Diego,USA), according to standard protocols. After demultiplexing, the resulting sequences were quality-filtered with fastp (0.19.6) and merged with FLASH (v1.2.11). The high-quality sequences were de-noised using the DADA2 plugin in the Qiime2 (version 2020.2) pipeline with the recommended parameters, which obtains single-nucleotide resolution based on error profiles within samples. DADA2 de-noised sequences are usually called amplicon sequence variants (ASVs). To minimize the effects of sequencing depth on alpha and beta-diversity measures, the number of sequences from each sample was rarefied to 20,000, which yielded an average Good’s coverage of 97.90%. Taxonomic assignment of ASVs was performed using the Naive Bayes consensus taxonomy classifier implemented in Qiime2 and SILVA 16S rRNA database (v138). Bioinformatics analysis of GM was performed using the Majorbio Cloud platform (https://cloud.majorbio.com).

Non-targeted liquid chromatography with tandem mass spectrometry (LC-MS/MS) based metabolomics analysis

Fresh feces (100 mg/mouse) were collected and sent to the company (Majorbio Bio-pharm Technology Co., Ltd, Shanghai, China) for metabolomic analysis. Briefly, 50 mg of the solid sample was added to the extraction solution (methanol: water = 4:1, 400 μL) containing 0.02 mg/mL internal standard (L-2-chlorophenylalanine), followed by low-temperature ultrasonic extraction and centrifugation. The supernatant was then transferred to a sample vial for LC-MS/MS analysis. The analysis was conducted on a SCIEX UPLC-Triple TOF 5600 system equipped with an ACQUITY HSS T3 column (100 mm × 2.1 mm, 1.8 μm; Waters, USA), with gradient elution of solvent A (0.1% formic acid in water: acetonitrile; 95:5; v/v) versus solvent B (0.1% formic acid in acetonitrile: isopropanol: water; 47.5:47.5:5; v/v/v). The flow rate was 0.40 mL/min, and the column temperature was set at 40 °C. The UPLC system was coupled to a quadrupole time-of-flight mass spectrometer (Triple TOFTM5600+, Sciex, USA) equipped with an electrospray ionization (ESI) source operating in the positive and negative modes. The optimal conditions were set as follows: source temperature at 550 °C; ion spray voltage floating at −4000 V in negative mode and 5000 V in positive mode; declustering potential (DP) at 80 V; collision energy (CE) at 20–60 eV rolling for MS/MS. Data acquisition was performed using information-dependent acquisition (IDA) mode. Detection was performed over a mass range of 50−1000 m/z. Pretreatment of LC-MS/MS raw data was performed using Progenesis QI (Waters Corporation, Milford, USA) software, and three-dimensional data in CSV format were exported. The data were analyzed using the free online Majorbio Cloud platform (cloud.majorbio.com).

Metagenomic sequencing

Fresh feces (100 mg/mouse) were collected and sent to Majorbio Bio-pharm Technology Co., Ltd (Shanghai, China) for metagenomic sequencing. Briefly, total genomic DNA was extracted from fecal samples using the E.Z.N.A.® Soil DNA Kit (Omega Bio-tek, Norcross, GA, U.S.) according to the manufacturer’s instructions. The DNA extract was fragmented to an average size of approximately 400 bp using a Covaris M220 (Gene Company Limited, China) for paired-end library construction. Paired-end sequencing was performed on an Illumina NovaSeq 6000 (Illumina Inc., San Diego, CA, USA) using NovaSeq Reagent Kits according to the manufacturer’s instructions. Briefly, paired-end Illumina reads were trimmed, and low-quality reads were removed using fastp. Reads were aligned to the mouse genome using BWA (http://bio-bwa.sourceforge.net, version 0.7.9a), and any hits associated with the reads and their mated reads were removed. The data were analyzed on the free online Majorbio Cloud Platform (www.majorbio.com).

Transcriptomic analysis by RNA-sequencing

The hippocampi of the mice were isolated, and the frozen tissues (50 mg/sample) were subjected to RNA extraction and RNA-seq analysis (Majorbio Bio-pharm Technology Co., Ltd, Shanghai, China). Briefly, total RNA was extracted from different groups in triplicate. The RNA-seq transcriptome library was prepared with 1 μg of total RNA using the TruSeqTM RNA sample preparation Kit from Illumina (San Diego, CA, USA). Then, the paired-end RNA-seq library was sequenced using the Illumina HiSeq xten/NovaSeq 6000 sequencer (2 × 150 bp read length). Raw paired-end reads were trimmed, and quality was controlled using SeqPrep (https://github.com/jstjohn/SeqPrep) and Sickle (https://github.com/najoshi/sickle), followed by alignment to the reference genome in orientation mode using HISAT2 software. The mapped reads of each sample were assembled using StringTie with a reference-based approach. The expression level of each transcript was calculated according to the transcript per million reads (TPM) method. The TPM values for each DEG are shown in the supplementary files. Transcript abundance was quantified using the RSEM software tool. DEGs analysis was performed using DESeq2 with an adjusted P-value ≤ 0.05.

Administration with CeM by gastric gavage

A 100 mg ceramide mixture (Topscience, T10760) was dissolved in 80 ml of a vehicle containing 5% Tween-80 and 0.5% sodium carboxymethyl cellulose to reach a final concentration of 1.25 mg/ml. Mice were administered a ceramide mixture (Cer-M, 10 mg/kg) or vehicle of ceramides (Veh/C, 8 ml/kg) daily for 36 days via gastric gavage. The dose of ceramide administration (10 mg/kg) was based on well-established protocols^60,61.

mRNA extraction and qPCR

Total RNA was extracted using SteadyPure quick RNA extraction kit (Accurate Biology, China) and diluted in RNAase-free water. One μg of RNA was reverse transcribed to cDNA using HiFiScript gDNA removal RT MasterMix (CW BIO, China). qPCR was performed using the SYBR Green Mix (Bio-Rad, USA) in a Bio-Rad CFX96 Thermal Cycler (Bio-Rad, USA). Housekeeping gene β-actin was used to normalize mRNA levels between different samples. Primer sequences are listed in Supplementary Table 1.

Hippocampal mitochondrial isolation and JC1 assay for isolated mitochondria

Mitochondrial isolation and JC1 assay were performed as previously described elsewhere⁶² with some modifications. Briefly, hippocampal mitochondria were isolated using a commercial mitochondrial isolation kit according to the manufacturer’s instructions. Briefly, fresh hippocampal tissues were homogenized in 400 μL of ice-cold mitochondria isolation buffer A using a low-temperature homogenizer. The homogenate was centrifuged (1000 × g, 5 min, 4 °C), the subsequent supernatant was transferred to another tube and centrifuged (3500 g, 10 min, 4 °C). The supernatant was discarded and the mitochondrial pellet was resuspended in 200 μL mitochondrial storage buffer, then the protein concentration was determined by BCA assay kit.

Mitochondrial membrane potential (ΔΨm) was determined to evaluate mitochondrial function using JC1 assay kit. Briefly, mitochondrial suspension in 100 μL of storage buffer was mixed with 10 μL of JC1; CCCP (final concentration: 50 μM), a well-established mitochondrial membrane potential disrupter, was added to the sample as negative control. After incubation for 15 min at 37 °C, the fluorescent intensity was measured by microplate reader. The ΔΨm was determined by the ratio of red (aggregated) (λ_ex = 525 nm, λ_em = 590 nm)/green (monomer) (λ_ex = 490 nm, λ_em = 530 nm) and normalized to the protein concentrations.

ATP measurement

Hippocampal ATP was determined using enhanced ATP assay kit according to the manufacturer’s protocol. Briefly, mice were euthanized and fresh hippocampal tissues were collected. The hippocampal tissues were homogenized in the presence of 400 μl of lysis buffer using a low-temperature homogenizer. Subsequently, the homogenate was centrifuged (12,000 × g, 5 min, 4 °C) and the supernatant was collected, and the protein concentration of each sample was determined by BCA assay kit. 100 μl of ATP detection working solution was added into the well and incubated for 3–5 minutes (room temperature, RT) to ensure complete consumption of endogenous ATP. Subsequently, 20 μl of the sample was added into the detection well, and the relative light unit value was measured using a luminometer. The ATP level was calculated using the ATP standard curve. For quantification, ATP levels were normalized to protein amounts of each sample.

Quantification of ceramides by using LC-MS/MS-based targeted lipidomics analysis

LC-MS/MS-based targeted lipidomics analysis allows for rapid identification and quantification of a broad spectrum of lipid molecular species, such as ceramides. Therefore, the samples (hippocampus, feces, serum) were flash frozen in liquid nitrogen and then sent to the company (APTBIO Co., Ltd, Shanghai, China) for targeted lipidomics analysis. The information of the types of detectable ceramides, dihydroceramide, and SM and their standard references are listed in Supplementary Table 2.

Lipids were extracted according to MTBE method. Briefly, samples were mixed with 200 μL methanol, and then 10 μL internal lipid standards and 800 μL MTBE were added. The mixture was adequately vortexed, sonicated for 20 min at 4 °C and then kept for 30 min at RT. Subsequently, 200 μL of MS-grade water was added, and the mixture was vortexed and centrifuged (14,000 rpm, 15 min, 4 °C). The upper organic solvent layer was obtained and dried under nitrogen. Subsequently, the samples were re-dissolved in 200 μL of IPA/ACN (9:1, v/v) solvent and centrifuged (14,000 rpm, 4 °C, 15 min), then the supernatant was injected.

The analysis was performed on a UHPLC system (LC-30AD, Shimadzu) coupled with QTRAP MS (6500+, Sciex). The analytes were separated on HILIC (Phenomenex, Luna NH2, 2.0 mm × 100 mm, 3 µm) and C18 column (Phenomenex, Kinetex C18, 2.1 × 100 mm, 2.6 µm). For C18 separation, the column temperature was set at 45 °C. Mobile phase A: 70% acetonitrile+30% H2O + 5 mM ammonium acetate, mobile phase B: IPA solution. A gradient (20% B at 0 min, 60% B at 5 min, 100% B at 13 min, 20% B at 13.1-17 min) was then initiated at a flow rate of 0.35 mL/min. The sample was placed at 10 °C during the whole analysis process.

For Amino separation, the column temperature was set at 40 °C. Mobile phase A: 2 mM ammonium acetate+50% methanol+50% acetonitrile; mobile phase B: 2 mM ammonium acetate +50% acetonitrile +50% water. A gradient (3% B at 0-3 min, from 3 to 100% B at 3-13 min, 100% B at 13–17 min, 3% B at 17.1 -22 min) was then initiated at a flow rate of 400 uL/min. 6500 + QTRAP (AB SCIEX) was performed in positive and negative switch mode. The ESI positive source conditions were as follows: Source temperature: 400 °C; Ion Source Gas1 (GS1): 50 Ion Source Gas2 (GS2): 55; Curtain Gas (CUR): 35; IonSpray Voltage(IS): +3000 V; the ESI negative source conditions were as follows: Source temperature: 400°C; Ion Source Gas1 (GS1): 50; Ion Source Gas2 (GS2): 55; Curtain gas (CUR): 35; IonSpray Voltage (IS): −2500 V. MRM method was used for mass spectrometry quantitative data acquisition. A pooled quality control (QC) samples were set in the sample queue to evaluate the stability and repeatability of the system. Sciex OS was used for quantitative data processing. The QCs were processed together with the biological samples. Metabolites in QCs with coefficient of variation (CV) less than 30% were denoted as reproducible measurements.

c-Fos induction and immunohistochemistry

In order to identify brain areas exhibiting remarkable changes following prolonged CORT treatment, c-Fos expression was detected. c-Fos, an immediate early gene, was initiated in activated neurons in response to various stimuli. c-Fos expression begins within 15 minutes of stimulation, and the c-Fos protein expression reaches peak between 90 and 120 min following stimulation⁶³. c-Fos detection was performed as previously described elsewhere⁶⁴ with some modifications. Briefly, FST (aversive stimulation) was performed to induce c-Fos expression, and the mice were sacrificed 1 hour after stimulation to collect the brain samples. Free-floating brain sections were prepared for subsequent c-Fos immunohistochemistry. Detailed information for antibodies can be found in Supplementary file (Supplementary Table 3). Imaging was performed using Olympus IX73 microscope system. The number of c-Fos-positive cells was counted using ImageJ software.

Maintenance and colonization of bacterial strains

L. reuteri (ATCC #23272) and B. pseudolongum (BNCC #135158) were cultured in MRS medium (pH 5.7 ± 0.2) at 37 °C in a vertical anaerobic culture bag.

To evaluate the bacterial capability to produce ceramides, B. pseudolongum or L. reuteri were cultured in MRS culture medium supplemented with 25 µM palmitic acid⁶⁵, respectively. The cultures were incubated anaerobically at 37 °C. Following 24 hours of incubation, the supernatant was collected and subjected to targeted lipidomics analysis.

Bacteria used in mouse experiments were cultured overnight anaerobically at 37 °C in MRS medium, and administered to individual mice via intragastric gavage. Abx mice received oral gavage of 2 × 10⁹ CFU of B. pseudolongum or L. reuteri in 200 μL of sterile anaerobic PBS daily for seven days. For gavage experiment, the bacteria were washed and concentrated in anaerobic PBS to a final concentration of 1 × 10⁹ CFU/mL under strict anaerobic conditions. 24 hours after the last gastric gavage experiment, the feces of mice were collected for LC-MS/MS-based lipidomics analysis.

Statistical analysis

Statistical analyses were performed using GraphPad PRISM 8 (GraphPad Software, San Diego, CA, USA). For experiments with larger sample sizes (N > 5), such as the behavioral experiments and LC-MS/MS-based metabolomics analysis, the Kolmogorov–Smirnov test confirmed the normal distribution of the data, justifying the use of an unpaired t-test for comparisons. For experiments with smaller sample sizes (n ≤ 5), such as in vivo proliferation assay, c-Fos detection, ATP and JC1 detection, the sample size is too small to validate the distribution of the data, so the nonparametric Mann-Whitney U test is selected to compare the differences. We included information about the sample sizes in the figure legends. For 16S rRNA sequencing-based microbial profiling and metagenomic sequencing, comparisons were performed using the Welch’s t test. For 16S rRNA-based MiSeq sequencing of this study, we identified an outlier that deviated significantly from its respective group (CORT-treated mice). Given the variability inherent in animal studies, such outliers can occasionally arise and potentially skew the overall clustering results. Because no universally accepted statistical method is available for identifying and removing outliers in this specific context, we decided to remove this outlier and reanalyze the whole data. This decision was based on the fact that the removal of the sample did not affect the final outcomes of our analysis but helped prevent any potential misinterpretation of the data. All data are expressed as mean ± standard error of the mean. *p < 0.05, **p < 0.01, and ***p < 0.001 were considered to be significant.