Microbiome and tryptophan metabolomics analysis in adolescent depression: roles of the gut microbiota in the regulation of tryptophan-derived neurotransmitters and behaviors in human and mice

The depressive patients were recruited from Nanjing Brain Hospital affiliated to Nanjing Medical University (Jiangsu, China) from September 2020 to May 2021. Eligible female subjects (aged 11 to 17 years) were diagnosed by Diagnostic and Statistical Manual of Mental Disorders 5 (DSM-V) criteria and their symptom severity was quantified by Revised Child Anxiety and Depression Scale-25 (RCADS-25) [15]. In addition, depressive participants (n = 25) were excluded by parental history of bipolar disorder, schizophrenia or other psychiatric disorders according to DSM-V, meanwhile, subjects suffering from chronic systematic diseases, including diabetes, cardiovascular disease, thyroid disease, cancer etc., as well as substance abuse and infection records of bacteria, fungi or virus were also precluded. The healthy controls (HC, n = 10) were enrolled based on matching criteria of gender and other demographic information. The baseline demographic information of all participants was collected by the Child Mental Health Questionnaire. No candidates showed other major psychiatric comorbidities and all of them were drug-naive, without using antibiotics, anti-inflammatory medicine, prebiotics, or anti-depressants in the last 2 months. After samples (serum, urine, and feces) collection, the depressive subjects were further treated with sertraline for clinical normal course, and then evaluated by RCADS-25 for the second time. Subjects with lower scores, compared with the first time, were included in the recurrence group (DEP-sertraline treated). The protocols of clinical research were reviewed and approved by ECNMU (Approval No. 2020-KY198-01). All participants provided written informed consent prior to the study.

Female C57/6 J mice (3 weeks old) were obtained from National Rodent Laboratory Animal Seed Center (Shanghai, China) and were maintained in the specific-pathogen-free level (SPF) environment with controlled room temperature (RT, 21–23 ℃), humidity (50–60%) and 12 h/12 h light/dark cycle. All mice had ad libitum access to dry food pellets and water and were left acclimatized for 1 week prior to any procedures. All the rodent experiments were approved by the Institutional Animal Care and Use Committee of Nanjing Medical University (Approval No. IACUC-2005051).

Roseburia intestinalis. (Ri., DSMZ 14610) strain was revitalized and cultured in YCFA broth (DSMZ Medium 1611) containing 5 mg Na-resazurin per liter (oxidation–reduction indicator). Sodium carbonate was added to the medium to adjust pH to 6.8 and the liquid medium was pre-reduced under anaerobic conditions for 24 h before culturing. The bacteria strain was then incubated in an anaerobic environment (gas atmosphere N₂: CO₂: H₂ of 80:10:10) at 37 ℃ for 48–72 h. For the animal experiment, Ri. cells were collected by centrifugation (5000 rpm, 5 min, RT) and washed twice with sterile anaerobic PBS, finally re-suspended in 1 × PBS (OD₆₀₀ = 0.2) and stored at 4 ℃ 1 day before mice gavage.

According to our previous study [14], mice exposed to CRS were placed in the 50 mL restraint cylinders fitted closely to body size with holes which allowed free breathing for 14 consecutive days (3–4 h per day). The control (CTR) and CRS mice were then examined by the first-round behavioral tests in order to assess depression- and anxiety-like changes, following by random assignment with 4 sub-groups for the FMT or target bacteria intervention.

Each of the experimental set (FMT set or target bacteria intervention set) contained two periods of behavioral tests. The first one was conducted after 2-week CRS modeling, while the other one was performed after microbiota transplantation or Ri. intervention. Mice were transferred to the testing room 3–4 h before all behavioral tests for adaption. In the spontaneous behavior experiments (open field test and elevated plus maze test), each mouse was gently placed in the cage frame or hands of the testers and was allowed free movement before formal tests. In the tail suspension test, open field test and elevated plus maze test, equipment was cleaned with 75% ethanol after every trial in order to minimize scents of the former test object. All the mice testing order was random.

The SPT was performed based on the previous study [14]. Mice anhedonia was described by the consumption ratio of sucrose in this assay. After the last CRS, mice were housed individually and were acclimatized with two 50 mL drinking tubes for the first 24 h, both containing 1%(w/v) sucrose solution. In the next 24 h, one tube of the solution was replaced with drinking water and during the ensuing test period (24 h), mice were exposed to tubes with sucrose solution and water. The tube position was reversed every 12 h in order to avoid place preference. The liquid consumption was measured, and the sucrose preference was calculated as the fraction of the sucrose solution compared to the total drinking liquid amount.

The OFT was implemented to evaluate mice spontaneous activity and anxiety-like behavior 1 day after SPT. Mice were placed in the center arena of a plastic open-field apparatus, consisting of one 50 × 50 cm base and four 50 × 30 cm walls, and were allowed to freely explore for 6 min. Time spent in different areas was recorded by ANY-maze software (Stoelting Co., USA).

Exploratory activity and anxiety-like behavior were measured in the elevated plus maze, which was 70 cm above the floor and composed of one central platform (5 × 5 cm), two open arms (30 × 5 cm) and two closed arms (30 × 5 × 20 cm). The closed ones were enclosed by non-transparent walls height in 20 cm. Mice were placed carefully in the center area and moved voluntarily for 6 min. Movement was recorded by ANY-maze software (Stoelting Co., USA).

Each mouse was suspended with adhesive tape affixed 1–2 cm to the tip of the tail and inverted for 6 min, with its nose 20–25 cm away from the ground. Immobile state was considered as no body movement and passive hanging. The ratio of immobility was quantified by Tail Suspension-scan software (Geneandi Co., China) over the observation period to assess the behavioral despair.

Mice were individually placed in one plexiglass cylinder (diameter 30 cm, height 50 cm) filled with 30 cm deep water (23–25 ℃) for 6 min and videotaped in every session using Superfst software (Soft-maze software Co., China). Immobility was defined as the absence of all motion except for movements required to keep mouse’s head above the water.

According to Feng Z.’s protocol with some modifications [16], mice distributed into FMT section experienced antibiotics cocktail (ABX; vancomycin 50 mg/kg/day, neomycin 100 mg/kg/day, metronidazole 100 mg/kg/day, ampicillin 100 mg/kg/day, amphotericin-B 1 mg/kg/day) treatment for 1 week before microbiota inoculation. Ablation of endogenous gut flora was evaluated by aerobic/anaerobic plating on TSA plates according to Sun’s procedures [17]. Fecal samples harvested from the HC and DEP subjects were suspended with sterile phosphate buffer saline (PBS) at a dilution volume ratio of 1:10 (w/v, feces/PBS). Before centrifugation (1000 rpm, 5 min, 4 ℃), the solution was vortexed for further homogenate. Then the supernate was transferred and another centrifugation (8000 rpm, 5 min, 4 ℃) was applied for the bacterial precipitates. Mixtures were re-suspended by equivalent volume of sterile PBS for the transplantation protocol.

For the FMT part, every single mouse in transferring groups (CRS + HC-tr and CRS + DEP-tr) received 200 μL fecal suspension once every 2 days for 14 days by oral gavage, while the remaining groups (CTR and CRS) mice were administered by equivalent volume of sterile PBS simultaneously.

For the target bacteria intervention, mice were assigned to 4 groups (CTR + PBS, CTR + Ri, CRS + PBS and CRS + Ri) and were orally administered by sterile PBS or Ri. suspension (200 μL/per mouse) once a day for 14 consecutive days.

Human fecal samples (2–3 g) were collected in sterile stool cups and were stored at -80℃ until further processing. The microbial DNA was extracted by the E.Z.N.A. Stool DNA kit (Omega inc., USA) and the total DNA was eluted in 50 μL of elution buffer, stored at -80℃ until measurement in the PCR (LC-Bio Technology Co., Ltd., China). The V3-V4 region of the prokaryotic 16S rRNA gene was amplified with 341F (5′-CCTACGGGNGGCWGCAG-3) and 805R (5′-GACTACHVGGGT ATCTAATCC-3) [18], with specific barcodes tagged in the 5′ end, using PCR conditions (initial denaturation at 98 ℃ for 30 s, 32 cycles of denaturation at 98 ℃ for 10 s, annealing at 54 ℃ for 30 s and extension at 72 ℃ for 45 s). The products were confirmed with 2% agarose gel electrophoresis and then purified with AMPure XT beads (Beckman Coulter Genomics, USA), finally quantified using Qubit (Invitrogen, USA). Following amplicon library assessment, the libraries were sequenced on Illumina NovaSeq PE250 platform. Pair-end reads was assigned to samples based on their unique barcode and truncated by cutting off the barcode and primer sequence, then further merged into the longer tags according to their overlap parts, using FLASH (version 1.2.8). Quality scanning on the raw reads was performed under specific filtering conditions to obtain the high-quality clean tags by fqtrim (version 0.94). Chimeric sequences were filtered using Vsearch software (version 2.3.4). After de-replication by DADA2, amplicon sequence variants (ASV) feature table and feature sequences were obtained. Relative abundance was analyzed by QIIME2 process. Based on this, the diversity evaluations were conducted, including alpha diversity and beta diversity analysis. As to the species annotation, we used NT-16S database of SILVA (Release 138, https://www.arb-silva.de/documentation/release138/↗) for annotation, according to ASV feature sequences (The confidence threshold for annotation is 0.7). Other diagrams were implemented using R package (version 3.5.2).

The prefrontal cortex (PFC; 30 mg) and clean colon tissues (2–3 cm) were collected in 2 mL screw cap microtubes containing 2 steel beads and 200–300 μL RIPA lysis buffer (25 mM Tris–HCl pH 7.6, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS; Sigma-Aldrich, USA), supplemented with protease inhibitor cocktail (Sigma-Aldrich, USA). Homogenization (4800 rpm, 10 s, 6 cycles) was performed, immediately followed by tissue lysis on ice for 20 min, and then the lysed mixture was centrifuged in 12,000 rpm for 15 min at 4 ℃. The supernate was transferred to 1.5 mL EP tube and the protein concentration was determined by Pierce BCA Protein Assay Kit (Thermo Scientific Rockford, USA). Thirty micrograms of tissue lysates boiled with 5 × loading buffer (Beyotime, China) were loaded onto 10% or 12.5% SDS–polyacrylamide gradient gel (Epizyme, China) and subsequently transferred onto PVDF membranes (Millipore, USA), using liquid transfer system (Bio-Rad, USA). The membranes were then probed with diluted primary antibodies overnight at 4 ℃, after 2 h blocking with 3% bovine serum albumin (BSA) in Tris-buffered saline and Tween 20 (TBST; 10 mM Tris-base pH 7.6, 150 mM NaCl and 0.1% (v/v) Tween 20) at RT. Following washed with TBST, bands were incubated with secondary antibodies in blocking buffer (1:10,000) for 2 h at RT. Visualization was then implemented by Chemiluminescent HRP Substrate (Millipore, USA) and quantification was performed using Fiji software (Image J; NIH, USA).

The KP metabolites were determined using ultra-high performance liquid chromatography tandem mass spectrometry (UHPLC-MS/MS). Serum and supernate of homogenized tissue were thawed, mixed with isotopically labeled internal standard solutions and then transferred to new tubes after vortex and centrifugation (29,700 g, 10 min, 4 ℃). The cryogenic vacuum centrifugal enrichment and evaporation system (CentriVap Labconco) was applied for further derivatization. 50 μL dansyl chloride solution (2 mg/mL, diluted with acetone) and sodium bicarbonate buffer (pH = 9.0, 0.2 M) was mixed at 1:1 volume ratio, mixture was added to the dry matter and vortexed for 1 min. After water-bath heating, evaporation, re-dissolution, and centrifugation, 5 μL supernatant was injected.

The kynurenine pathway metabolites were detected according to the manufacturer’s instructions. The ELISA kits including human 5-HT, DA, Trp, Kyn, Kyna, Quin, and NAD⁺ were used in the present work. Besides, parts of the KP metabolites of mice tissues were determined by ELISA kits. Data were normalized based on BCA determined protein concentrations.

The PFC and distal colon tissues were employed in this assay. Perfused tissues were overnight immersed in 4% paraformaldehyde at 4 ℃, then embedded in paraffin, cut into 3–4 μm thick sections. After dewaxing and antigen repairing, the slices were blocked with 5% BSA buffer for 2 h at RT, followed by overnight incubation with primary antibodies diluted in the blocking solution (F-actin 1:200, Drebrin 1:50, PSD95 1:500, Syn1 1:200, GFAP 1:1000, Iba1 1:500, Occludin 1:400) at 4 ℃. Washed 3 times, the slices were further incubated with secondary antibodies for 2 h at RT, shielded from light. Then, DAPI solution was applied before slices sealed with glycerol. Samples were imaged by confocal microscopy using a Zeiss Laser Scanning Inverted Microscope LSM-710 and images were exported and analyzed by Black ZEN software (Carl Zeiss, Germany) and Fiji, respectively.

The paraffin-embedded colon tissues were subjected to deparaffinization, rehydration and permeabilization successively. The probe used in this study was designed according to Hold’s research [19] with some modifications, diluted with pre-warmed hybridization buffer (1:200, 88 ℃) and then balanced for 5 min at 37 ℃ before applying. Slices were blocked with 0.2 N HCl and Proteinase K (50 μg/mL) for 15 min, RT and 30 min, 37 ℃ respectively, further incubated with diluted probe solution in the dark and humid chamber at 37 ℃ overnight. Followed by washing with pre-warmed washing buffer at 37 ℃ for 15 min, slices were air dried and then added with 20 μL DAPI-antifade solution in dark environment and incubated for 5 min. Confocal images were acquired and analyzed, and the probe sequence was displayed in Table S2.

Formalin-fixed and paraffin-embedded colon tissues were sliced into 4-μm-thick sections and stained with hematoxylin and eosin (H&E) for histopathological examination. Images were scanned using Pannoramic scanner (3DHISTECH, Germany), and then the histological scores were calculated.

Total RNA from mice colonic chunks (12 cm) was extracted using TRIZOL (Invitrogen, USA), while DNA from mice feces was isolated by Stool DNA extraction kit (TIANGEN, China). Chloroform and isopropanol were employed to precipitate the aqueous phase and RNA, or DNA concentration was examined by NanoDrop 2000 (Thermo Fisher, USA). cDNAs were generated from 500 ng total RNA with reverse transcription kit (Takara, Japan), followed by qPCR analysis using SYBR green kit (Takara, Japan). Primers sequences were listed in Table. S2

Data were analyzed using IBM SPSS 20.0 (Chicago, IL, USA) and GraphPad Prism 7 (San Diego, CA, USA), correlation images were created using R studio (Boston, MA, USA), and p ˂ 0.05 was considered statistically significant. Specific data significance values are reported in the figure legends, and most data are presented as mean ± SEM (except that the box chart shows the maximum and minimum values). Assumptions of normality and homogeneity of variance were checked using the Shapiro–Wilk and Levene tests, respectively. Significance between two independent groups that met the assumptions was determined using unpaired Student’s t tests, while those failed to meet the assumptions were examined using the Mann–Whitney U test. The chi-squared test was used for categorical data analysis. For multiple group comparisons, one-way ANOVA and two-way ANOVA followed by Tukey’s multiple comparison test were applied as appropriate. For the microbiome data analysis, due to its non-normal distribution, the Kruskal–Wallis test, along with Benjamini–Hochberg test were applied for multiple groups data analysis. Also, linear discriminant analysis effect size (LEfSe) analysis was used for comparing the differences in microflora composition of multiple groups. Receiver operating characteristic (ROC) curve was applied to depict the accuracy of targeted bacteria in distinguishing depressive adolescents and healthy controls. As to the type one error controlling, in the microbiota data analysis part, the Benjamini-Hochberg (original FDR method) was used for the multiple comparison among three groups. In the murine model data analysis part, the Tukey post hoc test was chosen due to its high testing efficiency in comparing equal or similar sample size data.

Healthy female controls (HC) and female unmedicated depressive (DEP) adolescents were enrolled in the present study, after the first diagnosis, the unmedicated patient samples were collected and the follow-up work was carried out for those patients who required sertraline treatment (DEP-sertraline-treated) until their depressive symptoms significantly improved, as evaluated by two attending psychiatrists or a clinician with a higher professional title. The HC adolescents were confirmed to be well matched with the depressive patients, as the majority of the demographic indicators were not significantly different between these two groups (Table). S1

To ascertain the association between deviations in gut microbiome composition and depression, the 16S ribosomal DNA sequencing was performed. As shown in Figure S1A, 272 ASV features accounting for the total richness were common to all the groups, whereas 881 and 2329 ASVs accounted for HC and DEP groups, respectively, along with 369 ASVs accounted for DEP-sertraline-treated samples. According to principal coordinate analysis (PCoA), gut microbiota (GM) from the HC, DEP and DEP-sertraline-treated group exhibited difference (Fig. 1A), albeit no significant alteration in alpha diversity among the three groups, as indicated by the Shannon, Simpson and Chao1 indices (Fig. 1B). The taxonomy results showed that the phylum Firmicutes was enriched in both HC and DEP-sertraline-treated subjects, meanwhile, the relative abundance of Actinobacteria, Proteobacteria, and Verrucomicrobia phyla increased in adolescents with depression when compared with the HC group, and this was partially reversed after sertraline use (Fig. 1C). Notably, at the genus level, Roseburia was more abundant in HC while decreased in unmedicated depressive adolescents and restored in DEP-sertraline-treated groups (Fig. 1D and E, Figure S1B). The specific bacterial genus showed fair sensitivity and specificity, considering that the area under curve (AUC) reached 0.7333 in the ROC curve (cut-off point value = 1.10303, sensitivity = 0.89, specificity = 0.56, Fig. 1F), the Roseburia may serve as a promising prediction in depression development and treatment.

Depression-like behavioral changes were highly associated with Trp-Kyn metabolism in the rodent model [14], thus, to fully unravel the alterations in Trp-derived metabolites (Trp, Kyn, kynurenic acid etc.) in adolescent depression, the neurotransmitters metabolized via the Trp pathway were examined. Compared with the unmedicated DEP group, the levels of Trp, 5-HT, and key KP metabolites, including Kyn and kynurenic acid (Kyna), were evidently increased in the serum and urine samples after the sertraline intervention, whereas Quin (serum and urine) and NAD⁺ (serum) levels were markedly suppressed (Fig. 2A, B, Figure S2B). To test whether the alterations in these neurotransmitters reflect the therapeutic effect of sertraline, a correlation analysis between the metabolite concentrations and RCADS-25 scores was conducted. In the serum, 5-HT, Kyn, and Kyna levels exhibited a pronounced negative association with self-rating scale scores (Fig. 3A), while the NAD⁺ level displayed a positive correlation (Figure S2C). In parallel with the results represented in the serum, a similar association between metabolite concentrations and RCADS-25 scores was observed in urine (Fig. 3B), suggesting that alterations in the targeted neurotransmitters specifically reflect the therapeutic effects of sertraline in adolescent depression.

To unravel the relationship among GM, KP metabolites, and adolescent depression, a human-to-mice FMT model was established (Fig. 4A) to determine whether an intestinal flora alteration can affect depressive behaviors by manipulating a host’s metabolism. In this study, 3-week-old female mice were used, corresponding to the higher prevalence of female adolescents in epidemiological studies [20]. The behavioral tests showed that the adolescent CRS mice exhibited obvious depression- and anxiety-like behaviors compared with the control (CTR) mice, with less consumption of sucrose solution in SPT and increased immobility states in both the TST and FST experiments. Meanwhile, the OFT results showed that CRS mice obviously spent less time in the center arena, and the EPM test data exhibited that they spent less time in the open arms, although the difference was not pronounced (Fig. 4B). These results demonstrated that depression-like behaviors in mice were induced by CRS. To assess whether the healthy human-to-mice FMT benefits CRS-induced depression, a gut microbiota depletion mouse model was constructed with antibiotic cocktail (ABX). Following 1-week ABX gavage and 2-week FMT, the depression-like behaviors in the SPT and TST of the adolescent CRS mice were markedly ameliorated after FMT from HC adolescents, and these changes were concomitant with relieved anxiety-like alterations in the OFT (Fig. 4C). In parallel, the gut microbiota transplanted from depressive adolescents showed no exacerbated effects on CRS-induced mice. Taken together, these results implied that some microbiota in the healthy adolescent feces may exert beneficial effects on CRS-induced depressive mice.

Neurotransmitter metabolism is strongly linked to depression- and anxiety-like behaviors, and Trp is a precursor of 5-HT and neuroactive KP metabolites [21]. Based on the depressive phenotype observed in the adolescent FMT rodent model aforementioned, the Trp-derived metabolic neurotransmitters were examined. As shown in Fig. 5A, B, a pronounced decrease in 5-HT in the PFC and colon was observed in the CRS mice, a process that was appreciably rescued following HC-GM transplantation. Reciprocally, Kyn levels were significantly elevated in the brains and guts of both CRS and CRS + DEP-tr mice, while microbiota from HC adolescents strikingly suppressed KP activation, as the levels of Kyn and its downstream neurotoxic metabolites, namely 3-HK and Quin, dramatically decreased (Fig. 5A, B). Similar results were also observed in the mouse serum, where the decreased levels of 5-HT were accompanied by increased levels of toxic KP products (3-HK, 3-HAA, and Quin), which were partially reversed by the HC microbiome transfer (Figure S3B). Moreover, the 5-HT pathway metabolites, including N-acetylserotonin (NAS) and melatonin (MLT), in the PFC were downregulated in CRS and CRS + DEP-tr mice, and these neurotransmitters were enhanced in the HC-GM treatment group (Figure S3A). Interestingly, the NAD⁺ levels increased in the CRS-treated mice, whereas the Trp levels showed no pronounced changes in the PFC, serum, and colon samples among the four groups (Fig. 5A, B and Figure S3B). Moreover, evidence for KP activation was also indicated in the CRS mice serum, as the metabolites of the KP, including Kyn, Kyna, 3-HK, 3-HAA, Quin, and NAD⁺ were significantly increased or displayed an increased tendency; synchronously, the 5-HT level in the serum was obviously reduced in the CRS-treated group (Figure S3B). To decipher the mechanisms of the neurotransmitter alterations mediated by CRS, the rate-limiting enzymes responsible for maintaining the balance of these neurotransmitters were dissected. As shown in Fig. 5C, D, the quantitative immunoblot assay displayed an evident decrease in tryptophan hydroxylase-2 (TPH2), an enzyme that plays a key role in 5-HT generation in the brain. Meanwhile, the levels of KP rate-limiting enzymes indole-amine oxygenase-1 (IDO1), 3-Hydroxyanthranilic acid 3, 4-dioxygenase (3HAO), and quinolinate phosphoribosyl transferase (QPRT) were markedly increased in both the PFC and colon samples from CRS and CRS + DEP-tr mice, and these effects could be rescued by HC-GM transplantation (Fig. 5C–E). Taken together, these results implied that some microbiota in the healthy adolescent feces modulate the rate-limiting enzymes, thereby, regulate the neurotransmitter metabolism both in the brain and colon.

The KP downstream metabolites, such as Kyna and Quin, play essential roles in synaptic plasticity, which are involved in the progression of adolescent depression [22]. Thus, the density of dendritic spines and synapses in the PFC was examined by double labeling with F-actin/Drebrin and Synapsin1 (Syn1)/Postsynaptic Density Protein95 (PSD95), respectively. As depicted in Fig. 6A–D, the HC fecal transplantation in adolescent CRS mice displayed more colocalization of F-actin and Drebrin, coupled with increased co-expression of Syn1 and PSD95 in the PFC. In accordance with the immunofluorescent results, CRS treatment remarkably decreased the protein levels of Drebrin, Syn1, and PSD95, whereas FMT from healthy adolescents substantially reversed these effects (Fig. 6E), suggesting that the normobiotic GM functions as a manipulator of synapse remodeling.

Microglial cells and astrocytes are vital sources of KP products in the brain [23]. Activated microglial cells are not only responsible for immune homeostasis in the brain, but also produce neurotoxic KP metabolites, such as Quin [24], under conditions of inflammation, which likely aggravates depressive symptoms in adolescents [25]. Interestingly, the activity undertaken in astrocytes, to some extent, exhibits neuroprotective effects by assuming the generation of Kyna, an effective neuroprotective agent in the case of excitotoxicity [26]. In this study, we evaluated the activities of both microglia and astrocytes. As shown in Fig. 6F–H, the protein level and mean fluorescence intensity (MFI) of GFAP increased in CRS + HC-tr group, compared to in adolescent CRS mice, indicating enhanced activation of astrocytes by HC-GM transplantation; meanwhile, as expected, Iba1-marked microglial activation was significantly ameliorated following normobiotic GM transplantation (Fig. 6F, I, J). These results, taken together, suggesting that FMT from healthy adolescent volunteers retards CRS induced depression-like changes may partially through ameliorating synaptic loss, microglial activation and improving astrocyte activity.

Perturbation of microbial homeostasis in the intestine is vulnerable to gut barrier damage and may further develop into “leaky gut” [27]. Lipopolysaccharide (LPS), along with other pro-inflammatory factors and toxic products, can easily translocate into the peripheral circulation and cause adverse effects on the CNS, such as synaptic injury and glial activation [28]. Since the pronounced increase of the toxic metabolites such as Kyn and Quin induced by CRS in the mouse colon, then the colonic structure integrity of the mice in this study was evaluated using their histological scores (Fig. 7A, B). CRS was found to cause evident epithelial structure deficiencies, manifesting as crypt distortion and goblet cell atrophy, which were remarkably recovered by HC-GM transplantation. Notably, DEP microbiota transplantation caused even worse barrier disruption, as the histological score in CRS + DEP-tr mice was higher than that in the CRS group (Fig. 7B). Collectively, FMT from healthy adolescent volunteers protects against CRS induced gut barrier damage, which may alleviate synapse injury and glial activation along the GBA.

Having documented the beneficial antidepressant effects of normobiotic FMT, we hypothesized that some microbiota in healthy adolescent volunteers may function as the probiotics to alleviate depression. Using 16S sequencing (Fig. 1E, F and Figure S1B) and combination of related studies [1, 29–31], we considered Ri., one favorable species of the genus Roseburia, as the target intervention microbial strain (Fig. 8A). Surprisingly, human Ri. was well customized in the mouse colon after FMT, as evidenced by its strong expression in the colon and feces of CRS + HC-tr mice (Fig. 8D). To further characterize the efficient colonization of Ri. in the mouse colon, green fluorescence-labeled Ri. was applied, in accordance with the results obtained in HC-GM transferring mice, the fluorescence intensity dramatically increased in the colonic segment of adolescent mice gavaged with Ri. in the CTR + Ri and CRS + Ri groups, respectively, as determined by FISH analysis (Fig. 8E, F), validating the efficient colonization of Ri. The question was then moved to whether Ri. colonization has beneficial effects on depression. Intriguingly, Ri. transplantation dramatically attenuated the depression- and anxiety-like behaviors induced by CRS (Fig. 8B, C), suggesting that Ri. and Ri. from healthy adolescent volunteers serve a salutary action on adolescent depression.

To decipher the role of Ri. in attenuating CRS-induced depressive behaviors, the Trp-Kyn metabolic pathway was investigated. As depicted in Fig. 9A, B; Figure S4A and S4B, CRS conspicuously suppressed 5-HT levels and reciprocally activated KP events in the PFC, serum, and colon, manifesting as elevated levels of Kyn, 3-HK, Quin, and NAD⁺. To support these phenomena, the Trp-derived rate-limiting enzymes aligned to these Trp-Kyn metabolites were detected, and significant reductions of TPH2 in the PFC and KAT2 in the colon were observed in CRS-treated mice, along with evident elevation of IDO1, 3HAO, and QPRT in the mouse brain and gut induced by CRS (Fig. 9C–E). Importantly, Ri. did restore the aberrant activation of KP metabolism. To gain more mechanistic insights into CRS-induced Kyn elevation in the PFC, large neutral amino acid transporter 1 (LAT1), a key transporter responsible for delivering larger, neutral amino acids and Kyn across the blood–brain barrier (BBB) from the peripheral circulation [32–34], was examined (Figure S5A and S5B). In line with the increased level of Kyn found in PFC, LAT1 expression in the CRS mouse brain was markedly increased.

Impairment of synaptic plasticity is one of the main causes of depression. To elucidate the mechanisms underlying the protective effects of Ri. on CRS-induced depression, synaptic markers and glial activities were investigated. As displayed in Fig. 10A–E, CRS induction decreased the densities of dendritic spines and synapses. This was significantly reversed by Ri. inoculation, as CRS induced downregulation of Drebrin, PSD95, and Syn1 was markedly rescued by Ri. administration. Similar to the results of healthy adolescent microbiota transplantation, aberrantly activated microglial cells and activity-suppressed astrocytes were ameliorated by Ri. transplantation, manifesting as decreased Iba-1 and increased GFAP expression (Fig. 10F–J), demonstrating the salutary effects of Ri. in anti-depression.

The relative abundance of Ri. was significantly decreased in inflammatory bowel disease (IBD) patients [35] which may be ascribed to its role in maintaining intestinal epithelial integrity and anti-inflammation [31]. Intestinal epithelial cells (IECs) serve as a physical and antimicrobial barrier against the microbiota, therefore, based on the protective effects of GM transplantation from healthy adolescents on colon endothelial cells, we wondered whether Ri. protects the structural integrity of the barrier in CRS mice. As expected, compared with the control mice, the epithelial integrity was significantly disrupted in the CRS mice, manifesting as increased crypt atrophy and lymphocyte infiltration at the base of the crypt and mucosal base (Fig. 11A, B). These changes were concomitant with decreased amounts of crypts and Occludin, which serves as one marker for barrier integrity (Fig. 11C, D). Of note, Ri. transplantation obviously mitigated CRS-induced intestinal epithelial impairment (Fig. 11A–D).

Epithelial regeneration is essential for barrier maintenance and organ function [36, 37]. Therefore, specific markers of intestinal stem cells (ISCs) were investigated. As depicted in Fig. 11E, CRS led to significantly decreased mRNA expression of ISCs, while Ri. intervention resulted in a pronounced elevation of MSI1 expression and slightly increased LGR5, BMI1, and HOPX levels, suggesting a favorable resolution of Ri. colonization in the colon and fewer effects on the neogenesis of colonic epithelial cells.

The present study has some limitations. First, due to the COVID-19 epidemic, the collection of population samples could not be conducted smoothly, thus the sample size is small and it will need to be amplified in the future. Second, the causalities among synaptic plasticity and KP, colon barrier integrity, and KP should be validated with stronger supports, for example, the usage of IDO1 and 3HAO knockout mice. Third, the most efficient compositions of the Ri. need to be clarified, and proteomics testing for bacterial secretions or structural substances is urgently needed in the future.

This study was approved by Nanjing Medical University and was conducted according to animal ethics guidelines and approved protocols.

Not applicable.

The authors declare that they have no competing interests.

The raw reads were deposited into the NCBI Sequence Read Archive (SRA) (BioProject ID: PRJNA894780). All data were available upon request from the authors.