Abstract

   The human microbiome has a pronounced impact on human physiology and
   behaviour. Despite its unique anatomical connection to the brain, the
   role of the nasal microbiome in neurological diseases is understudied.
   Here, using human data and experiments in mice, we show that nasal
   Staphylococcus aureus is linked to depression. Nasal microbiome
   analyses revealed a positive correlation between depression scores and
   S. aureus abundance among patients with depression and healthy
   controls. Metabolomics of the nasal cavity showed decreased sex
   hormones, estradiol and testosterone in patients with depression versus
   controls. Nasal microbiota transplants from patients reproduced
   depression-like behaviour in mice with differential abundance of S.
   aureus. Further homology and mutational analysis uncovered an S. aureus
   sex hormone-degrading enzyme, 17b-hydroxysteroid dehydrogenase (Hsd12),
   which degraded testosterone and estradiol in mice, leading to lower
   levels of dopamine and serotonin in the murine brain. These findings
   reveal a nasal commensal that influences depressive behaviour and
   provides insights into the nose–brain axis.

   Subject terms: Microbiome, Neurological disorders
     __________________________________________________________________

   Nasal colonization by Staphylococcus aureus is linked with depression
   in a human cohort and shown in a mouse model to cause decreased
   serotonin and dopamine in the brain.

Main

   Depression is a complex disorder whose pathogenesis is
   multifactorial^[66]1. It causes considerable personal suffering, an
   increased risk of suicide, cardiac disease and other comorbidities, as
   well as a substantial economic burden^[67]2. At the molecular level,
   depression is often accompanied by an abnormal abundance of cytokines,
   neurotrophic factors and sex hormones, leading to changes in
   neurotransmitter concentrations at the synapses^[68]3. Steroid sex
   hormones in particular serve as potent endogenous neuromodulators and
   have been shown to exert a substantial impact on the pathophysiology of
   depression, in large part through their modulation of dopamine and
   serotonin biosynthesis^[69]4.

   In recent years many human morbidities have been linked to a dysbiosis
   of the colonizing microbiota^[70]5. There is now also profound evidence
   that the gut microbiota influences brain function. This includes
   depression^[71]6, which may happen via alteration of the levels of
   dietary neurotransmitter precursors^[72]7 or degradation of sex
   hormones^[73]8,[74]9.

   It remains poorly understood if bacterial colonization or infection of
   body sites other than the gut impacts brain function. Hosang and
   colleagues reported that the lung microbiome influences brain
   auto-immunity^[75]10 and Wang and colleagues reported that
   Porphyromonas gingivalis, an oral pathogen, impacts depressive
   behaviour in mice^[76]11. Whether and how the nasal microbiome impacts
   brain function has remained virtually unaddressed. This is surprising
   given that the close vicinity of the nasal system to the brain and the
   fact that the intranasal route allows direct entry of drugs and
   biologics to the central nervous system indicate that metabolites
   produced by nose-colonizing bacteria may reach and affect the
   brain^[77]12,[78]13.

   In this study we analysed the nasal microbiome of people suffering from
   depression compared with that of healthy controls. Staphylococcus
   aureus emerged as the predominant bacterium associated with depression.
   We show that S. aureus colonization promotes depressive behaviour in
   mice via an enzyme that degrades sex hormones, leading to decreased
   levels of serotonin and dopamine in the brain. These findings establish
   a role of the nasal microbiome in the pathogenesis of depression and
   identify S. aureus, a nasal commensal in a considerable subset of the
   human population^[79]14, as responsible for the underlying mechanism.

Nasal S. aureus carriage is associated with depression

   We analysed nasal microbiome samples from 100 patients with untreated
   depression and 118 healthy controls. Other than for measures of anxiety
   and depression, there were no significant demographic or clinical
   differences between the two groups, including the serum concentrations
   of hormones and cytokines known to potentially affect anxiety and
   depression^[80]15 (Supplementary Table [81]1). The microbiome diversity
   (α-diversity) was higher in samples from healthy individuals than those
   with depression and the microbiota of the two groups were
   significantly^[82]16 dissimilar (β-diversity; Fig. [83]1a,b),
   indicating a significant association of the nasal microbiome
   composition with depression. Following analysis using multivariate
   logistic regression models adjusting for several covariates including
   sex, age, body mass index, education, income, psychosocial and
   biochemical determinants, the α-diversity remained significantly
   different (Supplementary Table [84]2).

Fig. 1. Nasal S. aureus carriage is associated with depression.

   [85]Fig. 1
   [86]Open in a new tab

   a–g, Results from 16S rRNA microbiome sequencing. a, The Shannon (left)
   and Simpson (right) methods were used to calculate the α-diversity in
   the nasal microbiomes of healthy individuals and (n = 118) and patients
   with depression (n = 100) at the species level. b, β-Diversity
   comparison of the composition of the nasal microbiomes of the two study
   cohorts at the species level. Shaded areas show the 95% confidence
   interval. c, Per-individual and overall relative abundances of the most
   abundant species in the nasal microbiomes of healthy individuals
   (n = 118) and individuals with depression (n = 100). d, Multivariate
   analysis adjusted for age, sex, body mass index, education, income,
   family relationships, adverse childhood experiences, available
   supportive friends and batch covariates. Fields marked with + and −
   signs represent significant associations (adjusted false discovery rate
   (FDR) < 0.05). A genus-level analysis is shown in Extended Data Fig.
   [87]1a,b. e, Actual abundance of S. aureus in the two cohorts (left) as
   well as among the males (middle) and females (right) among those
   cohorts. f, Correlation of S. aureus abundance with PHQ-9 depression
   score in males and females among all individuals. The line shows simple
   linear regression; r, correlation coefficient. Shaded areas show the
   95% confidence intervals. An analysis of the GAD-7 anxiety score is
   provided in Extended Data Fig. [88]1f. e,f, The S. aureus abundance was
   determined as the log-transformed number of nuc copies per μl of DNA.
   g, S. aureus carrier status in the two study cohorts. An individual was
   defined as a S. aureus carrier if at least one of 24 randomly selected
   bacterial clones in the culture-based analysis was found to be S.
   aureus using matrix-assisted laser desorption ionization–time-of-flight
   mass spectrometry (MALDI-TOF-MS) analysis. The number of individuals in
   the different groups are shown in the bars. Statistical analysis was
   performed using a two-sided Mann–Whitney test (a,e), permutational
   multivariate analysis of variance (PERMANOVA; b), Spearman’s
   correlation (f) or Fisher’s exact test (g). Data are the mean ± s.d.

   [89]Source data

   Using a method that allows species-level determination^[90]17,[91]18,
   we discovered that among the most abundant genera or species, only the
   genus Staphylococcus and specifically the species S. aureus were
   significantly more abundant in individuals with depression compared
   with healthy controls (Fig. [92]1c,d and Extended Data Fig. [93]1a,b).
   Culture-based analysis substantiated that assessment (Extended Data
   Fig. [94]1c,d). There was also a significant difference in absolute S.
   aureus abundance, both when males and females were analysed separately
   and in the combined groups (Fig. [95]1e). Moreover, Patient Health
   Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7)
   scores were significantly^[96]16 associated with S. aureus abundance in
   females and males (Fig. [97]1f and Extended Data Fig. [98]1e–g). These
   findings showed that S. aureus abundance in the nasal microbiome is
   positively associated with depression.

Extended Data Fig. 1. Nasal microbiome analysis in healthy versus depressed
cohorts—additional data.

   [99]Extended Data Fig. 1
   [100]Open in a new tab

   a, Per-individual and overall relative abundances of the most abundant
   genera in the nasal microbiomes of healthy (n = 118) and depressed
   (n = 100) cohorts as per 16S rRNA sequencing. b, Corresponding
   multivariate analysis adjusted for age, sex and batch covariates. c,
   Per-individual and overall abundances of the most abundant species as
   per culture-based analysis. d, Corresponding multivariate analysis
   adjusted for age, sex and batch covariates. e, Multivariate analysis of
   bacterial abundance and PHQ-9 scores adjusted for age, sex, BMI,
   education, income, family relationships, adverse childhood experiences,
   available supportive friends, and batch covariates. f,g, Multivariate
   analysis and correlation plot of bacterial and S. aureus (SA)
   abundance, respectively, and GAD-7 anxiety scores using the same
   analytical approach as applied to the PHQ-9. Statistical analysis is by
   Spearman correlation (f). The lines show simple linear regression.
   Shaded areas show 95% confidence intervals. h, Sequence-type
   distribution of randomly selected isolates (one per colonized
   individual; n = 31, healthy cohort; n = 45, depressed cohort). In
   multivariate plots (b,d,e,g), fields marked with “+” and “-“ signs
   represent significant associations (FDR < 0.05).

   [101]Source data

   S. aureus colonizes the nose (as well as, in a largely correlated
   fashion, other body sites) only in about a quarter to a third of the
   human population^[102]14. We found that there was a significantly
   higher abundance of S. aureus nasal carriers among patients with
   depression than healthy individuals (Fig. [103]1g), suggesting that S.
   aureus carriage is a risk factor for depression.

   Finally, sequence type analysis did not reveal that specific S. aureus
   strains are associated with depression, with sequence types 398 and 15
   (ST398 and ST15, respectively) dominating in both groups, suggesting
   that any potential causal relationship is probably not strain-specific
   (Extended Data Fig. [104]1h).

Nasal S. aureus induces depressive behaviour in mice

   To analyse experimentally whether the nasal microbiome impacts
   depression, we transferred nasal microbiome samples from healthy
   individuals and patients with depression into the nasal cavities of
   mice (one donor to one mouse, randomly selected but always male to male
   and female to female). We then used established mouse models for the
   subsequent analysis of depressive behaviour (Fig. [105]2a). We first
   ascertained that the model set-up appropriately measured the impact of
   the transferred microbiomes without cofounders and the mice did not
   show signs of disease^[106]19 (Extended Data Fig. [107]2a). Application
   of antibiotics to the nostrils resulted in a decrease of only
   approximately two orders of magnitude of colony-forming units (CFUs)
   and significant changes in the nasal, but not lung or gut microbiomes,
   indicating that any observed phenotypes are exclusively caused by nasal
   microbiome changes, as intended (Extended Data Fig. [108]2b–k).
   Following transplantation, the nasal microbiomes of the two groups mice
   (transplants from healthy individuals versus from patients with
   depression) differed significantly, reflecting the differences observed
   in humans (Extended Data Fig. [109]2l–n). Finally, the nasal
   microbiomes changed significantly after transplantation and recovered
   to bacterial abundance levels and a composition comparable to that
   before the experiment within a course of 14 days after the last
   transplantation (Extended Data Fig. [110]2b,o–q).

Fig. 2. Nasal S. aureus carriage induces depressive behaviour in mice.

   [111]Fig. 2
   [112]Open in a new tab

   a, Set-up of the experiment to investigate the impact of nasal
   microbiome transplantation on anxiety and depression-like behaviour in
   mice. b, Anxiety-like behaviour of mice in the OFT test (left) as well
   as depression-like behaviour in FST (middle) and TST (right) tests.
   Male and female mice received nasal transplants from male and female
   donors, respectively (one donor per one mouse); n = 16 mice (n = 8 male
   and n = 8 female; sex-specific analysis in Extended Data Fig. [113]3a)
   per group. Light blue and orange symbols depict S. aureus-positive (by
   culture analysis) transplanted microbiomes. c, Multivariate analysis
   (covariate-adjusted analysis using MaAsLin2) adjusted for sex. d,
   Set-up of the experiment to investigate anxiety and depression-like
   behaviour in female mice nasally colonized with S. aureus or S.
   epidermidis and control mice. e, Anxiety-like behaviour of mice in the
   OFT (left) as well as depression-like behaviour in FST (middle) and TST
   (right) tests. Data for the corresponding experiment with male mice in
   Extended Data Fig. [114]6. f, Set-up of the experiment to investigate
   anxiety and depression-like behaviour in male mice nasally colonized
   with S. aureus or S. epidermidis and control mice with CUMS treatment.
   Bacteria were administered seven times at 48-h intervals, including
   during the seven-day CUMS protocol. g, Anxiety-like behaviour of mice
   in the OFT (left) as well as depression-like behaviour in the FST
   (middle) and TST (right) tests. e,f, n = 18 (control and SE) or 20 (SA)
   mice per group. Data are the mean ± s.d. Statistical analysis was
   performed using a two-sided unpaired Student’s t-test (b), or two-sided
   one-way analysis of variance (ANOVA) or Kruskal–Wallis test (depending
   on the normality of distribution in the groups) and Tukey’s and Dunn’s
   post-tests, respectively (e,g). Colonization levels and evidence for
   absence of disease and inflammation in Extended Data Figs. [115]4 and
   [116]5. SA, S. aureus; SE, S. epidermidis. Credit: mouse illustrations
   in a,d,f, creazilla under a Creative Commons license [117]CC0 1.0.

   [118]Source data

Extended Data Fig. 2. Nasal transplant experiment—additional data 1.

   [119]Extended Data Fig. 2
   [120]Open in a new tab

   a, Weight, temperature and sickness scores over the time of the
   experiment. b, CFU over the course of the experiment. c,d, α-Diversity
   analyses of the nasal microbiomes of antibiotic-treated (Abx) versus
   control mice that did not receive antibiotic, measured at day 8 (see
   Fig. [121]2a). e, β-Diversity analysis of the nasal microbiomes of
   antibiotic-treated (Abx) versus control mice that did not receive
   antibiotic, measured at day 8 (see Fig. [122]2a). f–h, Corresponding
   analyses for the lung microbiomes. i–k, Corresponding analyses for the
   intestinal (fecal) microbiomes. l–n, α- And β-diversity analyses of
   mouse nasal microbiomes receiving nasal transplants from healthy versus
   depressed people, measured at day 17 (see Fig. [123]2a). o–q, α- And
   β-diversity analyses of mouse nasal microbiomes over the course of the
   experiment in comparison. n = 12/group (c–n); n = 6/group (a,o–q);
   n = 12 (b). Statistical analyses are by unpaired, two-sided t-tests
   (c,i,l), two-sided Mann–Whitney tests (e–g,j,m), two-sided
   Kruskal–Wallis tests versus control (o,p) and PERMANOVA (d,h,k,n).
   Shaded areas show 95% confidence intervals (d,h,k,n,q). All error bars
   show the mean ± standard deviation (SD).

   [124]Source data

   We observed significantly increased levels of anxiety-like behaviour in
   the open field test (OFT) and depression-like behaviour in forced
   swimming and tail suspension tests (FST and TST, respectively) in mice
   that received transplants from patients with depression (Fig. [125]2b
   and Extended Data Fig. [126]3a). Furthermore, the OFT, FST and TST
   values were significantly and at ‘fair’ levels^[127]16 correlated with
   the levels of depression and anxiety of the transplant donors, as
   measured by PHQ-9 and GAD-7 clinical scales, respectively (Extended
   Data Fig. [128]3b). These findings strongly indicated that the
   composition of the nasal microbiome influences depressive behaviour.
   Notably, in a multivariate analysis, mice that received microbiota
   transplants from depressed patients showed significantly higher S.
   aureus relative abundance compared with mice that received microbiota
   transplants from healthy controls (Fig. [129]2c).

Extended Data Fig. 3. Nasal transplant experiment—additional data 2.

   [130]Extended Data Fig. 3
   [131]Open in a new tab

   a, Sex-specific behavioural analyses. Male mice received nasal
   transplants from male and female from female donors. n = 8/group.
   Statistical analysis is by unpaired, two-tailed t-tests except for
   female OFT, two-sided Mann–Whitney test. Error bars show the
   mean ± standard deviation (SD). b, Correlation between anxiety and
   depression scores in humans with anxiety and depressive-like behaviour
   tests in nasal transplant-recipient mice. n = 16/group (n = 8 male,
   n = 8 female). Statistical analysis is by Spearman correlation. Lines
   show simple linear regression. Shaded areas show 95% confidence
   intervals. Credit: mouse illustrations in a, creazilla under a Creative
   Commons license [132]CC0 1.0.

   [133]Source data

   To provide direct experimental evidence for a causal relationship
   between S. aureus carriage and depression, we colonized the noses of
   mice with S. aureus and then measured anxiety and depressive behaviour
   separately in male and female mice (Fig. [134]2d–g). To demonstrate a
   specific role of S. aureus, we performed additional comparisons to mice
   that received the equal numbers of Staphylococcus epidermidis, a less
   virulent relative of S. aureus that also colonizes the
   nose^[135]20,[136]21. The two bacterial species achieved similar
   colonization levels in our mouse model, allowing direct comparison
   (Extended Data Fig. [137]4a). No changes in lung and nose cytokine
   levels, histological appearance of lung tissue and nasal epithelia,
   abundance of olfactory macrophages and neutrophils or expression of the
   olfactory marker protein were observed between the control and S.
   aureus or S. epidermidis nasally colonized mice (Extended Data Fig.
   [138]4b–h). Even shortly (6 h) after the last application of the
   bacteria, there were no differences in cytokine expression of the lungs
   and only minor changes in the transcriptome 3 h after the last
   application of the nasal mucosa, which are expected and reflective of
   physiological immune activation associated with commensal colonization
   (Extended Data Fig. [139]5a–c). Moreover, no differences in weight or
   temperature were noted between the groups and no signs of disease were
   observed (Extended Data Fig. [140]5d–f). Together, these controls
   substantiated that our model is a model of asymptomatic nasal
   colonization. Notably, S. aureus colonized the nose at >1 × 10^3 CFU
   over the entire course of the experiment and was undetectable in the
   lungs, whereas intestinal colonization^[141]14 was transient,
   indicating that the phenotypes observed in this experiment are due to
   nasal colonization rather than colonization at other body sites
   (Extended Data Fig. [142]5g).

Extended Data Fig. 4. Mouse colonization experiment—additional data 1.

   [143]Extended Data Fig. 4
   [144]Open in a new tab

   a, Colonizing CFU in the nose at the time of euthanasia (day 21).
   n = 6/group. b,c, Nose and lung cytokine concentrations in S. aureus
   (SA) nasally colonized, S. epidermidis (SE) nasally colonized, and
   control female mice at the time of euthanasia. n = 6/group. d,e, Nose
   and lung cytokine concentrations in S. aureus nasally colonized, S.
   epidermidis nasally colonized, and control male (CUMS-treated) mice at
   the time of euthanasia. n = 6/group. f, Nose and lung histology in S.
   aureus nasally colonized, S. epidermidis nasally colonized, and control
   male (CUMS-treated) mice at the time of euthanasia. g, Histological
   analysis of OMP and Iba1 expression in the olfactory epithelium at the
   time of euthanasia (day 21). h, Ly6G expression in the olfactory
   epithelium at the time of euthanasia (day 21). Statistical analysis is
   by two-sided one-way ANOVA with Tukey’s post-tests (b–e) and two-sided,
   unpaired t-tests (a). There were no significant differences. All error
   bars show the mean ± standard deviation (SD). Credit: mouse
   illustrations in b,d,f, creazilla under a Creative Commons license
   [145]CC0 1.0.

   [146]Source data

Extended Data Fig. 5. Mouse colonization experiment—additional data 2.

   [147]Extended Data Fig. 5
   [148]Open in a new tab

   a,b, Lung cytokine concentrations 6 h post the last bacterial
   application (day 18) in female (a) and CUMS-treated male mice (b).
   n = 3/group. Statistical analysis is by two-sided 1-way ANOVAs with
   Tukey’s post-tests. There were no significant differences. c, Cnet
   plots of transcriptomic changes (5 top enriched KEGG pathways) in the
   nasal mucosa 3 h post the last bacterial application (day 18).
   n = 3/group (male CUMS-treated mice). d–f, Weight, temperature and
   sickness scores over the time of the experiment. n = 6/group. g,
   Bacterial loads in the noses, lungs, and faeces measured starting 2 h
   past the time of the last bacterial application. n = 6/group. All error
   bars show the mean ± standard deviation (SD). SA, S. aureus; SE, S.
   epidermidis. Credit: mouse illustrations in a,b, creazilla under a
   Creative Commons license [149]CC0 1.0.

   [150]Source data

   Female, but not male, mice colonized with S. aureus exhibited increased
   anxiety and depressive behaviour (Fig. [151]2e and Extended Data Fig.
   [152]6a,b). When male mice were additionally treated with a protocol of
   chronic unpredictable mild stress (CUMS; Fig. [153]2f), the results
   were similar to those observed in female mice (Fig. [154]2g). These
   findings indicate that S. aureus nasal carriage promotes depression and
   does so more strongly in female than male mice.

Extended Data Fig. 6. Induction of anxiety and depression-like behaviour in
male mice and metabolome and sex hormone measurement in healthy and depressed
cohorts—additional information.

   Extended Data Fig. 6
   [155]Open in a new tab

   a, Set-up of the experiment (as in Fig. [156]2d, but using male mice).
   b, Anxiety-like behaviour of mice in OFT and depression-like behaviour
   in FST and TST tests. n = 8/group. SA, S. aureus; SE, S. epidermidis.
   c, Serum estradiol and testosterone levels in female and male cohorts,
   respectively. n = 19 (male, healthy), n = 34 (male, depressed), n = 24
   (female, healthy), n = 42 (female, depressed). Serum levels were
   determined in all individuals who consented to having blood drawn. d,
   Correlation of nasal testosterone or estradiol levels with GAD-7
   anxiety scores. Statistical analysis is by two-sided one-way ANOVA with
   Tukey’s post-tests (b), two-sided Mann–Whitney tests (c: serum
   estradiol), two-sided, unpaired t-test (c: serum testosterone) and
   Spearman correlation (d). Error bars show the mean ± standard deviation
   (SD) (b,c). Lines show simple linear regression and shaded areas show
   95% confidence intervals (d). Credit: mouse illustrations in a,b,
   creazilla under a Creative Commons license [157]CC0 1.0.

   [158]Source data

Reduced nasal sex hormone levels are associated with depression in mice

   To explain our findings mechanistically, we hypothesized that an
   increased abundance of S. aureus alters metabolism in the nose, which
   could modulate brain physiology and behaviour. We therefore performed
   tandem mass spectrometry (MS/MS)-based metabolomic profiling of nasal
   metabolites in randomly selected individuals from both groups (n = 40;
   population characteristics of those subsets in Supplementary Table
   [159]3).

   According to a Bray–Curtis distance-based evaluation of principal
   coordinate analysis (PCoA; Fig. [160]3a) and an orthogonal partial
   least squares-discriminant analysis (OPLS-DA; Fig. [161]3b), there was
   a significant difference between the nasal metabolomes of healthy and
   individuals and patients with depression. Significantly altered
   pathways predominantly included pathways related to steroid hormone
   biosynthesis (Fig. [162]3c). Furthermore, nasal testosterone and
   estradiol levels were significantly decreased in patients with
   depression compared with healthy individuals (Fig. [163]3d,e), whereas
   the serum levels of these hormones were similar (Extended Data Fig.
   [164]6c). Moreover, the nasal estradiol and testosterone levels were
   significantly and at ‘fair’ to ‘moderate’ degrees^[165]16 correlated
   with the anxiety and depression scores (Fig. [166]3f,g and Extended
   Data Fig. [167]6d). These findings demonstrated that differential
   concentrations of sex hormones in the nose are associated with
   depression, suggesting that depression may be impacted by nasal
   microbiome members via alteration of sex hormone levels.

Fig. 3. Reduced nasal sex hormone levels are associated with depression.

   [168]Fig. 3
   [169]Open in a new tab

   a,b, The nasal metabolomes of individuals in the two cohorts (n = 40
   each; randomly selected) were analysed using PCoA with Bray–Curtis
   distance-based PERMANOVA. R^2 proportion of variance explained by the
   principal coordinate (a) and OPLS-DA analyses, with the model fit
   parameters (R^2x, R^2y) and predictive ability (Q^2) shown (b). c,
   Significantly altered metabolic pathways by KEGG enrichment. d,e, Nasal
   estradiol (d) and testosterone (e) levels in the same randomly selected
   cohort subsets. Data are the mean ± s.d. Serum levels are provided in
   Extended Data Fig. [170]6c. f,g, Correlation between the levels of
   nasal estradiol (f) and testosterone (g), and PHQ-9 depression scores.
   GAD-7 anxiety score analysis in Extended Data Fig. [171]6d. a,b,f,g,
   Lines show simple linear regression and shaded areas the 95% confidence
   intervals. Statistical analysis was performed using a two-sided
   Mann–Whitney test (d,e) or Spearman’s correlation (f,g).

   [172]Source data

Nasal S. aureus degrades sex hormones and is associated with decreased
dopamine and serotonin levels in the brain

   In humans testosterone is converted to androstenedione and estradiol to
   estrone by 17β-hydroxysteroid dehydrogenases (HSDs; Fig. [173]4a). We
   hypothesized that S. aureus produces a similar activity. In accordance
   with the idea that S. aureus degrades sex hormones in the nose, nasal
   estradiol and testosterone in humans were at a ‘fair’ degree^[174]16
   negatively correlated with the abundance of S. aureus, with the
   correlation being stronger for estradiol (Fig. [175]4b). Only S. aureus
   was associated with increases in metabolites belonging to steroids and
   their derivatives in a multivariate metabolomics analysis
   (Supplementary Fig. [176]1). Furthermore, S. aureus had the ability to
   convert estradiol to estrone and testosterone to androstenedione in
   vitro, with the estradiol-converting activity stronger than the
   testosterone-converting activity (Fig. [177]4c). Although some other
   nose-colonizing bacteria that we tested exhibited similar activities in
   vitro, these were significantly less pronounced than in S. aureus (Fig.
   [178]4d). All tested S. aureus sequence types exhibited similar
   activities (Extended Data Fig. [179]7a). In the mouse nasal
   colonization model, S. aureus colonization led to significantly reduced
   levels of nasal estradiol and testosterone compared with controls and
   S. epidermidis colonization (Fig. [180]4e), whereas the levels of sex
   hormones in serum remained unchanged (Extended Data Fig. [181]7b). S.
   aureus also had a similar effect on sex hormone levels in the brain,
   substantiating our hypothesis about the impact of the nasal microbiome,
   and particularly S. aureus on, brain metabolites. The changed levels of
   sex hormones in the brain were not due to differences in de novo
   biosynthesis, as the levels of the precursors pregnenolone and
   progesterone were unchanged (Extended Data Fig. [182]7c). Furthermore,
   transcriptomic analysis of midbrain samples of mice nasally colonized
   with S. aureus showed that expression of the CYP17A1 and HSD genes
   involved in sex hormone biosynthesis were unchanged between the mice
   that were nasally colonized with S. aureus and those colonized with S.
   epidermidis as controls (Fig. [183]4f,g).

Fig. 4. Nasal S. aureus degrades sex hormones and is associated with
decreased dopamine and serotonin levels in the brain.

   [184]Fig. 4
   [185]Open in a new tab

   a, Testosterone and estradiol biosynthesis. Note the HSD-catalysed
   conversion of androstenedione/testosterone and estradiol/estrone.
   Enzyme names are in green font; hydroxyl and keto groups affected by
   the HSD enzymes are in red. b, Correlation between nasal testosterone
   (right) and estradiol (left) concentrations, and nasal S. aureus
   abundance in combined selected cohort subsets (n = 80). The line shows
   the simple linear regression. Shaded areas show 95% confidence
   intervals. c, In vitro conversion of estradiol to estrone (left) and
   testosterone to androstenedione (right) by S. aureus culture filtrate
   within 24 h (n = 3 per group). d, Conversion of estradiol to estrone
   (left) and testosterone to androstenedione (right) by culture filtrates
   from different bacteria. Tryptone soya broth (TSB) control, n = 4;
   n = 3 for all other groups. e, Nasal and midbrain levels of
   testosterone and estradiol in mice colonized with S. aureus or S.
   epidermidis, and in controls. Estradiol was measured in female mice
   (n = 6 per group) and testosterone in male mice (n = 8 per group). f,g,
   Transcriptome analysis of midbrain gene expression in male and female
   mice nasally colonized with S. aureus compared with S. epidermidis.
   Red, increased in S. aureus-colonized mice; green, increased in S.
   epidermidis-colonized mice. Dopamine and serotonin biosynthesis enzymes
   are labelled; n = 3 per group. e–g, Experimental set-up as depicted in
   Fig. [186]2d,f. h, Dopamine and serotonin biosynthesis. Enzyme names
   are in green font. i, Concentrations of serotonin (left) and dopamine
   (right) in the midbrains of S. aureus-colonized, S.
   epidermidis-colonized and control mice; n = 7 mice (n = 4 male and
   n = 3 female) mice per group. The GABA and glutamate concentrations are
   in Extended Data Fig. [187]9f. c–e,i, Data are the mean ± s.d.
   Statistical analysis was performed using a two-sided one-way ANOVA or a
   Kruskal–Wallis test, depending on the normality of distribution in the
   groups, and Tukey’s and Dunn’s post-tests, respectively (d,e,i), or
   Spearman’s correlation (b). SA, S. aureus; SE, S. epidermis. Steroid
   formulae in a and h created with [188]BioRender.com. Credit: mouse
   illustrations in e–g, creazilla under a Creative Commons license
   [189]CC0 1.0.

   [190]Source data

Extended Data Fig. 7. Nasal S. aureus degrades sex hormones and increases
dopamine and serotonin levels in the brain—additional information.

   [191]Extended Data Fig. 7
   [192]Open in a new tab

   a, Conversion of estradiol and testosterone to estrone and
   androstenedione, respectively, by randomly selected isolates of main
   isolate STs. n = 3/group. b, Serum estradiol and testosterone levels in
   mice colonized with S. aureus (SA) or S. epidermidis (SE), or in
   controls. n = 6–8/group (n = 8, CUMS-treated male; n = 6, female). c,
   Pregnenolone and progesterone midbrain levels. n = 6/group (n = 3,
   CUMS-treated male; n = 3, female). d, Expression of serotonin and
   dopamine biosynthesis genes in the midbrain of mice colonized with S.
   aureus or S. epidermidis, or in controls. n = 7/group (n = 4,
   CUMS-treated male; n = 3, female). e, Results of KEGG pathway
   enrichment. f, GABA and glutamate midbrain concentrations in S.
   aureus-colonized, S. epidermidis-colonized, and control mice.
   n = 7/group (n = 4, CUMS-treated male; n = 3, female). Statistical
   analysis is by two-sided one-way ANOVAs or Kruskal–Wallis tests,
   depending on normality of distribution in the groups, and Tukey’s and
   Dunn’s post-tests, respectively (b–d,f). All error bars show the
   mean ± standard deviation (SD). Credit: mouse illustrations in e,
   creazilla under a Creative Commons license [193]CC0 1.0.

   [194]Source data

   Estradiol and testosterone are known to increase the levels of dopamine
   and serotonin in the midbrain^[195]22,[196]23, a region that is known
   to play a crucial role in emotional regulation^[197]24. These
   neurotransmitters have a well-known impact on depression. Dopamine is
   synthesized from tyrosine via dihydroxyphenylalanine (DOPA) in two
   steps by tyrosine hydroxylase (TH) and DOPA decarboxylase (DCC).
   Serotonin (5-hydroxytryptamine, 5-HT) is synthesized from tryptophan
   via 5-hydroxytryptophan (5-HTP) by tryptophan hydrolase (TPH2) and
   5-HTP decarboxylase (Fig. [198]4h). The transcriptomic analysis also
   revealed that TPH2, TH and DOPA decarboxylase transcripts were
   significantly less abundant in both male and female mice than in their
   S. epidermidis-colonized counterparts (Fig. [199]4f,g), indicating that
   sex hormone degradation by S. aureus in the nose affects dopamine and
   serotonin levels in the brain via downregulation of their biosynthesis.
   Changes in the expression of these genes were verified by quantitative
   PCR with reverse transcription of midbrain tissue samples (Extended
   Data Fig. [200]7d). In accordance with these findings, Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis
   revealed significant changes in related metabolic pathways (for
   example, tyrosine and tryptophan metabolism; Extended Data Fig.
   [201]7e). Importantly, we then verified that decreased biosynthesis
   results in decreased serotonin and dopamine concentrations in the
   brains of mice nasally colonized with S. aureus (Fig. [202]4i). In
   contrast, the concentrations of other neurotransmitters were not
   affected (Extended Data Fig. [203]7f).

An S. aureus HSD degrades sex hormones and induces depressive behaviour in
mice

   To elucidate the mechanism by which S. aureus degrades sex hormones, we
   searched the genome of the S. aureus nasal isolate that we used in this
   study and whose genome we sequenced for putative hsd genes. To that
   end, we searched for members of the short-chain dehydrogenase/reductase
   (SDR) family, which have physiological roles in steroid hormone,
   prostaglandin and retinoid metabolism^[204]25 (Supplementary Table
   [205]4). Excluded from further analysis were genes with known different
   functions. We then expressed the selected putative hsd genes in
   Escherichia coli and tested for the ability to degrade estradiol and
   testosterone. Only one expression strain, expressing the putative hsd
   gene number 12 (hsd12), showed a strong capacity to degrade estradiol
   and testosterone. This identified the hsd12 gene as responsible for sex
   hormone degradation in S. aureus and as a previously unknown 17β-HSD,
   isozymes of which are used by the host to maintain homeostasis of
   active estradiol and testosterone^[206]26 (Fig. [207]5a,b). The hsd12
   gene, annotated in the National Center for Biotechnology Information
   (NCBI) as an SDR family oxidoreductase, was present in all S. aureus
   isolates obtained in this study as well as highly conserved in
   published S. aureus sequences. It is in a region with predominantly
   genes without known functions (Extended Data Fig. [208]8a).

Fig. 5. An S. aureus HSD degrades sex hormones and induces depressive
behaviour in mice.

   [209]Fig. 5
   [210]Open in a new tab

   a, Conversion of testosterone to androstenedione and estradiol to
   estrone by putative S. aureus hsd genes expressed in E. coli. The
   putative hsd7 gene was not included as it could not be expressed,
   probably due to its large size (n = 3 per group). b, Conversion of
   testosterone to androstenedione and estradiol to estrone by culture
   filtrates from S. aureus wild type and its hsd12 isogenic mutant. TSB
   control, n = 4; other groups, n = 3. c, Set-up of the model to test for
   induction of anxiety and depression-like behaviour in female mice by
   nasal colonization with S. aureus wild type, its hsd12 isogenic mutant
   and controls. d, Anxiety-like behaviour of female mice in the OFT
   (left) as well as depression-like behaviour in FST (middle) and TST
   (right) tests (n = 17 per group). e,f, Nose (e) and midbrain (f)
   estradiol levels in female mice (n = 12 per group). g, Set-up of the
   model to test for induction of anxiety and depression-like behaviour in
   male CUMS-treated mice by nasal colonization with S. aureus wild type,
   its hsd12 isogenic mutant and controls. h, Anxiety-like behaviour of
   male mice in the OFT (left) as well as depression-like behaviour in FST
   (middle) and TST (right) tests (n = 17 per group). i,j, Nose (i) and
   midbrain (j) testosterone levels in male mice (n = 12 per group). k,
   Relative expression of TH and TPH2 genes in midbrain samples (n = 6 per
   group; n = 3 females and n = 3 CUMS-treated males). l, Midbrain
   serotonin (left) and dopamine (right) levels (n = 12 per group; n = 6
   females and n = 6 CUMS-treated males). Statistical analysis was
   performed using a two-sided one-way ANOVA or Kruskal–Wallis test,
   depending on normality of distribution in the groups, and Tukey’s and
   Dunn’s post-tests, respectively. Data are the mean ± s.d. SA, S.
   aureus; SAΔhsd, S. aureus hsd12 isogenic mutant. Credit: mouse
   illustrations in c,g, creazilla under a Creative Commons license
   [211]CC0 1.0.

   [212]Source data

Extended Data Fig. 8. Location and deletion of the S. aureus hsd12 gene.

   [213]Extended Data Fig. 8
   [214]Open in a new tab

   a, Genetic context of the hsd12 gene, shown for S. aureus strain USA300
   FPR3757. b, Growth comparison of S. aureus wild-type (SA) and isogenic
   hsd12 deletion (SAΔhsd) strains of strain ST398 in two different media.
   n = 4/group. c, RNA-Seq transcriptomic comparison of gene expression in
   ST398 wild-type versus isogenic hsd12 deletion strains grown in TSB to
   logarithmic growth phase. Significantly differentially expressed genes
   are coloured (red, up in Δhsd; blue, up in wild-type). d,e,
   Corresponding analyses for ST88 wild-type versus isogenic hsd12
   deletion strains. n = 4/group. All error bars show the mean ± standard
   deviation (SD).

   [215]Source data

   To link the observed S. aureus-mediated phenotypes directly to the
   hsd12 gene, we created an S. aureus hsd12-deletion mutant. Deletion of
   hsd12 did not change in vitro growth in different media and had only a
   very minor influence on the expression of other genes in the genome by
   RNA-sequencing (RNA-Seq) analysis (Extended Data Fig. [216]8b,c).
   Colonization with the hsd12-deletion strain in female and CUMS-treated
   male mice led to phenotypes indistinguishable from those observed in
   the controls but significantly different from those observed in mice
   colonized with the isogenic wild-type S. aureus strain (Fig.
   [217]5c–l). We also verified that bacterial colonization did not lead
   to changes in weight and temperature (Extended Data Fig. [218]9a,b),
   the mice did not show signs of illness (Extended Data Fig. [219]9c),
   and inflammation and cytokine and histology analyses did not show
   significant differences between the S. aureus and isogenic
   hsd12-deletion strains (Extended Data Fig. [220]9d–g). These findings
   indicated that the hsd12 gene is mechanistically responsible for the
   link between S. aureus nasal colonization with depression.

Extended Data Fig. 9. Absence of inflammation and histology differences in
mice nasally colonized by S. aureus wild-type and isogenic hsd12 (ST398)
deletion strains.

   [221]Extended Data Fig. 9
   [222]Open in a new tab

   a–c, Weight, temperature and sickness scores over the time of the
   experiment. n = 6/group. d, Nose and lung cytokine concentrations in S.
   aureus (SA) nasally colonized, S. aureus Δhsd (SAΔhsd) nasally
   colonized, and control female mice. n = 6/group. e, Nose and lung
   cytokine concentrations in S. aureus nasally colonized, S. aureus Δhsd
   nasally colonized, and control (CUMS-treated) male mice. n = 6/group.
   f, Colonizing CFU in the nose at the time of euthanasia. n = 6/group.
   g, Nose and lung histology in S. aureus nasally colonized, S. aureus
   Δhsd nasally colonized, and control male (CUMS-treated) mice.
   Statistical analysis is by two-sided one-way ANOVA with Tukey’s
   post-tests (d,e) and two-sided, unpaired t-tests (f). There were no
   statistically significant differences except for where marked. All
   error bars show the mean ± standard deviation (SD). Credit: mouse
   illustrations in d,e,g, creazilla under a Creative Commons license
   [223]CC0 1.0.

   [224]Source data

Depression-inducing capacity of a mouse-adapted S. aureus strain

   To rule out that the observed impact of S. aureus on depressive
   behaviour is strain-specific, we also performed key experiments using a
   strain (JSNZ)^[225]27 of sequence type 88 (ST88) that has been reported
   to be mouse-adapted^[226]27. S. aureus ST88 degraded estradiol and
   testosterone similarly to the ST398 strain that we used generally in
   this study (Fig. [227]6a). We created an hsd-deletion mutant of ST88,
   which did not show any growth defects or considerable changes in gene
   expression, excepting the hsd gene (Extended Data Fig. [228]8d,e), and
   exhibited markedly reduced capacity to degrade estradiol and
   testosterone (Fig. [229]6b).

Fig. 6. Depression-inducing capacity of a mouse-adapted S. aureus strain.

   [230]Fig. 6
   [231]Open in a new tab

   a, In vitro sex hormone conversion by S. aureus ST88 culture filtrate
   (n = 3 per group). b, Sex hormone conversion by culture filtrates from
   S. aureus ST88 wild type and hsd12 isogenic mutant (n = 3 per group).
   c, Set-up of model to test for the induction of anxiety and
   depression-like behaviour in mice by nasal colonization with S. aureus
   wild type, its hsd12 isogenic mutant and controls. d, Anxiety-like
   behaviour in the OFT (left) as well as depression-like behaviour in FST
   (middle) and TST (right) tests (n = 12 per group). e, Set-up of the
   model to measure degradation of applied sex hormones to the noses of
   mice. ORX, orchidectomy; OVX, ovariectomy. f, Detected sex hormone
   levels (estradiol in female and testosterone in male mice) with and
   without S. aureus application (n = 6 per group). g, Set-up of the model
   to test for depression and anxiety-like behaviour in female mice after
   application of sex hormones (in the background of S. aureus
   colonization). h, Anxiety-like behaviour in the OFT (left) as well as
   depression-like behaviour in FST (middle) and TST (right) tests (n = 10
   per group). i,j, Corresponding set-up (i) of the experiment in g and
   results in male mice (j; n = 10 per group). Statistical analysis was
   performed using a two-sided one-way ANOVA with Tukey’s (b,d) post-test
   with the exception of the male mice TST in d where a two-sided
   Kruskal–Wallis test with Dunn’s post-test was used, or and unpaired
   two-sided Student’s t-test (f,h,j) with the exception of female TST and
   OFT and male FST in h,j where a two-sided Mann–Whitney test was used.
   Data are the mean ± s.d. SA, S. aureus ST88 wild type; SAΔhsd, S.
   aureus hsd12 isogenic mutant. Credit: mouse illustrations in c–e,g,i,
   creazilla under a Creative Commons license [232]CC0 1.0.

   [233]Source data

   We then set up a mouse colonization model that replicated the general
   set-up of the model used with strain ST398 (Fig. [234]2d,f) but only
   using two nasal instillations with a strongly reduced bacterial amount
   seven days apart—a change prompted by the reported better overall mouse
   colonization capacity of that strain^[235]27 (Fig. [236]6c). Strain
   ST88 colonized the antibiotic-treated noses over the time course of the
   experiment like ST398 and like ST398 did not colonize the lungs but
   showed much higher capacity to colonize the gut (Extended Data Fig.
   [237]10a; compare with Extended Data Fig. [238]5g). Although the use of
   a mouse-adapted strain might better reflect a ‘natural’ colonization
   scenario; this means that results obtained with ST88 should be
   interpreted with the caveat that they might not exclusively be due to
   the nose–brain but also the gut–brain axis. ST88 colonization did not
   affect weight, temperature, disease or cytokine expression in the noses
   and lungs, and deletion of hsd did not affect colonization (Extended
   Data Fig. [239]10b–e). In most of the behavioural tests, S. aureus ST88
   increased depressive behaviour, whereas application of ST88Δhsd led to
   phenotypes indistinguishable from the control (Fig. [240]6d). These
   results showed that the promotion of depressive behaviour in mice by
   the Hsd enzyme of S. aureus is not strain-dependent.

Extended Data Fig. 10. ST88 experiments—additional information.

   [241]Extended Data Fig. 10
   [242]Open in a new tab

   a, Bacterial loads in noses, lungs, and guts (faeces) after second
   colonization (day 13; time 0 h). n = 6/group. b, Weights, body
   temperatures, and sickness scores over the time of the experiment.
   n = 6/group. c, Colonizing CFU in the nose at day 20. n = 6/group. d,e
   Nose and lung cytokine concentrations in S. aureus nasally colonized,
   S. aureus Δhsd nasally colonized, and control female (d) and
   (CUMS-treated, e) male mice. n = 6/group. f, Ex vivo imaging of nasal
   hydrogel-loaded FITC retention in the nose or excretion. Note prolonged
   retention in the nose (yellow arrows) in the hydrogel versus liquid
   control samples. g, In vitro cumulative release of estradiol or
   testosterone from the nasal hydrogel. Statistical analysis is by
   two-sided one-way ANOVA with Tukey’s post-tests (d,e). There were no
   significant differences. n = 3/group. All error bars show the
   mean ± standard deviation (SD). Credit: mouse illustrations in d,e,
   creazilla under a Creative Commons license [243]CC0 1.0.

   [244]Source data

   Finally, using strain ST88, we tested whether S. aureus degrades sex
   hormones artificially introduced into the noses of mice. To that end,
   we first eliminated the effects of physiological hormonal fluctuations
   via a bilateral ovariectomy in female mice and bilateral orchidectomy
   in male mice, and subsequently used a nasal hydrogel for sustained
   release of estradiol or testosterone (Fig. [245]6e). We used a hydrogel
   as it enables better brain targeting and avoids rapid renal or
   intestinal excretion, which we demonstrated with fluorescein
   isothiocyanate (FITC; Extended Data Fig. [246]10f). Furthermore, we
   ascertained that the gel releases loaded estradiol and testosterone
   rapidly and virtually completely over a period of two days (Extended
   Data Fig. [247]10g). Application of S. aureus together with estradiol-
   or testosterone-loaded hydrogel significantly decreased nasal and brain
   estradiol and testosterone levels (Fig. [248]6f). Finally, we showed
   that hydrogel-mediated nasal application of sex hormones significantly
   alleviated anxiety and depressive behaviour induced by S. aureus
   application (Fig. [249]6h–k). In addition to confirming the impact of
   nasal sex hormones on depression and anxiety-related phenotypes, these
   findings provide further direct evidence for S. aureus-mediated
   degradation of sex hormones in the nasal cavity and the impact of this
   mechanism on depression-like behaviour.

Discussion

   In this study we show that the nasal microbiome influences depression
   dependent on the presence of S. aureus in the nose. We demonstrate that
   S. aureus degrades the sex hormones testosterone and estradiol by
   expressing a previously unknown 17β-HSD isozyme that is encoded by a
   gene we named hsd12, which led to decreased concentration of dopamine
   and serotonin in the brain in mouse models. Although we found that some
   other nasal commensal bacteria also had some sex hormone-degrading
   activity in vitro, their abundance was not increased in humans
   suffering from depression.

   Peripheral bacterial infection^[250]28–[251]30 as well as the
   composition of the gut microbiome^[252]6–[253]9,[254]31 have been shown
   to impact neurological disease. However, a role of the nasal microbiome
   in neurological diseases has remained speculative or based merely on
   correlative evidence^[255]12,[256]32. Our discovery of a mechanism that
   causes an impact of the nasal microbiome on depression provides
   previously unavailable evidence for a key role of the nose–brain axis
   in the development of neurological disease.

   The field of microbial endocrinology is rapidly
   advancing^[257]33,[258]34, with several recent studies illustrating
   hormone conversion by gut bacteria^[259]8,[260]9,[261]35,[262]36. By
   identifying an enzyme in the nasal microbiome that converts steroid sex
   hormones, our research indicates the presence of isozymes involved in
   steroid metabolism within host-associated environments beyond the
   intestine.

   Both male and female sex hormones have been implicated in determining
   depressive behaviour^[263]22,[264]37 and there are distinct
   sex-dependent differences in the incidence of depression^[265]38.
   Accordingly, we observed a stronger correlation between nasal estradiol
   levels, compared with testosterone, with depressive behaviour in humans
   and more pronounced induction of depressive behaviour by S. aureus in
   female mice compared with males. Our finding that S. aureus degrades
   the female sex hormone estradiol much more rapidly than the male sex
   hormone testosterone may explain these differences, and S. aureus nasal
   colonization may thus contribute to the observed sex-dependent
   difference in depression prevalence.

   The impact of commensal S. aureus in the nasal microbiome on depression
   is categorically different from that of infection on neurological
   disease that is inflammation-associated^[266]28–[267]30. Asymptomatic
   S. aureus presence in the nose, however, represents a known risk factor
   for S. aureus infection^[268]14,[269]39. Our study adds the
   neurological disease depression to the list of morbidities for which S.
   aureus represents a risk factor, and S. aureus nasal decolonization
   could thus also have an impact on the incidence of depression.

   Our study has limitations. We investigated a relatively small human
   cohort and relied predominantly on mechanistic studies in mice to
   support our conclusions. Our cohort of young adults without
   comorbidities cannot represent populations with socioeconomic adversity
   and adolescent or elderly depression. Although we matched key
   confounders, unmeasured factors may contribute to residual confounding.
   Furthermore, no method of S. aureus detection is entirely accurate,
   which is why we included several different methods in our study.
   Finally, it is not entirely clear how the bacteria benefit from sex
   hormone-degrading enzymes. Potentially, they utilize them as carbon
   sources or to eliminate estradiol-mediated stress^[270]40.

   In conclusion, our discovery of a significant impact of the nasal
   microbiome on a widespread neurological disease highlights the
   importance of the nose–brain axis and nasal bacterial colonization for
   such diseases.

Methods

Inclusion and ethics

   The research complied with all applicable ethical regulations. The
   human clinical study was approved by the ethics committee of Renji
   Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai,
   China (approval number KY2022-139-B). All animal procedures followed
   the ethical guidelines outlined in the Guide for the Care and Use of
   Laboratory Animals proposed by the Institute for Laboratory Animal
   Research of the National Academy of Sciences and all protocols were
   approved by the Animal Welfare Committee of Renji Hospital, Shanghai
   Jiao Tong University School of Medicine, Shanghai, China (approval
   number RJ2023-025B).

Human participants

   Healthy volunteers were recruited from the Physical Examination Center
   of Renji Hospital. Patients who had been newly diagnosed with
   depression were recruited from the Department of Psychological Medicine
   of Renji Hospital. All participants were ethnically Han Chinese. Each
   participant was provided with a detailed written questionnaire by the
   research physicians. Informed consent was obtained from all human
   research participants. Patients with depression and healthy volunteers
   were enrolled between November 2022 and November 2024. All participants
   were between 18 and 44 years of age. The patients with depression were
   diagnosed according to the Diagnostic and Statistical Manual of Mental
   Disorders (DSM-IV)^[271]41 by clinical psychologists and had no other
   psychiatric disorders or family history of any psychiatric disorders.
   The severity of depression and anxiety was evaluated using PHQ-9 and
   GAD-7 scores, respectively. The healthy controls were physically
   healthy and lacked any neurological illness or related family history,
   with PHQ-9 and GAD-7 scores both below four. Other exclusion criteria
   for both control and depression groups included nasal or oral diseases,
   previous use of any type of antidepressant medication within the past
   three months, use of antimicrobial drugs within the past four weeks,
   thyroid dysfunction, diabetes, hypertension, autoimmune diseases,
   tumours and pregnancy or lactation.

   Due to the minimal risks involved in the sampling process of this
   study, only non-economic compensation measures were adopted.
   Professional psychologists were provided for participants when they
   experienced significant discomfort or needed immediate psychological
   support during the research process.

Sample collection

   Two nasal swab samples were collected from the nasal cavity of each
   participant using Copan swabs. With the participant’s head tilted
   slightly backward, a trained physician gently inserted the swab along
   the nasal cavity to a depth of 2–3 cm without endoscopic guidance or
   local anaesthesia. The swab was rotated firmly against the mucosal
   surface for 15–20 s to ensure adequate specimen collection before
   slowly withdrawing. For microbiota analysis, the collected swabs were
   immediately immersed in 1 ml sterile saline. All collected swabs were
   maintained at 4 °C and processed within 4 h of collection. For
   metabolome analysis, the collected swabs were immediately stored at
   −80 °C without sterile saline. We collected 119 serum samples from
   participants who agreed to provide blood samples. After collection, the
   serum samples were immediately aliquoted and stored at −80 °C.

Culture-based analysis

   Nasal swabs were vortexed in 1 ml sterile saline for 2 min. Aliquots of
   100 μl from each swab were diluted, plated on 5% sheep blood agar and
   incubated at 37 °C for 24 h. Twenty-four random colonies were isolated
   and identified by matrix-assisted laser desorption
   ionization–time-of-flight mass spectrometry (MALDI-TOF-MS, Bruker
   Daltonics) in each sample.

Full-length 16S rRNA gene sequencing

   Nasal swabs were immersed in 1 ml sterile saline and vortexed for
   2 min. Each sample (500 μl volume) was centrifuged at 13,000g and 4 °C
   for 10 min. The pellets were resuspended in 180 μl Buffer ATL (QIAamp
   DNA mini kit, QIAGEN, 51306) with 5 μl lysozyme (50 mg ml^−1; Sigma,
   L6876) and 5 μl lysostaphin (1 mg ml^−1; Sigma, L4402), and incubated
   at 37 °C for 30 min. Next, DNA was extracted according to the
   manufacturer’s instructions. The full-length 16S rRNA gene was
   amplified using the primers 27F (5′-AGRGTTYGATYMTGGCTCAG-3′) and 1492R
   (5′-RGYTACCTTGTTACGACTT-3′). The amplicons were then sequenced on the
   PacBio platform. For negative controls, a full sequencing protocol was
   applied to a sterile swab.

Amplicon sequence analysis

   Sequence analysis was performed using the DADA2 workflow in the QIIME2
   software pipeline^[272]42. Initially, primers and adaptors were removed
   and sequences with a quality score of less than three or expected
   errors greater than two were filtered out, retaining sequences with
   lengths between 800 and 1,800 base pairs (bp). Further processing with
   DADA2 included the removal of duplicate sequences, learning error
   models, inferral of amplicon sequence variants (ASVs) and removal of
   chimaeras to obtain an ASV feature table. To mitigate potential
   contamination effects, all ASVs detected in the negative controls were
   removed from subsequent analyses (Supplementary Data [273]1). Batch
   correction was performed with MMUPHin^[274]43. Only ASV sequences
   present in at least two samples were retained to eliminate spurious
   features. The taxonomy of ASV sequences was analysed using RDP
   Classifier version 2.13 against the NT_16S (v20221012) database.
   Taxonomic α-diversity was estimated using Shannon and Simpson indices.
   The β-diversity between groups was measured using PCoA based on
   Bray–Curtis dissimilarity and compared using the PERMANOVA method with
   the vegan R (v2.6-4) package. Microbiome taxonomic abundance data were
   analysed using general linear models by MaAsLin2 (ref. ^[275]44),
   adjusted for demographic variables (age, sex and body mass index),
   socioeconomic factors (education and income), family relationships,
   adverse childhood experiences, social support (availability of
   supportive friends) and technical factors (batch). The P values were
   adjusted for multiple comparisons using FDR based on the
   Benjamini–Hochberg method. Differentially abundant genera and species
   were determined using FDR < 0.05. Absolute quantification of S. aureus
   genomic copies was achieved through quantitative PCR analysis utilizing
   a standard curve developed via successive 1:10 dilutions of linearized
   plasmid DNA templates containing cloned nuc sequences with
   pre-determined copy number.

Whole-genome sequencing

   A single S. aureus isolate per specimen was randomly selected for
   whole-genome sequencing. Genomic DNA was isolated from the bacterial
   cell pellets using a Bacterial DNA kit (OMEGA) according to the
   manufacturer’s instructions. Paired-end libraries with insert sizes of
   150 bp were prepared following Illumina’s standard genomic DNA library
   preparation procedure. The qualified Illumina paired-end library was
   used for Illumina NovaSeq 6000 sequencing (150 bp × 2; Shanghai
   BIOZERON Co., Ltd). The raw paired-end reads were filtered using fastp
   v0.12.5 and de novo assembly was performed using SPAdes v3.15.4. The
   resulting scaffolds were annotated using Prokka v1.14.6-. Multilocus
   sequence typing was performed using mlst v2.23.0
   ([276]https://github.com/tseemann/mlst) based on whole-genome
   sequencing data.

Non-targeted metabolomics

   Extract solution (1,000 μl; 3:1 methanol:water containing
   isotope-labelled internal standard mixture) was added to the nasal swab
   sample and the mixture was sonicated for 10 min in an ice-water bath.
   The samples were incubated for 1 h at −40 °C and centrifuged at 13,800g
   and 4 °C for 15 min. The resulting supernatant was transferred to a
   fresh glass vial for analysis. Liquid chromatography with MS/MS
   (LC–MS/MS) analyses were performed using an ultra-high-performance
   liquid chromatography (UHPLC) system (Vanquish, Thermo Fisher
   Scientific) with a UPLC HSS T3 column (2.1 mm × 100 mm; 1.8 μm) coupled
   to an Orbitrap Exploris 120 mass spectrometer (Orbitrap MS, Thermo
   Fisher Scientific). The mobile phase consisted of 5 mmol l^−1 ammonium
   acetate and 5 mmol l^−1 acetic acid in water (A) and acetonitrile (B).
   The auto-sampler temperature was 4 °C and the injection volume was
   2 μl. The mass spectrometer was used to acquire MS/MS spectra based on
   information-dependent acquisition mode in the control of the
   acquisition software (Xcalibur, Thermo Fisher Scientific). Raw data
   were converted to the mzXML format using ProteoWizard and processed
   using the XCMS R package (v3.22.0) for peak detection, extraction,
   alignment and integration^[277]45. An in-house MS2 database (BiotreeDB)
   was used for metabolite annotation. The cutoff for peak annotation was
   set at 0.3. The identification level of metabolites was annotated based
   on the method by Alseekh et al.^[278]46, and metabolites identified at
   the MS2 level with definitive KEGG compound assignments were used for
   further analysis. The raw peak areas of metabolites were normalized
   using internal standards. Differentially abundant metabolites were
   determined based on the variable importance for the projection value
   calculated by OPLS-DA analysis and FDR using the ropls R package
   (v1.30.0)^[279]47, with a cutoff of variable importance for the
   projection > 1 and FDR < 0.05. KEGG pathway enrichment analysis was
   performed using MetaboAnalyst 6.0 (ref. ^[280]48). HAllA-based
   multivariate analysis (v0.8.20) was conducted to explore associations
   between species showing differential abundance in MaAsLin2 analysis and
   metabolites meeting statistical significance criteria^[281]49.

Animals

   Specific-pathogen-free C57BL/6J mice were purchased from GemPharmatech
   and bred in-house under specific-pathogen-free conditions. The vendor
   specifically guaranteed that the mice were free of S. aureus. The mice
   were provided with food and water ad libitum and housed at consistent
   ambient temperature (22 ± 1 °C) and humidity (50 ± 5%) with a 12-h
   light–dark cycle. The mice were maintained on a ɣ-gamma-irradiated
   standard rodent diet (Suzhou Shuangshi Laboratory Animal Feed
   Technology Co., Ltd). The diet contained corn, wheat bran, soybean
   meal, fish meal, alfalfa meal, calcium hydrogen phosphate, sodium
   chloride, vitamin premix and mineral premix, with the following
   nutritional composition: crude protein, ≥200 g kg^−1; crude fat,
   ≥40 g kg^−1; crude ash, ≤80 g kg^−1; crude fibre, ≤50 g kg^−1 and
   moisture content, ≤100 g kg^−1.

Animal experiments

Nasal microbiota transplantation

   Nasal microbiota transplantation was performed in male and female
   C57BL/6J mice that were about 6 weeks old. Briefly, 10 μl antibiotic
   cocktail (0.5 g l^−1 ampicillin, 0.5 g l^−1 metronidazole and
   0.25 g l^−1 vancomycin) was instilled daily into each mouse’s nostril
   (total of 20 μl per mouse) for seven days. Nasal microbiota
   transplantation was performed three days after the last antibiotic
   instillation. Male mice received nasal transplants from male donors and
   female mice from female donors (one donor per mouse). Donor nasal
   microbiota samples were randomly selected from the healthy and
   depression cohorts, and were not expanded through cultivation before
   inoculation. Each mouse was inoculated with the microbiome of a
   different donor. The nasal microbiota sample was dropped into the
   nostrils (10 μl per nostril; 20 μl total per mouse) of
   antibiotic-treated mice every day for seven days to improve microbial
   community stability. Four mice were housed in one cage. Behavioural
   assessments were performed afterwards. To evaluate the bacterial load,
   noses were collected at different time points, homogenized in 0.5 ml
   cold PBS, and homogenates were serially diluted and plated on 5% sheep
   blood agar plates for enumeration.

   DNA was extracted according to the manufacturer’s instructions (EZNA
   soil DNA kit, Omega). The V3–V4 region of the 16S rRNA gene was
   amplified using the primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R
   (5′-GGACTACHVGGGTWTCTAAT-3′). The amplicons were then sequenced on the
   Illumina platform. Sequence analysis was performed using the DADA2
   workflow in the QIIME2 software pipeline. The taxonomy of ASV sequences
   was analysed using the SILVA v138.2 database. Taxonomic α-diversity was
   estimated using Shannon and Simpson indices. The β-diversity between
   groups was measured using PCoA based on Bray–Curtis dissimilarity and
   compared using the PERMANOVA method with the vegan R (v2.6-4) package.
   Microbiome taxonomic abundance data were analysed using general linear
   models by MaAsLin2.

Bacterial culture and nasal colonization

   S. aureus strain P24-2 (ST398) or S. epidermidis strain P24-1, which
   were isolated from the nasal microbiome of patient 24 in the depression
   cohort, were cultured in TSB at 37 °C for 8 h, harvested by
   centrifugation at 13,000g and 4 °C for 2 min, and then resuspended in
   sterile PBS. Each nostril of male and female C57BL/6J mice
   (approximately 6 weeks old) received 10 μl of an antibiotic cocktail
   (0.5 g l^−1 ampicillin, 0.5 g l^−1 metronidazole and 0.25 g l^−1
   vancomycin; 20 μl total per mouse). After three days of daily
   antibiotic pretreatment, the mice were subjected to treatment with
   10 μl bacterial solution in each nostril (total of 20 μl per mouse
   containing approximately 1 × 10^7 CFU) once every two days for seven
   total treatments. For S. aureus strain JSNZ (ST88), mice without
   antibiotic pretreatment directly received two treatments of 2.5 μl
   bacterial solution seven days apart (total of 5 μl per mouse containing
   approximately 1 × 10^6 CFU). The mice were euthanized at pre-determined
   intervals throughout the experimental period. Nose and lung tissues
   were collected and homogenized in 0.5 ml cold PBS. Caeca and faeces
   were collected and homogenized in 5 ml cold PBS. For CFU counts,
   homogenates were serially diluted and plated on 5% sheep blood agar
   plates, ChromAgar Staph aureus selective plates, or selective plates
   containing 0.5 mg ml^−1 streptomycin (for ST88), which were then
   incubated at 37 °C for 24 h. S. aureus and S. epidermidis colonies were
   identified using MALDI-TOF-MS (Bruker Daltonics).

Chronic unpredictable mild stress

   Male C57BL/6J mice were exposed to one of the following various
   low-intensity social/environmental stressors sequentially for a total
   duration of seven days based on the method by Yu and
   colleagues^[282]50. The stressors were: (1) food deprivation for 24 h,
   (2) overnight illumination for one night, (3) absence of sawdust in the
   cage for 24 h, (4) moistened sawdust for 24 h, (5) water deprivation
   for 24 h, (6) physical restraint for 6 h and (7) 45° cage‐tilt along
   the vertical axis for 3 h.

Hormone replacement therapy

   To eliminate the effects of physiological hormonal fluctuations, a
   bilateral ovariectomy was performed on female C57BL/6J mice and a
   bilateral orchidectomy on male C57BL/6J mice (all approximately 6 weeks
   old). After 14 days of postoperative care and no antibiotic
   pretreatment, 5 μl JSNZ bacterial solution (containing about
   1 × 10^6 CFU) was instilled into the right nostrils of the mice. The
   next day, the mice received hormone replacement therapy using a nasal
   hydrogel for sustained release of estradiol or testosterone^[283]51.
   The female (ovariectomy) mice received 1 μg estradiol in 5 μl in situ
   nasal hydrogel. The male (orchidectomy) mice received 0.1 mg
   testosterone in 5 μl in situ nasal hydrogel.

   To study the effects of intranasal administration of pure sex hormones,
   female mice colonized with JSNZ (ST88) received 0.1 μg estradiol or
   vehicle in 5 μl in situ nasal hydrogel every two days, and CUMS-treated
   male mice colonized with JSNZ (ST88) received 10 μg testosterone or
   vehicle in 5 μl in situ nasal hydrogel every two days.

Nasal hydrogel

   The thermosensitive in situ nasal hydrogel was prepared as follows.
   Solubilization of estradiol, testosterone or FITC was achieved through
   hydroxypropyl-β-cyclodextin (HP-β-CD) complexation, sonication (40 kHz,
   ice bath) in 2% HP-β-CD in PBS (pH 6.5), followed by magnetic stirring
   (500g, 25 °C) and purification via centrifugation. PLGA-PEG-PLGA and
   CS-PEG polymers were dissolved in chilled ultrapure water and acetate
   buffer (pH 5.0), respectively, and allowed to electrostatically
   self-assemble (2:1 vol/vol) with drug-loaded HP-β-CD under continuous
   stirring (4 °C for 6 h). The gel was reconstituted by blending
   complexes with 20% chilled poloxamer 407 (1:9 vol/vol)^[284]52, which
   is a thermosensitive molecule widely used in mucosal delivery. The
   gelation time was about 30 s at 37 °C. In vitro release studies
   employed dynamic dialysis (molecular weight cutoff, 3.5 kDa) in the
   synthetic nasal medium SNM3 (pH 6.5)^[285]53 containing 0.5% HP-β-CD,
   with sustained drug release monitored over 48 h using LC–MS/MS.
   Anaesthetized mice received unilateral intranasal administration of
   5 μl thermosensitive hydrogel (containing about 10 μg FITC) or control
   liquid formulation (FITC dissolved in 2% HP-β-CD/PBS, pH 6.5). The
   animals were euthanized at pre-determined intervals post
   administration, with major organs excised for fluorescent imaging using
   an IVIS in vivo imaging system.

Behavioural tests

   Behavioural tests were performed based on the previously reported
   methods^[286]50,[287]54.

Tail suspension test

   The TST test was performed by wrapping the tail of each mouse with
   tape, with the tip sticking out by about 1 cm, and then suspending the
   mouse head-downwards at a height of 15 cm above the floor. The animals
   were videotaped from the front for 6 min. The total duration of
   immobility within the last 4 min was analysed using the VisuTrack
   software (Shanghai XinRuan Information Technology Co., Ltd.). Duration
   of immobility was defined as the time when animals did not seem to
   struggle.

Forced swimming test

   The FST test was performed by placing individual mice in a water-filled
   cylinder (diameter of 12 cm and a height of 25 cm; water temperature of
   25 ± 1 °C). The animals were videotaped from the front for 6 min. The
   time of immobility during the last 4 min was counted using the
   VisuTrack software. It was defined as the time when the animals
   remained floating or motionless with the only observed movements being
   those necessary for keeping their balance in the water.

Open field test

   The OFT test was performed by placing the mice in the centre of an
   arena (50 cm × 50 cm × 40 cm). The animals were videotaped from above
   for 6 min. The time spent in the central area during the last 4 min was
   counted using the VisuTrack software. Additional data related to all
   OFT experiments are in Supplementary Data [288]2.

Cytokine concentration assessment

   Mice nose and lung tissues were collected and homogenized in 0.5 ml
   RIPA solution (containing 0.1 mM phenylmethylsulfonyl fluoride). After
   centrifugation at 4 °C and 10,000g for 15 min, the supernatants were
   collected and assessed using a bicinchoninic acid assay for protein
   quantification (Sangon Biotech). Concentrations of interleukin-6 and
   1β, tumour necrosis factor-α and CXCL1 were determined using an
   enzyme-linked immunosorbent assay kit (CUSABIO).

Histopathology

   Mouse noses and lungs were fixed in 4% paraformaldehyde at 4 °C for
   three days. The noses were then decalcified in 0.12 mol l^−1 EDTA
   solution (pH 7.4) for 7–14 days at room temperature. Hematoxylin and
   eosin staining was then performed following standard procedures.

   Nasal tissues were collected four days after the final colonization
   (before behavioural testing) and fixed in 4% paraformaldehyde at 4 °C
   for 72 h. Following EDTA decalcification (0.12 M, pH 7.4; 7–14 days
   with solution replacement every 48 h), the tissues were dehydrated
   through graded ethanol, cleared in xylene and paraffin-embedded for
   5-μm sectioning. Immunofluorescence protocols included sequential
   dewaxing with eco-friendly agents, EDTA-based heat-induced epitope
   retrieval (98 °C for 20 min), blocking with 10% donkey serum and
   incubation with primary antibodies—anti-IBA1 (guinea pig polyclonal
   antibody; Oasis Biofarm, OB-PGP049) or anti-OMP (rabbit polyclonal
   antibody; Bioss, bs-19568R) at a dilution of 1:100—and the fluorescent
   secondary antibodies Alexa Fluor 594 goat anti-guinea pig IgG (Oasis
   Biofarm, G-GP594; 1:200) and Alexa Fluor 488 donkey anti‐rabbit IgG
   (Thermo Fisher Scientific, A21206; 1:400). Nuclei were counterstained
   with 4′,6-diamidino-2-phenylindole (1:500), followed by mounting with
   antifade medium and fluorescence microscopy analysis.

   Lung tissues were harvested four days after the final colonization
   (before behavioural testing) and fixed in 4% paraformaldehyde at 4 °C
   for 72 h. Paraffin sections underwent sequential dewaxing with
   eco-friendly agents and graded ethanol, followed by antigen retrieval
   using either EDTA-based pressure cooking (1,200 W, 1.5 min boiling) or
   microwave-mediated retrieval (medium power, 12 min). Endogenous
   peroxidase activity was quenched with 3% H[2]O[2] before blocking with
   10% goat serum. Anti-LY6G primary antibody (rabbit monoclonal antibody;
   Abcam, ab238132, EPR22909-135; 1:500) was applied overnight at 4 °C,
   followed by corresponding horseradish peroxidase-conjugated secondary
   antibody (goat anti-rabbit IgG; Abcam, ab205718; 1:2,000) at 37 °C for
   45 min. Immunoreactivity was visualized using 3,3′-diaminobenzidine
   chromogen with reaction monitoring, counterstained with haematoxylin
   for 1 min, and differentiated with acid-alcohol. The sections were
   dehydrated, cleared and mounted with eco-friendly medium for
   bright-field microscopy analysis.

Sickness scores

   Sickness scores were determined according to Huet et al.^[289]19, where
   eight parameters (fur aspect, activity, posture, behaviour,
   respiration, chest sounds, eyes and body weight) were given a score
   between one, representing the minimum (healthy), and four, representing
   the maximum. Total scores thus ranged from eight for entirely healthy
   mice to 32.

Neurotransmitter detection

   Midbrain tissues were collected, weighed and homogenized in 80 μl
   extract solvent (0.1% formic acid in acetonitrile; precooled to
   −20 °C)/20 μl water. The samples were kept at −20 °C overnight and then
   centrifuged at 13,800g and 4 °C for 15 min. The 80-μl supernatant
   samples were incubated with 40 μl of 100 mmol l^−1 sodium carbonate
   solution and 40 μl 2% benzoyl chloride in acetonitrile solution for
   30 min. The samples were then centrifuged at 13,800g and 4 °C for
   15 min after the addition of 10 μl internal standard. Next, 40-μl
   aliquots of the supernatants were added to 20 μl water and transferred
   to an auto-sampler vial for UHPLC–MS/MS analysis (Waters ACQUITY
   Premier or SCIEX Triple Quad 6500+ MS). The mobile phase A was 0.1%
   formic acid and 1 mmol l^−1 ammonium acetate in water, and the mobile
   phase B was acetonitrile. SCIEX Analyst Work Station Software (version
   1.6.3) was employed for multiple-reaction-monitoring data processing.

Anaesthesia

   Mice received nasal antibiotics, microbiome transplants or bacteria and
   were euthanized by cervical dislocation under anaesthesia using
   isoflurane inhalation.

Serum cytokine and thyroid hormone detection

   Serum cytokine concentrations were measured using a cytokine detection
   kit (BD Biosciences). Serum concentrations of thyroid-stimulating
   hormone (TSH), free triiodothyronine (FT3) and free thyroxine (FT4)
   were determined using a Roche Cobas fully automated
   electrochemiluminescence immunoassay system.

Steroid hormone detection

   Steroid hormones were detected using LC–MS/MS. For serum or supernatant
   samples of bacterial late-exponential-phase cultures (TSB), 200 μl of
   the samples was mixed with 20 μl internal standard solution and 400 μl
   methanol, followed by vortexing for 30 s and centrifugation for 10 min
   at 13,800g and 4 °C. Next, 600 μl water was added to 500-μl supernatant
   aliquots, followed by vortexing for 30 s and centrifugation under the
   same conditions. A 950-μl aliquot of the obtained supernatant was
   further purified with solid phase extraction cartridges (Agela). The
   cartridges were washed with 200 μl methanol and then equilibrated with
   200 μl water. After sample application loading, the cartridges were
   washed with 200 μl of 10% acetonitrile in water (vol/vol) and 200 μl
   hexane. The cartridge was then rinsed with 40 μl of 90% acetonitrile in
   water (vol/vol) and 60 μl water was added to the eluent. All samples
   were shaken for 3 min, after which the mixed solution was subjected to
   UHPLC–MS/MS analysis (Waters ACQUITY UPLC/Xevo TQ-S MS). The mobile
   phase A was 0.5 mmol l^−1 ammonium fluoride in water, and the mobile
   phase B was methanol. Waters MassLynx V4.1 was employed for
   multiple-reaction-monitoring data acquisition and processing.

   For tissue samples, nose and midbrain tissues were collected, weighed
   and homogenized in 200 μl water. After the addition of 20 μl internal
   standard solution and 400 μl methanol, the samples were processed as
   described for the serum samples.

Degradation of estradiol and testosterone

   Nasal bacteria were cultured in TSB medium containing 100 ng ml^−1
   estradiol or testosterone at 37 °C. After 24 h of cultivation, the
   supernatant was collected, and testosterone and estradiol
   concentrations were determined using LC–MS/MS. For time-dependent
   degradation of testosterone and estradiol by S. aureus supernatant, the
   same conditions were used and the sampling time points were set to 0,
   24, 48 and 60 h.

RNA-sequencing

   Nasal mucosal tissues and midbrain samples were carefully isolated
   using microscopy forceps. Total RNA was extracted using TRIzol reagent
   according the manufacturer’s instructions (Invitrogen) and genomic DNA
   was removed using DNase I (TaKara). RNA-seq transcriptome libraries
   were prepared following RNA preparation with a TruSeqTM RNA sample
   preparation kit (Illumina). Paired-end libraries were sequenced by
   Illumina NovaSeq 6000 sequencing (150 bp × 2; Shanghai BIOZERON Co.,
   Ltd). The raw paired-end reads were trimmed and quality-controlled
   using Trimmomatic v0.39. Next, clean reads were separately aligned to
   the reference genome with orientation mode using the hisat2 v2.2.1
   software. FeatureCounts v2.0.3 was used to count each gene read.
   Differentially expressed genes were determined using edgeR v3.42.4
   based on FDR and fold change, with a cutoff of FDR < 0.05 and
   |log[2](fold change)| >1. KEGG pathway enrichment analysis was
   performed using clusterProfiler v4.8.2. For bacteria, RNA was prepared
   from bacteria cultured to logarithmic growth phase and RNA-seq was
   performed as described above.

Quantitative real-time PCR

   Total RNA was extracted from the midbrain using an RNeasy kit (Qiagen)
   according to the manufacturer’s instructions. The extracted RNA was
   reverse-transcribed into cDNA using a PrimeScript RT reagent kit with
   gDNA Eraser (TaKaRa). Real-time PCR was performed using the Hieff
   UNICON Universal Blue qPCR SYBR Green Master Mix (Yeasen) on a 7500
   real-time PCR system (Applied Biosystems). Gene expression was
   calculated using the 2^−ΔΔCT method.

SDR family identification using hidden Markov models

   The SDR superfamily encompasses multiple Pfam entries: PF00106,
   PF01073, PF01370, PF05368, PF08659 and PF1356 (ref. ^[290]25). Hidden
   Markov models of these Pfam profiles were extracted from the Pfam
   database and hmmer v3.0 was used to scan protein sequences of the P24-2
   (ST398) S. aureus genome for proteins belonging to these Pfam protein
   families. Proteins with an E-value of less than 1 × 10^−5 were recorded
   as positive hits. The sequences of putative SDR superfamily proteins
   were compared using blastp with the Uniprot database to exclude
   candidates with known other functions.

Heterologous expression of putative SDR proteins in E. coli

   The plasmid pET28a was used as an overexpression vector to express SDR
   proteins from S. aureus in E. coli. Genomic DNA of S. aureus P24-2 was
   used as a template to amplify putative SDR genes via PCR (primers in
   Supplementary Table [291]5). Amplified genes were cloned into the
   pET28a plasmids and the resulting plasmids were transformed into E.
   coli BL21 (DE3). The recombinant E. coli strains were cultured in 3 ml
   Luria–Bertani medium containing 100 ng ml^−1 testosterone or estradiol
   at 37 °C. When the optical density at 600 nm reached a value of
   0.6–0.8, 0.5 mM isopropyl β-d-1-thiogalactopyranoside was added to the
   medium to induce protein expression. After 24 h of cultivation, the
   supernatant was collected, and the concentrations of testosterone and
   estradiol and their degradation products androstenedione and estrone,
   respectively, were determined using LC–MS/MS.

Construction of S. aureus hsd12-deletion mutant

   The hsd12-deletion mutants were generated in the strains S. aureus
   P24-2 (ST398) and JSNZ (ST88) using the pKOR1 allelic replacement
   strategy. Briefly, sequences of approximately 1 kb flanking the hsd12
   gene were amplified by PCR. The recombinant plasmids pKOR1-HSD12 (for
   deletions in ST88 and ST398) were constructed using a
   ligation-independent cloning method and transformed into E. coli Top10
   cells as the cloning host. The plasmids purified from E. coli Top10
   cells were electroporated into S. aureus P24-2 and JSNZ using S. aureus
   RN4220 as an intermediary host. Allelic replacement was induced by
   temperature shift. The pKOR1 transformants were selected by plating on
   tryptic soy agar containing 10 μg ml^−1 chloramphenicol and incubation
   at 30 °C, followed by incubation in TSB containing 10 μg ml^−1
   chloramphenicol at 43 °C to allow plasmid integration into the
   chromosome. Non-plasmid-carrying mutants were selected by plating on
   tryptic soy agar containing 1 μg ml^−1 anhydrotetracycline. Successful
   deletion of hsd12 was verified by PCR. To verify the growth behaviour
   of the hsd12-deletion strain (SAΔhsd), the same amount
   (1 × 10^6 CFU ml^−1) of mid-logarithmic phase S. aureus P24-2 and its
   hsd12-deletion mutant were cultured in 200 μl of fresh TSB medium in
   96-well plates. The plates were incubated at 37 °C in a BioTek Synergy
   H1 multi-mode microplate reader with continuous shaking for 14 h. The
   optical density at 600 nm was measured every 15 min.

Statistics and reproducibility

   GraphPad Prism version 10.2.0 for Macintosh was used for simple
   statistical analyses including comparisons of two or more groups and
   correlation analyses. Two-group comparisons were performed with
   unpaired two-sided Student’s t-tests or Mann–Whitney tests. Comparisons
   of three or more groups were performed using a one-way or two-sided
   ANOVA, or a Kruskal–Wallis test, as appropriate, depending on the
   assessments of normal distribution using the Shapiro–Wilk test, with
   Tukey’s and Dunn’s post-tests, respectively. Spearman’s correlation was
   used to assess the association between two quantitative variables,
   interpreted according to Akoglu et al.^[292]16 and Chan et al.^[293]55,
   and Fisher’s exact tests or χ^2 tests were used to assess contingency,
   as indicated. Principal coordinate analyses were performed and
   evaluated using the vegan R (v2.6-4) package. Further tests are
   described in their specific method sections. Each independent
   experiment was performed with at least three biological replicates.
   Values were expressed as the mean ± s.d., unless otherwise indicated in
   the figure legend.

   Where indicated, randomization was performed using the randomization
   functions of Microsoft Excel 2019, which randomly assigns a number
   between zero and one to every sample, after which the numbers are
   sorted in ascending order and the top samples of the desired n
   selected. Animals were randomly assigned to experimental groups using
   computer-generated random numbers following the completion of an
   acclimation period. Experimental conditions (including bacterial
   administration sequence and behavioural testing order) were randomized
   across subjects.

   No statistical method was used to pre-determine sample size. The sample
   sizes were similar to those reported in previous
   publications^[294]9,[295]56. No data were excluded from the analyses
   except when using the described data filtering processes. The
   investigators were not blinded to allocation during experiments and
   outcome assessment, except for histological analysis, where slides were
   examined independently by a histopathologist who was blinded to the
   treatment.

Availability of materials

   All unique biological materials are available from M.L.
   (rjlimin@shsmu.edu.cn) on request.

Reporting summary

   Further information on research design is available in the [296]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [297]Supplementary Information^ (1.3MB, pdf)

   Supplementary Tables 1–5 and Supplementary Fig. 1.
   [298]Reporting Summary^ (2.9MB, pdf)
   [299]Peer Review File^ (1.1MB, pdf)
   [300]Supplementary Data 1^ (122.2KB, xlsx)

   All taxa detected in 16S rRNA gene sequencing of individual extraction
   blanks (n = 21).
   [301]Supplementary Data 2^ (88.7KB, xlsx)

   Detailed OFT data.

Source data

   [302]Source Data^ (2.1MB, xlsx)

   Statistical source data for Figs. 1–6 and Extended Data Figs. 1–10.

Acknowledgements