Case study: a probiotic and the gut-brain axis

Following the gut–brain chain: what a probiotic changes, and how the host responds

How Microbiome Xplorer combines standardised sampling, strain-resolved metagenomics, metatranscriptomics, metabolomics and host markers to show which communities change, which functions they carry and how the host responds, an illustrative study.

Illustrative case study. This scenario shows how Microbiome Xplorer is applied to a gut–brain question. It does not describe a specific client project or product: study size, results and timelines are indicative and depend on the intervention, the matrices and the study design. Microbiome Xplorer is a research service; it does not diagnose or treat any condition.

1Sample

Gut, oral, skin, vaginal or any matrix, collected and stabilised to standard

2Communities

Which taxa and strains change, in absolute terms

3Functions

Which functions they carry, and which are switched on

4Host

Which signals reach the host, and what it means

At a glance

Question

Does a probiotic act on the gut–brain axis: and through which microbial functions and host signals?

Client profile

Nutrition company running a randomised, placebo-controlled trial of a three-strain probiotic

Participants

120 healthy adults with moderate perceived stress; 8 weeks of intake + 4-week washout; 112 completers

Matrices

Stool (4 timepoints), saliva (oral microbiome, cortisol), blood

Layers

Strain-resolved shotgun metagenomics, metatranscriptomics, 16S (saliva), stool and plasma metabolomics, host markers; WGS of the product strains

Volume

≈1,200 molecular profiles

Outcome

Transient engraftment; butyrate, GABA and indole functions ↑; tryptophan shifted towards indoles; no change in inflammation or cortisol; modest stress effect, mediation exploratory

Duration

≈ 14 weeks from last sample to final report

The challenge

A nutrition company had developed a three-strain probiotic for stress and well-being and was running a randomised, double-blind, placebo-controlled trial in 120 healthy adults with moderate perceived stress. The clinical endpoints were in place. What was missing was the mechanism: does the product actually act through the microbiota–gut–brain axis¹, and how?

The brief: show which communities change, which functions they carry and how the host responds, and where the chain breaks.

01Sample: the right matrices, collected the right way

The trial was the client’s; the sampling design was built with them before the first participant was enrolled. Each matrix was chosen for the question it answers: stool for the gut community and its products, saliva for the oral microbiome and the cortisol awakening response, and blood for circulating metabolites and inflammatory markers.

02Which communities change

Microbiome Xplorer combines layers according to the question.

Number of molecular profiles per layer: about 1,200 in total — **Figure 1. What was measured.** Number of molecular profiles generated per layer, about 1,200 in total for 112 participants who completed the trial: stool metagenomes at four timepoints, stool and plasma metabolomes, saliva microbiomes, and stool metatranscriptomes for a subset of 40 participants at weeks 0 and 8. Indicative volumes.

PCoA with overlapping arms and unchanged Shannon diversity at week 8 — **Figure 2. Community composition and diversity at week 8.** Left: each point is one participant’s gut community, placed so that similar communities sit close together (PCoA on Bray-Curtis distances). The two arms largely overlap: the probiotic shifts overall composition only slightly (PERMANOVA R² 0.8%, p = 0.03), as expected in a healthy adult gut. Right: Shannon diversity does not change (p = 0.62). Simulated data matching the case study.

Bar chart of participants carrying the product strains at week 8 and after washout — **Figure 3. Engraftment of the product strains.** Share of participants in whom the product strains were detected, using strain-specific markers that tell them apart from related resident strains. 82% of the probiotic arm carried them at week 8, only 9% four weeks after stopping, and no one in the placebo arm: engraftment is real but transient. Indicative results.

04How the host responds

A microbiome change matters only if the host responds. Each link of the chain, from product to perceived stress, was tested rather than assumed.

In the gut

1Product strains engraft

Detected in 82% at week 8

9% after washout

Supported · transient

2Community shifts

Small overall shift (R² 0.8%)

Butyrate producers ↑

Modest

3Functions switch on

Butyrate, GABA, indole pathways ↑ (DNA and RNA)

Supported

In the host

4Metabolites reach the host

Faecal butyrate ↑ 15%

Plasma IPA ↑

Kynurenine/tryptophan ↓

Supported

5Host signalling

CRP, IL-6, LBP unchanged

Cortisol response n.s.

Not shown

6Perceived stress

PSS −1.6 vs placebo

~24% mediated by butyrate

Exploratory

Figure 4. The gut-brain chain, link by link. Each box is one link between the product and perceived stress, with its evidence and a verdict. The first four links are supported, from engraftment to metabolites reaching the circulation. Host signalling (inflammation, cortisol) did not change, so the final link to perceived stress remains exploratory. Dashed arrows mark links not demonstrated at this dose and duration. Indicative results.

Interpretation. The data support a coherent but partial mechanism. The strains engraft transiently and raise the community’s capacity, and actual activity, for butyrate, GABA and indole production; these products reach the circulation and shift tryptophan metabolism away from the kynurenine branch. The host signals expected downstream, systemic inflammation and the cortisol response, did not change at this dose and duration. The gut-to-host link is established; the host-to-brain link is not yet.

What the data support
Statement	Evidence	Verdict
“Product strains colonise the gut during intake”	Strain-level detection in 82% at week 8; 9% after washout	Supported: transient
“Increases the microbiota’s butyrate production”	Metagenome ↑, metatranscriptome ↑, faecal butyrate ↑	Supported: 3 layers
“Shifts tryptophan metabolism towards indoles”	Plasma IPA ↑; kynurenine/tryptophan ↓; tryptophanase expression ↑	Supported: 2 layers
“Gut-specific, without disturbing the resident microbiome”	Oral microbiome and resistome unchanged; no acquired resistance genes in the strains	Supported
“Reduces systemic inflammation”	CRP, IL-6, LBP unchanged	Not supported
“Dampens the cortisol stress response”	−6%; 95% CI −14% to +2%	Not supported
“Reduces perceived stress via the microbiome”	PSS −1.6 vs placebo; ~24% mediated by butyrate (exploratory)	Hypothesis: to confirm in a dedicated trial

In the EU, health claims on foods and food supplements must be authorised under Regulation (EC) No 1924/2006 after EFSA assessment of human efficacy data. Mechanistic microbiome data strengthen the scientific rationale, trial design and scientific communication; they do not by themselves establish a claimed health effect.

Faecal butyrate rises 15% at week 8 in the probiotic arm and returns towards baseline after washout — **Figure 5. Faecal butyrate over the trial.** Mean and 95% CI, indexed to 100 at baseline. In the probiotic arm, butyrate rises by 15% at the end of intake (p = 0.02) and returns towards baseline after the 4-week washout, in line with transient engraftment; under placebo it stays flat. Simulated data matching the case study.

Changes versus placebo across communities, functions, metabolites and host markers — **Figure 6. What changed, layer by layer.** Change versus placebo at week 8 for each readout, grouped from microbes to host: butyrate producers (absolute abundance), butyrate synthesis genes (DNA), expressed butyrate, GABA and indole pathways (RNA, fold change), metabolites reaching the circulation, and the cortisol response. Points to the right of zero increased, to the left decreased. The cortisol bar is a 95% CI that crosses zero: no effect is shown. Indicative results.

What the client received

An integrated dataset: taxonomic, strain, functional, metabolite and host-marker tables linked by participant and timepoint, with raw data (FASTQ, MS files).

A strain-tracking report documenting the engraftment and persistence of each product strain.

A mechanism map showing which links of the gut–brain chain are supported, partial or open.

A statistical report following the pre-specified analysis plan, including the exploratory mediation.

Recommendations for the confirmatory trial: a higher dose or longer intake to test host signalling; stratification by baseline community to account for colonisation resistance; butyrate and IPA as pre-specified mechanistic endpoints.