Childhood gut microbiome is linked to internalizing symptoms at school age via the functional connectome

Abstract

The microbiome-gut-brain-axis plays a critical role in mental health. However, research linking the microbiome to brain function is limited, particularly during development, when tremendous plasticity occurs and many mental health issues, like depression and anxiety, initially manifest. Further complicating attempts to understand interactions between the brain and microbiome is the complex and multidimensional nature of both systems. In the current observational study (N = 55), we use sparse partial least squares to identify linear combinations of brain networks (brain signatures) derived from resting state fMRI scans at age 6 years that maximally covary with internalizing symptoms at age 7.5 years, before identifying microbe abundances (microbial profiles) derived from 16S rRNA sequencing of stool samples at age 2 years that maximally covary with those brain signatures. Finally, we test whether any early microbial profiles are indirectly associated with later internalizing symptoms via the brain signatures, highlighting potential microbial programming effects. We find that microbes in the Clostridiales order and Lachnospiraceae family are associated with internalizing symptoms in middle childhood through connectivity alterations within emotion-related brain networks.

Fig. 1. Component loadings for brain signatures and microbial profiles.

A SOFA, MTL, SAL, PMN Network Signature. B Microbial Profile 1. C SOFA Between Network Signature. D Microbial Profile 2. E Microbial Profile 3. Colors (blue or salmon) and rows correspond to the brain signature that the microbial profile was derived from (blue = SOFA, MTL, SAL, PMN Network Connectivity Brain Signature; salmon = SOFA Between Network Connectivity Brain Signature). Only variables with VIP > 1 are shown. SOFA = striatal-orbitofrontal-amgydalar, MTL = medial temporal lobe, DMN = default mode network, SAL = salience, PMN = parietomedial network, VAN = ventral attention network, AUD  = auditory, CON = cingulo-opercular, SMD = somatomotor dorsal. For microbes that were unable to be classified to the genus level, the family name is shown. Source data are provided as a Source data file.

Fig. 2. Regions of interest in the networks with loadings onto the SOFA, MTL, SAL, PMN Network Connectivity Brain Signature.

Regions of interest and network assignments were derived from the Seitzman et al. atlas. SOFA striatal orbitofrontal amygdalar, MTL medial temporal lobe, DMN default mode, SAL salience, VAN ventral attention, PMN parietomedial, AUD auditory, CON cingulo-opercular, SMD dorsal somatomotor.

Fig. 3. Regions of interest in the networks with loadings onto the SOFA Between Network Connectivity Brain Signature.

Regions of interest and network assignments were derived from the Seitzman et al. atlas. SOFA = striatal orbitofrontal amygdalar, VIS = visual, FPN = frontal parietal, DAN = dorsal attention, SML = lateral somatomotor, DMN = default mode, VAN = ventral attention, CON = cingulo-opercular, SMD = dorsal somatomotor.

Fig. 4. Significant association between Faith’s phylogenetic diversity and SOFA Between Network Connectivity Brain Signature scores.

Each dot represents a participant. Faith’s Phylogenetic Diversity and SOFA Between Network Connectivity Brain Signature scores are shown after partialling out variance accounted for by covariates. Predictors in the model included Faith’s phylogenetic diversity, child sex, gestational age at birth, birthweight, delivery mode, mean FD, maternal education and consumption of fiber, total fat and polyunsaturated fat. The dark purple line represents the predicted best-fit function predicting x from y using a linear model, and the lighter purple shaded area around the line is the 95% CI level for that predicted best-fit function. Dots have been jittered horizontally (along the x-axis) by 0.3 to increase visibility of individual data points. Source data are provided as a Source data file.

Fig. 5. Diagram summarizing results from the mediation analysis testing the effect of Microbial Profile 3 on child internalizing symptoms through the SOFA Between Network Connectivity Brain Signature.

Covariates include child sex, gestational age at birth, birthweight, delivery mode, mean FD, maternal education, and consumption of fiber, total fat and polyunsaturated fat for both the a and b paths.