By Michael Widdowson
Principal supervisor: Professor Jakob Stokholm, COPSAC
Principal co-supervisor: Professor Bjørn H. Ebdrup, CNSR
Co-supervisors: Professor Morten Arendt Rasmussen, KU FOOD and Senior scientist, PhD Shiraz Shah, COPSAC
Chairperson: Kenneth K. Barfod, PhD , Associate Professor, Section of Food, Microbiology, Gut health, and Fermentation, Department of Food Science, University of Copenhagen, Copenhagen, Denmark
Opponent: John F. Cryan, PhD, Professor and Chair at the Deptartment of Anatomy & Neuroscience, University College Cork, Ireland
Opponent: Martin F. Laursen, PhD, Associate Professor, National Food Institute Research Group for Gut, Microbes and Health DTU Microbes Initiative National Food institute, Technical University of Denmark, Denmark
Summary
Postnatal life represents a critical window for neurodevelopment, overlapping with the initial colonization and rapid diversification of the gut microbiome, through which neural development may be modulated via the microbiota-gut-brain axis, potentially shaping long-term trajectories of neurodevelopment. Early-life exposures, such as mode of delivery, can profoundly shape microbial composition and, in turn, influence the host’s health. Alterations in the gut microbiome have been linked to various neuropsychiatric and neurodegenerative conditions; however, the timing-specificity and compositional features underlying microbial influences on early neurodevelopment remain incompletely understood.
The work presented in this PhD thesis aimed to identify key windows during the first year of life when early microbial influences on neurodevelopment a decade later may be particularly important. Furthermore, we sought to characterize the perinatal factors shaping early microbial composition and to elucidate microbe-host metabolic pathways that may mediate associations between the gut microbiome and long-term neurodevelopmental outcomes, including neuropsychiatric diagnoses, cognitive and behavioral functioning, and brain structure encompassing both morphology and white matter microstructural and macrostructural properties.
This PhD thesis is based on data from the Copenhagen Prospective Studies on Asthma in Childhood 2010 (COPSAC2010) longitudinal birth cohort, comprising 700 Danish mother-child dyads. Participants were enrolled during the 24th week of pregnancy and prospectively followed throughout childhood and early adolescence with frequent, detailed clinical assessments at the COPSAC research clinic. Child gut microbiome (16S rRNA gene sequencing and/or shotgun metagenomics) data were generated from fecal samples collected at 1 week, 1 month, and 1 year of age and served as the primary exposure, alongside other early-life factors influencing microbial composition. At 10 years of age, children underwent an comprehensive neurodevelopmental evaluation as part of the COpenhagen Prospective Study on Neuro-PSYCHiatric Development (COPSYCH), which included structured assessments of neuropsychiatric diagnoses, cognitive and behavioral domains, and MRI scans conducted by clinicians and research staff trained in child and adolescent psychiatry. External cohorts, COPSAC2000 and the iPSYCH twin study, which include comparable early-life data and neurodevelopmental end points, were used for replication in Paper I.
The thesis contains three scientific papers:
In Paper I, we investigated differences in the early-life gut microbiome in relation to later ADHD diagnosis. Using 16S rRNA gene sequencing at 1 week, 1 month, and 1 year in the COPSAC2010 cohort, we found that children who developed ADHD by age 10 years had a higher relative abundance of Bifidobacterium in their 1-week gut microbiome. Bifidobacterium relative abundance was positively associated with levels of the tryptophan-derived metabolite indole-3-lactic acid (ILA) in neonatal dried blood spots, and ILA mediated the relationship between Bifidobacterium and ADHD. Certain species of Bifidobacterium are known to encode the ALDH gene required for ILA production. However, given the limited species-level resolution of 16S rRNA, we applied a three-step triangulation approach to evaluate the biological plausibility of our findings. First, we identified the specific Bifidobacterium amplicon sequence variant (ASV) driving the association with ADHD. Second, we aligned the 1-week Bifidobacterium ASV nucleotide sequence to NCBI reference sequences, observing 100% identity and coverage with ALDH-encoding Bifidobacterium species. Lastly, because this same ASV was also present at 1 month in our data, we linked it to species-level metagenomic profiles at that time point. The ASV relative abundance correlated specifically and strongly with ALDH-encoding Bifidobacterium species – the only Bifidobacterium taxa capable of synthesizing ILA – rather than with non-ALDH species. Together, these findings indicate that the observed ILA and clinical associations are driven by the functionally relevant ALDH-positive subset, rather than diluted by other Bifidobacterium species. The association between ILA and ADHD was replicated in two independent cohorts (COPSAC2000 and the iPSYCH twin study). Overall, this work shows that Bifidobacterium relative abundance within the 1-week gut microbiome, but not at 1 month or 1 year, is associated with an increased risk of ADHD, and that this relationship is mediated by the Bifidobacterium-derived metabolite ILA.
In Paper II, we examined whether mode of delivery (vaginal vs. C-section) influences the brain’s morphology (i.e., total white matter surface area and mean cortical thickness) and white matter microstructure and macrostructure (i.e., apparent fiber density, fiber bundle cross-section, and the combined fiber density and cross-section) at 10 years of age, using structural magnetic resonance imaging and fixel-based analysis (FBA), respectively. Because mode of delivery largely determines the initial seeding of the gut microbiome, we additionally investigated the role of the microbiome by deriving C-section gut microbiome scores from supervised models using 16S rRNA data at 1 week, 1 month, and 1 year of age to predict these brain outcomes. While mode of delivery was not significantly associated with surface area or cortical thickness, it showed significant associations with all FBA metrics after adjustment for sex, intracranial volume, and relevant early-life factors. In both crude and adjusted models using the C-section microbiome scores as predictors of FBA outcomes, a consistent signal emerged from the 1-month microbiome score for fiber density. However, this association was not robust enough to pass multiple testing corrections. A differential abundance analysis was conducted to identify genera that overlapped as strong predictors of both C-section and these FBA metrics. These results identified a set of specific genera that appear to be shared between birth mode and white matter microstructure and macrostructural features. Collectively, these findings suggest that white matter structure may be significantly shaped by mode of delivery, specifically C-section, but this association is not likely explained by early-life perturbations of the gut microbiome.
In Paper III, we examined the joint variation between widespread measures of brain development (i.e., total surface area and mean cortical thickness) at age 10 years and the early-life gut microbiome (16S rRNA) at 1 week, 1 month, and 1 year of age, to test for timing-specific effects. Beta diversity analyses indicated structured, abundance-weighted associations between subsets of taxa within the gut microbiome in early life and both global brain measures. Using a sparse partial least squares (sPLS) framework with nested cross-validation, we modeled the multivariate covariation between microbial genera and both global and regional brain morphology. Significant covariational patterns were observed specifically at 1 week of age. Higher relative abundance of Streptococcus and Staphylococcus in the neonatal gut was associated with widespread reductions in cortical surface area. For cortical thickness, higher Streptococcus and lower relative abundance of a genus within the family Acutalibacteraceae (annotated as UMGS1071) were associated with widespread increases in cortical thickness. Among the genera associated with both brain features, Streptococcus appeared to be a dominant driver, as supported by differential abundance analyses and mutual adjustment in linear regression models. Unbiased component scores from the sPLS model were used to derive microbiome and brain scores for downstream analyses. Both microbiome scores were associated with being born by C-section, and the surface-area microbiome score partially mediated the relationship between delivery mode and total surface area. The cortical-thickness brain score was associated with fewer CBCL internalizing problems (i.e., fewer somatic complaints, less anxiety/depression, and potentially less withdrawn behavior). Collectively, these results indicate that the relative abundance of Streptococcus within the neonatal 1-week gut microbiome may contribute to lasting effects on brain structure and perhaps behavior.
In summary, the findings presented in this PhD thesis highlight a critical window in early life – around one week of age – during which microbial influences may have lasting effects on neurodevelopment. Future work can build on these insights to better elucidate the underlying mechanisms and to inform strategies for promoting healthy neurodevelopment through early-life microbial modulation.
