The Microbiota Shapes Central Nervous System Myelination in Early Life

ABSTRACT

Maturation of the gut microbiota coincides with neurodevelopmental processes such as myelination, essential for efficient neural signal transmission. While a role for the microbiome in regulating adult prefrontal cortex (PFC) myelination is known, its effects on early‐life myelin formation, growth, and integrity remain unclear. Using a cross‐species approach in germ‐free (GF) mice and zebrafish, we examined how the microbiota influences early myelination and neural development. Multi‐system, multi‐level analyses showed that the microbiota impacts glial maturation and myelination across species. In GF mice, we observed sex‐ and age‐dependent alterations in pathways linked to neuronal activity and myelination, with myelin‐related transcriptomic changes correlating with functional shifts in neurotransmission‐ and metabolism‐related metabolites over time. Myelin growth and integrity were also affected in a sex‐ and time‐dependent manner. As microglia regulate neuronal activity and engulf myelin, we examined microbiota–microglia interactions and found altered expression of genes involved in microglia maturation and synaptic pruning in both species. In zebrafish larvae, the microbiota influenced the spatial distribution of microglia and oligodendrocytes within the brain and spinal cord. These findings reveal conserved microbiota‐mediated modulation of neuronal activity, myelination, and glial maturation in early life, providing a foundation for future studies into these mechanisms.

Gut microbiota shapes brain development by regulating myelination and glial cell maturation in early life. Using germ‐free (GF) mice and zebrafish, this study reveals sex‐ and age‐dependent effects on myelin growth, integrity, and related gene expression. Findings highlight conserved microbiota‐driven mechanisms linking metabolism, neuronal activity, and immune function during critical windows of neurodevelopment.

1. Introduction

Myelination of axons within the central nervous system (CNS) is a life‐long process, however, during development, it plays a crucial role in shaping neural circuitry and facilitating efficient signal transmission [1, 2]. This complex process involves the insulation of axons by myelin sheaths, a specialized, proteolipid‐rich membrane produced by oligodendrocytes [3]. Beyond enhancing nerve impulse conduction velocity, myelination plays a fundamental role in neural plasticity and is essential for sensory, motor, and cognitive function [4, 5, 6, 7]. In mammals, the prefrontal cortex (PFC) is a dynamically evolving brain region undergoing myelination at later stages, rendering it particularly susceptible to external influences [8]. The PFC is integral to neuronal circuitry and essential for emotional regulation. Specifically, dysfunction of the PFC, including perturbed myelination, is strongly associated with various neuropsychiatric and neurodevelopmental disorders, including autism spectrum disorder [9]. While conventional paradigms have predominantly emphasized genetic and environmental factors in regulating CNS myelination, emerging evidence implicates the gut microbiota and its signaling molecules as pivotal mediators of this process, potentially influencing neural circuit structure and function [10, 11, 12, 13].

The gut microbiota, encompassing diverse microbial communities residing within the gastrointestinal tract, has gained considerable attention for its profound influence on various aspects of host physiology, spanning metabolism, immune function, and neurodevelopment [14, 15]. Recent studies have begun unravelling the intricate communication networks between the gut microbiota and the CNS, collectively termed the microbiota‐gut‐brain axis. Within this axis, mounting evidence suggests that the gut microbiota mediates region‐specific activity, glial maturation including microglia, and myelination within the CNS, prompting inquiries into the mechanisms underlying this regulatory crosstalk and its implications for brain function and behavior [11, 13, 16, 17, 18].

Gnotobiotic model systems, particularly rodents, have been instrumental in advancing our understanding of how the gut microbiota influences myelination. However, zebrafish have recently emerged as a promising model organism to complement rodent models and mechanistically investigate microbiota‐brain interactions [16]. Seminal studies in germ‐free (GF) mice have revealed the essential role of the gut microbiota in the dynamic regulation of cortical myelination in adulthood, with notable sex‐specific effects observed, where GF male mice exhibit upregulation of activity‐induced and myelin‐related genes, and hypermyelination of axons in the PFC [11, 13]. Intriguingly, the impact of microbes on myelination varies with age, as colonization of GF mice with a normal gut microbiota at weaning can recover myelin‐related transcriptomic changes, but not the hypermyelination phenotype [11]. Moreover, alterations in microbiota composition and associated microbial‐derived metabolites have been found to impact neural circuit formation and myelination, subsequently impacting complex behaviors such as social and anxiety‐like behaviors [12, 13, 19]. However, the timing and extent of interactions between the microbiota and myelin across various developmental stages and levels of evolutionary complexity remain unclear.

To this end, we hypothesized that the developmental trajectory of the microbiota aligns with myelination and neural circuit formation in early life, and perturbations during this period may have implications for myelin and neural circuit refinement across species. We employed a cross‐species approach to characterize fundamental aspects of microbiota‐myelin communication in early life. Through integrated transcriptomic and metabolomics analyses, we assessed the temporal trajectory of myelin formation at four distinct timepoints in both sexes in mice during early life. Moreover, to verify our findings, we turned to the zebrafish model to parse potential mechanisms involved.

2. Results

2.1. The Microbiota Influences the Developmental Trajectory of the PFC Transcriptome Associated with Neuronal Activity and Myelin‐Related Pathways

2.1.1. Young Adulthood

Previous research has indicated that the gut microbiota influences young adult cortical myelination [10, 12], with only GF male mice showing elevated expression of activity‐induced and myelin‐related genes, as well as hypermyelination of axons [11, 13]. To further interrogate the molecular mechanisms, we built upon our previously published dataset by conducting targeted transcriptomic analysis on the prefrontal cortex (PFC) of young adult mice. This analysis included both conventionally‐raised (CON) and germ‐free (GF) mice (Figure 1A–D), based on new observations derived from the reanalysis of our previously published dataset [11]. To determine the functional implications of differentially expressed genes, we employed targeted enrichment analysis using a gene ontology (GO) database. In addition to the previously shown changes in myelination and neuronal activity, differential gene expression was seen in several pathways relevant to oligodendrocyte maturation, myelin regulation, nodal physiology, and gamma‐aminobutyric acid (GABA) neurotransmission (Figure 1B,C). We examined 66 genes associated with myelin and oligodendrocyte maturation terms and found that 22 genes (33.3%) were upregulated in the PFC of GF mice (Figure 1D). Specifically, out of 66 genes linked to myelin regulation, 9 genes (13.6%) were upregulated (Nkx6.2, Egr2, Myrf, Mag, Rara, S100b, Sirt2, Mtmr2) and 2 genes (Ctsc, Rxrg) were downregulated. Furthermore, analysis of 34 genes relating to activity‐induced genes revealed that 31 genes (91%) were upregulated in GF mice. Among the 40 genes associated with the node of Ranvier term, 8 genes (20%) were differentially expressed in GF mice compared to controls, with 7 of them being upregulated. These genes encode voltage‐dependent ion channels (Kcna1, Scna1) and structural components of the nodes (Ermn, Mbp, Tubb4a, Mag). These findings collectively suggest that young adult GF male mice experience widespread changes in genes associated with myelination and neuronal activity. However, the influence of microbial signals on myelin formation and the temporal neurodevelopmental stage during which sex differences emerge remained unknown.

FIGURE 1.

The microbiota alters the developmental trajectory of PFC transcriptome associated with oligodendrocyte maturation and myelin‐related pathways. (A) Schematic representation of the experimental timeline for RNASeq of PFC. (B) Volcano plot illustrating the most differentially expressed genes (DEGs) in germ‐free (GF) male mice in young adulthood (p < 0.05). Genes (dots) to the left of “0” are downregulated, while genes to the right are upregulated. (C) Targeted enrichment analysis of genes associated with myelin‐related pathways in young adult GF mice relative to controls. Enrichment analysis was conducted by hypergeometric testing, where a solid grey circle indicates p < 0.05. (D) Heatmap showing the expression fold change (logFC) and significance of differentially expressed genes associated with myelin maturation, activity‐induced, myelin regulation, and nodal physiology of young adult GF versus control mice. Colour indicates logFC, with purple indicating increased expression and blue indicating decreased expression in GF mice compared to controls. (E) Volcano plot illustrating the most differentially expressed genes (DEGs) in GF mice during each of the four early developmental timepoints (p < 0.05). Genes (dots) to the left of “0” are downregulated, while genes to the right are upregulated. (F) Targeted enrichment analysis of genes associated with myelin‐related pathways in early development of GF mice relative to controls. Enrichment analysis was conducted by hypergeometric testing, where a solid grey circle indicates p < 0.05. (G–I) Heatmap showing the expression fold change (logFC) and significance of differentially expressed genes associated with (G) myelin maturation, (H) activity‐induced, myelin regulation, and (I) nodal physiology of GF versus control mice during all four early developmental timepoints. Colour indicates logFC, with purple indicating increased expression and blue indicating decreased expression in GF mice compared to controls. (J) Venn diagrams illustrating the overlap of significantly different genes (p < 0.05) between GF and CON groups across different developmental timepoints and sexes. Sample size: Young Adulthood: n = 8 mice/group, Early Development: n = 8–10 mice/group. Microbiota; *p < 0.05, **p < 0.01, ***p < 0.001, Sex; #p < 0.05, ##p < 0.01. Abbreviations: P: postnatal day, F: females, M: males, GF: germ‐free. Detailed statistical analysis can be found in Supplementary Table S1.

2.1.2. Early Development

Regarding early development, previous studies have shown that colonizing GF male mice during the weaning period recovered myelin‐related transcriptomic changes but did not resolve the hypermyelination phenotype observed in young adulthood [11]. This suggests the developing microbiota plays a significant role in shaping myelin formation in a sex‐specific manner. To investigate whether developmental myelination is regulated by microbiota‐derived signals, we performed transcriptomic analysis on the PFC in male and female conventionally raised and GF mice at four developmental timepoints (PN postnatal days [P]2, P8, P14, and P21). Principal component analysis (PCA) followed by PERMANOVA indicated a significant effect of microbiota and sex in all the timepoints assessed (Figure S1A). Notably, the effect of sex and microbiota increased alongside sexual development and maturation of the gut microbiota, with the weaning period (P21) representing the phase characterized by the most significant microbiota effects.

Germ‐free status resulted in differential expression of >771 genes, compared to controls, with the P2 and P21 timepoints representing the most marked effects, irrespective of sex (Figure 1E and Figure S1B). Untargeted enrichment analysis revealed disruptions in pathways relevant to lipid regulation, including regulation of long‐chain fatty acid import, ganglioside metabolic and catabolic processes, fatty acid beta‐oxidation, lipid catabolic processes, and glycosphingolipid metabolic processes (Figure S1C). Notably, the pathways related to long‐chain fatty acid import and fatty acid beta‐oxidation were disrupted at P21, both of which are critical for providing the energetic fuel required for oligodendrocyte maturation and function [20]. These changes in lipid metabolism suggest that GF status impacts the energetic support needed for myelination. Additionally, significant alterations were observed in glycosphingolipid metabolism, ganglioside metabolism and catabolism, and ceramide catabolic processes, all of which are key to myelin structure and function [21]. Taken together, these findings suggest that GF status influences lipid regulation in a manner highly specific to developmental timing, with disruptions at P21 reflecting alterations in both energetic and structural lipid pathways essential for myelination.

To further investigate the relevance of these changes for myelin and neuronal activity, which influences myelin formation, we performed targeted enrichment using the same GO terms utilized in the young adult male dataset (Figure 1F). Similar to the young adult dataset, we observed changes in myelin‐related genes, specifically in GF males at the P14 and P21 timepoints. Additionally, genes within pathways relating to microglia dynamics were also significantly perturbed in GF males across the P8 and P14 timepoints. Analysis of genes associated with myelin and oligodendrocyte maturation terms revealed time‐dependent and sex‐specific effects of GF status (Figure 1G). While myelin‐related gene expression did not exhibit robust upregulation as observed in young adulthood, signatures indicative of perturbed oligodendrocyte maturation were evident. At the P8 timepoint, male GF mice exhibited an overall downregulation of maturation markers compared to controls, which seems to be compensated by an upregulation at P14 and a subsequent downregulation again at P21.

In terms of activity‐induced genes, we observed an overall downregulation in GF‐reared mice in the early P2 and P8 timepoints, irrespective of sex. However, at the P14 timepoint, there was an upregulation of activity‐induced genes (Spry4, Nr4a2) in GF mice of both sexes, indicating that heightened neuronal activity may be contributing to the time‐dependent changes in myelination. Consistent with this, as neuronal circuit formation progressed, a significant upregulation of activity‐induced genes was observed in GF males at P21 (Klf4, Tsc22d3, Trib1, Sik1, Btg2, Csrnp1, Dusp1, Per1, Npas4), reminiscent of those observed in GF males in young adulthood (Figure 1H).

Alongside temporal changes in myelin and activity‐related gene expression, at P21 GF male mice exhibited upregulation of genes responsible for myelin regulation including those observed in adult GF male mice (Sirt2, Egr2). Akt1, a central component of the PI3K‐Akt signaling pathway, was also upregulated in GF males compared to controls at the P8 and P21 timepoints. Interestingly, across all developmental timepoints, male GF mice exhibited upregulation of the cell–cell adhesion molecule Jam2, which is expressed in the somatodendritic compartment and inhibits aberrant myelination ensuring selective axonal myelination. This upregulation suggests a potential compensatory mechanism for myelin regulation that merits further exploration [22]. Further, in terms of nodal structure and function sex‐ and time‐dependent effects were observed, with GF males exhibiting downregulation of nodal genes at P2, P8, and P21 timepoints but showing overall increased expression at P14, with significant effects in the nodal and myelin structural component Ermn (Figure 1I). Taken together, these results highlight the complexity of developmental gene expression, underscoring the need for further research to determine the direct implications of these dynamic changes on myelination.

Utilizing Venn diagrams, we evaluated the overlap of significantly different genes (p < 0.05) between GF and CON across sexes and developmental timepoints (Figure 1J). Several overlapping genes were observed across development for both sexes. In GF males, six genes were shared across all four developmental timepoints, three of which are associated with cell signaling and neural development (Jam2, Gstt2, Dclk3) [22, 23, 24, 25]. Similarly, GF females shared 3 genes (Lmbr1, Rnf32, Pla2g) across all four timepoints, each linked to cellular and developmental functions [26, 27, 28]. These findings suggest that the microbiota influences certain conserved cellular processes throughout development. When comparing young adult and P21 males, 171 differentially expressed genes were shared between these two timepoints. Among these shared genes, 8 genes (4.67%) were related to myelin and oligodendrocyte maturation, and 16 (9.36%) were associated with neuronal activity (Table S1). Taken together, these findings underscore the impact of the microbiota on cortical myelination and neuronal activity throughout different stages of life. They reveal intricate, sex‐ and time‐dependent effects of microbiota‐derived signals on shaping neural activity and myelination patterns during development.

2.2. The Microbiota Influences the Developing PFC Metabolome, Which Correlates with Transcriptomic Signatures

Given that the early gut microbiota has been implicated in a variety of neurodevelopmental processes, we postulated that transcriptomic alterations in the PFC could arise from alterations in key brain metabolites. Therefore, we performed untargeted metabolomic profiling of the PFC in both GF and conventionally‐raised mice across the same four developmental timepoints (Figure 2A). Following Principal Component Analysis (PCA) and subsequent PERMANOVA, a significant effect of microbiota was evident across all developmental timepoints, regardless of sex (Figure 2B), mirroring the transcriptome profile. Particularly noteworthy was the increased effect of microbiota at the P2 timepoint. Despite the absence of an effect of sex across timepoints, an interaction between sex and microbiota was observed at P8. Overall, we detected a total of 112 metabolites, among which 36 metabolites were significantly altered in GF mice (p < 0.05) (Figure 2B,C).

FIGURE 2.

The microbiota influences the developing PFC metabolome, which correlates with transcriptomic signatures. (A) Schematic representation of the experimental timeline for metabolomics of PFC. (B) Principal Component Analysis (PCA) displaying differences in the PFC metabolome of germ‐free (GF) and conventionally‐raised (CON) in each of the early developmental timepoints assessed. Data analysed using PERMANOVA analysis. (C) Heatmap demonstrating the concentration (z‐scored) of metabolites that are differentially abundant in GF versus CON mice in at least one of the assessed timepoints (p < 0.05). Each square represents one mouse. (D) Untargeted enrichment analysis of metabolites that were significantly altered in GF compared to CON mice during early developmental timepoints. (E) Heatmap demonstrating the concentration (z‐score) of metabolites that are differentially abundant in young adult GF versus CON mice (p < 0.05). Each square represents one mouse. (F,G) Venn diagrams illustrating the overlap of significantly different metabolites (p < 0.05) between GF and CON groups across different developmental timepoints and sexes. (H) Schematic illustrating the procedure involved in multi‐omics analysis. (I) Representative schematic of glutamine metabolism in the tripartite synapse. Data presented as Mean ± SEM. Sample size: Young Adulthood: n = 8 mice/group, Early Development: n = 8‐10 mice/group. Microbiota; *p < 0.05, **p < 0.01, ***p < 0.001, Sex; #p < 0.05, ##p < 0.01. Detailed statistical analysis can be found in Supplementary Table S1.

Untargeted enrichment analysis revealed that the absence of microbiota in early life disrupted several pathways critical for maintaining homeostatic brain function (Figure 2D and Figure S2A). Specifically, pathways associated with glutamate, arginine, and aspartate metabolism were most significantly affected. Focusing on P8 in males and P14 in females, there was a significant enrichment of pathways relevant to glutathione, D‐glutamine, and D‐glutamate metabolism. Coinciding with the disruption of these pathways was the enrichment of interconnected processes, including nitrogen, alanine, aspartate, and glutamate metabolism, which are crucial for the maintenance of oxidative stress, neuronal function, and neurotransmission. Additionally, arginine biosynthesis was another important pathway perturbed by GF status, with males exhibiting enrichment across all timepoints, except for P14. Furthermore, there was significant enrichment of arginine and proline metabolism, with males exhibiting disturbances during the early P2 and P8 timepoints, whereas females showed effects at the later P14 and P21 timepoints.

We next investigated whether changes in metabolite profiles were evident in young adult male GF mice (Figure 2E and Figure S2B,C). Utilizing untargeted metabolic profiling, we detected a total of 120 metabolites. While Principal Component Analysis (PCA) and subsequent PERMANOVA analysis did not reveal a significant overall effect of microbiota, we observed alterations in 7 metabolites in adult GF mice. Notably, disrupted metabolites included glutamine, asparagine, and N‐acetyl‐ornithine. Moreover, similar to the developmental dataset, untargeted enrichment analysis revealed disruption of pathways linked to nitrogen metabolism, arginine biosynthesis, and alanine, aspartate, and glutamate metabolism in GF mice, with the latter exhibiting significant effects (Figure S2C). To investigate potential differences in metabolites between GF and CON mice across development, we used Venn diagrams to identify shared metabolites between experimental timepoints (Figure 2F). In contrast to transcriptomic signatures, we found limited shared metabolites across development for both male and female GF mice, suggesting that developmental timepoint has a strong effect on metabolite profiles. Notably, when comparing young adult and P21 GF mice, only one overlapping metabolite, glutamine, was identified (Figure 2G). Taken together, these findings demonstrate the dynamic influence of the microbiota on biochemical profiles crucial for neurotransmission, nitrogen metabolism, and cell signaling pathways within the brain across lifespan.

While it is known that the absence of the microbiota in early life disrupts metabolic profiles in the developing brain, the interaction between myelin‐ and activity‐related genes and metabolites remains largely unexplored [29, 30]. To address this gap, we employed a multi‐omics integration approach to examine the association between the targeted myelin‐related transcriptome and metabolome profile in the PFC, leveraging the KEGG database [31] (Figure 2H). Through joint‐pathway analysis, we identified enriched metabolites associated with neurotransmission, cellular metabolism, protein modification, and oxidative stress response. These metabolites showed significant correlations with functionally relevant genes identified in our targeted enrichment analysis (Figure S2D). Notably, disruptions in both metabolite and transcript pathways were central to the composition of the tripartite GABAergic and glutamate synapse in GF mice (Figure 2I). Additionally, alterations in metabolites, particularly amino acid derivatives relevant to myelin regulation and cellular metabolism, were also observed (Figure S2E).

2.3. The Microbiota Induces Temporal Changes in Myelin Maturation, Growth, and Integrity

Given the temporal increase in activity‐induced and myelin‐related gene expression across developmental and young adult stages, and previous findings of increased myelination in male young adult mice within the mPFC, our subsequent analysis focused on assessing the impact of GF status on oligodendrocyte maturation at P21, a critical period when both myelination and neural development are actively progressing [32, 33]. Using the pan‐oligodendrocyte marker Olig2 and PDGFR⍺, a marker of oligodendrocyte progenitor cells, we found that GF mice exhibited no deficits in either the quantity of Olig2 or PDGFR⍺ positive cells, nor in the ratio of oligodendrocytes (Olig2+) expressing PDGFR⍺ within the mPFC (Figure 3A–E). Quantification of CNPase, a marker indicative of mature myelin, was performed across the entire PFC, striatum, and other regions of interest to mirror the scope of our transcriptomic analysis. This revealed a significant increase in the percentage of CNPase per area in GF male mice, suggesting that oligodendrocyte maturation and myelination are altered in a sex‐specific manner (Figure 3F–H).

FIGURE 3.

The microbiota influences myelination. (A) Schematic representation of PFC region of interest. Olig2+/PDGFR⍺+ analysis was performed on the mPFC: anterior cingulate (AC), prelimbic (PL), and infralimbic (IL) (green outline). CNPase staining was performed across prefrontal cortex, including grey matter, white matter (corpus callosum), and the striatum (caudate and putamen), as well as other regions of interest, outlined in magenta. (B–D) Total number of (B) Olig2+ (C) PDGFR⍺+ and (D) Olig2+/PDGFR⍺+ cells per mm² within the mPFC of germ‐free (GF) and conventionally‐raised (CON) at P21. (E) Representative images of pan oligodendrocytes (Olig2⁺; green) and oligodendrocyte progenitor cells (PDGFR⍺⁺; magenta) immunofluorescence in the mPFC of GF and CON mice at P21. Scale bar, 100 µm. (F) Representative images of mature oligodendrocytes (CNPase⁺; white) immunofluorescence across GF and CON brain sections from rostral to caudal at P21. Magenta lines illustrate areas of interest. Scale bar, 200 µm. (G) Percentage of CNPase per mm² from rostral to caudal brains in GF and CON mice at P21. (H) Total percentage of CNPase per mm² in the PFC of GF and CON mice at P21. Data presented as Mean ± SEM. Sample size: Early Development P21: n = 6 mice/group. Microbiota; *p < 0.05, **p < 0.01, ***p < 0.001, Sex; #p < 0.05, ##p < 0.01. Detailed statistical analysis can be found in Supplementary Table S1.

Subsequently, we investigated if these alterations could result in changes in myelin at the ultrastructural level. Quantification of the number of myelinated axons throughout development revealed subtle reductions in the number of myelinated axons in GF male mice compared to controls, with significant effects emerging at P14 (Figure 4A,B). Axon diameter, which has previously been shown to influence myelin sheath thickness, was not affected by GF status at P21 (Figure 4C). Further characterization of the growth and integrity of myelinated axons involved quantifying the number of outfoldings, bleb‐like structures which protrude from the compact myelin sheath, and the unravelling of myelin, characterized by separation of the myelin laminae [34].Interestingly, during early myelination, GF mice displayed no discernible differences in the number of normally myelinated axons compared to controls (Figure 4D,E).

FIGURE 4.

The microbiota induces temporal changes in myelin growth and integrity. (A) Electron micrographs of axons in the mPFC of germ‐free (GF) and conventionally‐raised (CON) mice. Scale bar, 1 µm. (B) Total number of myelinated axons in the mPFC of GF and CON mice throughout the early life developmental timepoints. (C) Average axon diameter of myelinated axons in the mPFC of GF and CON mice at P21. (D) Representative images of myelin abnormalities (arrowheads) in GF male mice at P21. Scale bar, 1 µm. (E) Proportion of axons with and without outfolding and unravellings per mm² in the mPFC of GF and CON mice during the early developmental timepoints. (F) Average inner tongue thickness in GF and CON mice at P21. (G) Scatter plot of mean inner tongue thickness per axon diameter in the PFC of GF and CON mice at P21. (H) Representative electron micrographs of GF and CON mice at P21. (Green; inner tongue region). Scale bar, 0.2 µm. (I)Average number of laminae per myelinated axon in the mPFC of GF and CON mice during the early developmental timepoints. (J) Average myelin thickness in GF and CON mice at P21. (K‐L) Myelin thickness versus axon diameter in the PFC of GF and CON mice at P21. Data presented as Mean ± SEM. Sample size: Early Development: n = 3 mice/group, >50 axons/per mouse. Microbiota; *p < 0.05, **p < 0.01, ***p < 0.001, Sex; #p < 0.05, ##p < 0.01. Detailed statistical analysis can be found in Supplementary Table S1.

Furthermore, GF mice displayed no deficits in the inner tongue area (uncompacted myelin) when compared to controls, implying that GF status does not appear to affect the structural integrity of developmental myelination (Figure 4F,G). Subsequently, we assessed myelin growth, and while we did not observe significant effects indicative of hypermyelination, we did observe subtle increases in the number of laminae and myelin sheath thickness in GF males (Figure 4H–L). Taken together, these findings indicate that the microbiota influences the timing of myelin maturation and growth in male mice. Furthermore, the period from postnatal days 14 to 21 (P14‐P21) appears to be an important window for microbiota‐driven regulation of myelination.

2.4. The Microbiota Modulates Oligodendrocyte Maturation and Myelination in a GF Zebrafish Model of Microbiota Perturbation

While previous research in rodents has implicated the microbiota in myelination, the translatability and mechanistic underpinnings remain underexplored. Therefore, we utilized GF zebrafish larvae as a mechanistic model to characterize the impact of the microbiota on CNS myelin formation across species. Morphological assessment of GF larvae demonstrated no gross morphological defects, though they did exhibit increased whole‐body and head area compared to conventionally‐raised larvae at 5 dpf (Figure S3A,B). Initial characterization of the GF phenotype in zebrafish involved analysis of myelin‐related gene expression (Figure 5A). Similar to GF mice, larvae demonstrated an upregulation of markers associated with oligodendrocyte differentiation (nkx2.2b), myelin formation and structure (mbp, olig2, plp1, mag, Cx47.1), and neuronal activity (gad1). Interestingly, GF larvae also exhibited an upregulation of markers relating to synaptic pruning (c1qb), microglial and astrocytic function (csfr1a, s100b). To confirm our findings at the protein level, we performed whole‐body live imaging of zebrafish larvae expressing a myelin fluorescent reporter [Tg(mbp:memScarlet)] at 5 dpf and found that GF larvae exhibited increased myelination relative to controls (Figure 5B–D and Figure S3C). Overall, these findings suggest that the effects of the microbiota on neurodevelopment may be conserved across species.

FIGURE 5.

The microbiota modulates oligodendrocyte maturation and myelination across species. (A) Relative gene expression (RT‐qPCR) of glia, myelin, and neurotransmitter markers in GF and conventionally‐raised (CON) whole larvae at 5 dpf. (B) Schematic of the lateral larval zebrafish showing the region of interest and the method for calculating the g‐ratio. Imaging was conducted at the urogenital opening level, around somite 15. (C) Representative images of whole‐body myelin (white—Tg(mbp:memScarlet); nacre) in GF and CON larvae at 5 dpf. Scale bar, 300 µm. (D) Quantification of the total area of myelin (µm²) proportional to larval length at 5 dpf. (E) Super‐resolution confocal live‐imaging depicting the diameter of the Mauthner axon (magenta—Tg(hspGFF62A:Gal4)); Tg(UAS:mem‐Scarlet)) with myelination (green—Tg(mbp:eGFP‐CAAX)) at somite 15 at 5 dpf. The white boxed region depicts the area where analysis was performed (somite 15). Scale bar, 10 µm. (F,G) Quantification of Mauthner (F) axon diameter growth and (G) number of synaptic boutons at somite 15 at 5 dpf. (H) Representative images of the Mauthner axon tracing (region outlined with magenta) along with myelin sheath tracing (region outlined with green) for both GF and CON larvae at 5 dpf. Scale bar, 10 µm. (I–K) Quantification of (I) myelin diameter (J) myelin thickness, and (K) (g‐ratio) in GF larvae compared to controls at 5 dpf. Data presented as Mean ± SEM. Sample size: RT‐qPCR; n = 10–13, Whole body myelin; n = 21–26 larvae, confocal Imaging; n = 16–18 axons from individual larvae. Microbiota; *p < 0.05, **p < 0.01, ***p < 0.001. Detailed statistical analysis can be found in Supplementary Table S1.

To investigate the influence of the microbiota on axonal diameter and myelin sheath development in 5 dpf zebrafish, we performed super‐resolution live‐imaging of fluorescent reporters expressed in the Mauthner neuron [Tg(hspGFF62A:Gal4); Tg(UAS:mem‐Scarlet)] and myelin sheath [Tg(mbp:eGFP‐CAAX)] (Figure 5E). Myelination along the Mauthner axon is easily identifiable, beginning at approximately 2.5 dpf with substantial myelination observed at 3.5 dpf. Following this initial myelination phase, both axonal diameter and myelin thickness continue to grow [35, 36]. Consistent with findings in GF mice, our analysis revealed that GF larvae did not display any differences in axon diameter compared to controls (Figure 5F). Additionally, we evaluated the number of synaptic boutons, which enable neurotransmission along the Mauthner axon, and found no significant difference between GF and control larvae (Figure 5G). Subsequently, we assessed myelin diameter and thickness along the Mauthner axon and found that GF larvae exhibited increased myelin diameter and thickness compared to controls, indicating that the observed increased myelin thickness was independent of axonal diameter (Figure 5H–K). Correspondingly, GF zebrafish demonstrated an increase in total fluorescence intensity of the myelin sheath (Figure S3.3D). To further understand the relationship between axon diameter and myelin thickness, we calculated the g‐ratio, which is a ratio of the axon diameter relative to the diameter of the axon plus myelin sheath. Notably, GF larvae showed a reduced g‐ratio (thicker myelin) compared to controls, whereby a lower g‐ratio is thought to enhance conduction velocities due to increased sheath thickness [37] (Figure 5K). Taken together, our findings demonstrate for the first time that the microbiota consistently modulates myelination across species.

2.5. The Microbiota Modulates Microglial Maturation and Localization Across Species

Microglia, the innate immune cells of the brain, play a crucial role in sensing neural activity, regulating developmental myelination, and shaping neuronal circuits accordingly [38, 39]. Notably, studies in GF mice and zebrafish have shown that the gut microbiota is required for microglia maturation and function [17, 19]. Thus, using targeted transcriptomic analysis, we investigated the functional state of microglia in our mouse datasets. During development, we identified a significant loss of the microglial homeostatic signature in males but not females at P8 (Figure 6A). Specifically, GF males demonstrated an overall downregulation of homeostatic genes (Siglech, P2ry12, P2ry13, Mertk, Hexb, Cx3cr1) compared to controls. Similarly, at P8, we observed a decrease in markers relevant to microglia priming and activation. This effect persisted at P14 with both male and female GF mice exhibiting overall decreased levels compared to their respective controls. Due to the role of microglia in regulating myelination and synaptic circuits, we evaluated the effects of GF status on the genes associated with the term synapse pruning. Interestingly, we observed a significant reduction in pruning gene expression levels in P8 GF male mice (Itgam, C3, Cx3cr1, C1qc, C1qb, C1qa) compared to controls. Similar to the developmental dataset, adult male GF mice also disrupted microglia function, at least on a transcriptomic level, with GF mice demonstrating disrupted expression of markers relevant to activation and priming (Lgals3bp) and markers of synaptic pruning (Plxnc1, C1qa, C1qb) (Figure 6B). Taken together, these results support previous findings in GF mice suggesting the microbiota impacts the transcriptomic profile of microglia function in a sex‐specific and time‐dependent manner.

FIGURE 6.

The microbiota modulates microglial homeostasis and localization across species. (A‐B) Heatmap showing the expression fold change (logFC) and significance of differentially expressed genes associated with microglia homeostasis, activation and priming, and synaptic pruning in germ‐free (GF) versus conventionally‐raised (CON) mice during (A) early development and (B) adulthood. Colour indicates logFC, with purple indicating increased expression and blue indicating decreased expression in GF mice compared to controls. (C) Representative images of microglia/macrophages (magenta; Tg(mpeg1:eGFP) in the head of GF and CON larvae at 5 dpf. Scale bar 50 µm. (D,E) Quantification of the total number of (D) mepg1⁺ cells and the (E) total area and total area occupied per mepg1⁺ cell within the head of GF and CON larvae at 5 dpf. (F) Representative images of mepg1⁺ cells (magenta; Tg(mpeg1:eGFP) and oligodendrocytes (green; Tg(mbp:mScarlet)) in the dorsal and ventral spinal cord of GF and CON larvae at 5 dpf. Scale bar 300 µm. (G) Quantification of the total number of oligodendrocytes and microglia within the dorsal and ventral spinal cord of GF and CON larvae at 5 dpf. (H) Quantification of the total volume of oligodendrocyte and microglia within the dorsal and ventral spinal cord of GF and CON larvae at 5 dpf. (I) Schematic illustrating the region of interest, with emphasis on the lateral and dorsal spinal tracts as well as the peripheral tract. (J) Representative images of microglia/macrophages (magenta; Tg(mpeg1:eGFP) and oligodendrocytes (green; Tg(mbp:mScarlet)) in the spinal cord of GF and CON larvae at 5 dpf. Scale bar 50 µm. (K) Quantification of the total number of microglia‐oligodendrocyte contacts within the dorsal and ventral spinal cord of GF and CON larvae at 5 dpf. Data presented as Mean ± SEM. Sample size: Adulthood: n = 8 mice/group, Early Development: n = 8–10 mice/group, Zebrafish Head: n = 33–47 larvae. Zebrafish whole spinal cord: n = 21–23 larvae, Zebrafish microglia‐oligodendrocyte contacts: n = 30–45 larvae, Microbiota; *p < 0.05, **p < 0.01, ***p < 0.001. Abbreviations: P: postnatal day, F: females, M: males, GF: germ‐free. Detailed statistical analysis can be found in Supplementary Table S1.

Specifically in zebrafish, microbiota‐derived signals are necessary for microglia‐mediated synaptic refinement of neural circuits involved in complex behaviors [19]. Thus, we investigated whether the microbiota influences the relationship between oligodendrocytes and microglia during development. To do so, we performed live‐imaging of fluorescent reporters for microglia/macrophages [Tg(mpeg1:eGFP)] and oligodendrocytes [Tg(mbp:mScarlet)] in the spinal cord and head of 5 dpf larvae. Whole brain microglia counts demonstrated that GF larvae exhibited reduced numbers compared to controls (Figure 6C,D). Because microglia display an immature phenotype in GF models, we assessed the total volume occupied by microglia. Interestingly, GF microglia also demonstrated a reduced total volume occupied per cell (Figure 6E), potentially suggesting alterations in activation state or function. However, in the dorsal and ventral spinal cord tracts, there was no difference in the number of microglia or oligodendrocytes between GF larvae relative to controls (Figure 6F,G). Further, the total volume occupied by microglia and oligodendrocytes demonstrated no significant effect in either of the spinal cord tracts (Figure 6H). As microglia are motile cells, we sought to determine the extent of their surveillance along myelinated regions. To investigate this, we conducted live imaging and quantified the total number of microglia‐oligodendrocyte contacts within a segment of the spinal cord in both GF and conventional larvae at 5 dpf. We found that spinal cord microglia in GF larvae established less contacts with oligodendrocytes during development, indicating that microbial signals may be involved in microglia recruitment to oligodendrocytes within the dorsal spinal cord tract (Figure 6I,J). The results demonstrate that the microbiota does not influence the total number of oligodendrocytes or microglia but does influence the localization of microglia within the spinal cord. Further, the microbiota was required for microglia cell morphology and brain localization in early life which may have enduring effects on neural connectivity, myelination, and subsequent brain function.

3. Discussion

Adopting a cross‐species approach to characterize microbial regulation of myelination processes enhances confidence in the translational relevance of our findings. Here we demonstrated that such a strategy lays the foundation for an increased understanding of neurodevelopmental processes relevant to disrupted myelination‐microbiota interactions. Specifically, we show that the ablation of microbiota‐derived signals during early life has enduring effects on neuronal activity, myelination and microglial maturation in a sex and time‐dependent manner. Furthermore, to our knowledge, these findings are among the first to demonstrate that the impact of microbiota perturbations on myelination is conserved across species during early development.

Expanding on previous research, we demonstrated that the microbiota not only impacts myelination and neural activity‐induced pathways but also influences critical components involved in nodal physiology and neurotransmission in the adult PFC, implying dynamic regulation of the structure and function of myelinated axons in response to microbial signals. Previously, colonization of GF mice with a normal gut microbiota reversed the transcriptomic signatures relating to myelin but not the hypermyelination phenotype, suggesting a critical window of microbial influence [11]. Investigation of the developmental trajectory of the PFC transcriptome reflects the dynamic interplay between intrinsic developmental programs and microbial cues in shaping the maturation of myelination and neuronal circuits. Although subtle, our findings align with the young adult GF phenotype, demonstrating that GF males exhibited time‐specific effects that emerge during development. Specifically, we observed that coinciding with the maturation of the gut microbiota and neurodevelopment in early life, GF male mice exhibit dysregulation of pathways relating to myelin‐related maturation and neuronal activity. To our knowledge, this is the first temporal exploration of the brain transcriptome in GF mice across critical windows of neurodevelopment. Our data demonstrate that the P14–P21 period represents a critical, sensitive window, particularly during the pre‐weaning phase, where microbiota maturation, immune development, and intrinsic myelination programs intersect to shape brain development. This aligns with existing literature highlighting the importance of microbiota perturbations during this period, with significant implications for immune priming and neurodevelopmental outcomes [40, 41, 42, 43].

Consistent with the transcriptomic signatures, the gut microbiota influenced the developing PFC metabolome, leading to significant dysregulation of crucial biochemical pathways. Employing a multi‐omics approach, we identified enriched metabolites closely associated with neurotransmission, cellular metabolism, protein modification, and oxidative stress response to be perturbed in GF mice. Importantly, these metabolites exhibited significant correlations with functionally relevant genes identified through targeted enrichment analysis, reinforcing the concept that microbially derived signals influence transcriptomic signatures linked to myelin formation and activity‐related pathways during early life disruptions were observed in pathways related to GABA and glutamate synapses in GF mice, aligning with previous research that highlights the microbiota's profound impact on neurotransmitter activity and cellular metabolism [44]. This supports the broader view that the microbiota plays a vital role in shaping brain function and behavior [12, 13, 17, 30, 45]. Although temporal profiling of the brain metabolome in GF mice during early life has been limited, our data strengthen the concept that microbial‐derived signals intrinsically influence critical developmental processes, with neuronal activity and cellular metabolism likely serving as key mechanisms. Notably, it is also important to recognize that the gut microbiota produces a diverse array of bioactive molecules beyond metabolites that may also influence activity and myelin‐related pathways. While the mechanisms underlying these influences remain to be fully elucidated, they may involve direct pathways—such as crossing the blood–brain barrier to modulate neuronal activity—or indirect routes, including regulation of immune responses and hormonal signaling [46].

Our findings also indicate that the microbiota orchestrates temporal changes in oligodendrocyte maturation. Surprisingly, myelin ultrastructural analysis revealed subtle reductions in the number of myelinated axons in male GF mice, with no changes in axon diameter, indicating that structural integrity remains intact. However, the absence of the microbiota did subtly influence myelin growth in a sex and time‐dependent manner. While the hypermyelination phenotype wasn't evident during development, we observed slight increases in the number of laminae and myelin sheath thickness in GF males at P21. The second and third postnatal weeks are when synaptogenesis, neural circuit formation, and myelination processes are most active and dynamically regulated [9]. As neural circuits mature myelination proceeds, these critical processes may go unchecked in GF male mice and therefore result in a strong and hardwired hypermyelination phenotype later in life. Importantly, myelination occurs in a spatiotemporal manner and can even continue in the PFC in humans until the third decade of life making it extremely susceptible to environmental influences such as the gut microbiota [47].

Myelination is tightly regulated by activity‐dependent mechanisms that facilitate the dynamic adaptation of neural circuits to environmental stimuli [48]. Evidence supports the microbiota's role in regulating neuronal activity, with GF mice exhibiting heightened neuronal activity in specific brain regions such as the amygdala and PFC [11, 49]. Consequently, it is reasonable to suggest that although GF male mice may maintain intact structural connectivity, alterations in functional connectivity may occur due to network properties impacting neural transmission speed. Interestingly, specific bacterial species, including prominent genera like Bifidobacterium, dominantly found in the developing gut microbiota, have been implicated in modulating neuronal circuits [50].

Paralleling studies in zebrafish with those in mice offers unique advantages in understanding complex biological processes with both a translational and evolutionary‐informed lens [51, 52]. Consistent with observations in rodents, we noted significant upregulation of myelin, glia, and neurotransmitter‐related gene expression in GF zebrafish larvae as early as 5 dpf. By combining in‐vivo imaging, we evaluated the growth and integrity of the myelinated Mauthner axon in the zebrafish spinal cord. Our data show that similar to mice, the microbiota in zebrafish does not influence axon diameter but significantly influences myelin thickness and associated g‐ratios, underscoring a conserved influence of microbial signals on myelin architecture across species.

Microglia are key regulators of neuronal activity, synaptic pruning, and myelination [38, 39]. Previous research demonstrates that the gut microbiota regulates the maturation and function of microglia in a sex and age‐dependent manner [17]. Our data demonstrate that GF mice displayed sex‐ and time‐dependent disruption of microglia homeostatic signatures alongside significant reductions in genes associated with activation and synaptic pruning. Specifically, we found that microglia were more profoundly perturbed in male mice during early life, underscoring their heightened vulnerability to microbial influence. This is consistent with prior findings that microglial perturbations are more severe in male embryos and female adults [53].

Recent work indicates that microglia are dispensable for initial developmental myelination but are crucial for regulating myelin growth and preventing hypermyelination later in life [34]. Additionally, microglia phagocytose developmental myelination in an activity‐dependent manner [54]. Therefore, we hypothesized that deficits in microglia maturation and localization due to GF status could impair myelin regulation and neuronal function. Furthermore, similar to previous findings, we observed that the microbiota influences microglia localization in the brain of zebrafish larvae, which may impact myelination and network formation [19]. However, it remains unclear whether these changes occur after microglia infiltrate the brain or due to impaired recruitment into the CNS. We also demonstrate that while microbiota does not affect microglia or oligodendrocyte numbers in the spinal cord, it reduces their co‐localization, suggesting GF status disrupts microglial migration in early life. Together, we postulate that altered brain localization and lack of reactive functional microglia may contribute to the perturbed pruning, myelin alterations and dysfunctional neurotransmission. Ultimately, this may affect the development of mature and functional neural networks and alter behavioral development [38, 54].

These findings underscore the critical role of sex in microbiota‐mediated neurodevelopment, with sex‐specific differences emerging during key developmental windows. The observed sex dimorphism in transcriptomic signatures, along with subtler metabolomic changes, suggests that microbial signals may differentially influence pathways such as microglial activation and immune maturation in a sex‐dependent manner. Supporting evidence indicates that males may be more susceptible to microbial influences, though the underlying mechanisms remain unclear [55, 56]. Notably, sex hormones like estrogens and androgens are known to be modulated by the gut microbiota, and these hormones can further influence myelination, microglial activity, and neuronal circuit refinement, potentially mediating microbiota‐derived effects. This highlights the need for further investigation into these hormonal pathways throughout life. Additionally, microglial sex dimorphism, which affects neuroinflammatory responses and synaptic pruning, may also contribute to divergent developmental trajectories in males and females [43].

The timing of microbial influence is particularly critical, as early life represents a key window for microbiota maturation, immune priming, and myelination. However, from an evolutionary and developmental perspective, zebrafish may differ, with sex effects often only emerging later due to environmental sex determination and distinct developmental pathways [57, 58, 59]. This suggests that, while our findings on myelination may be conserved across species, there could also be species‐specific effects related to developmental timing and sex dimorphism. Understanding whether these sex‐ and species‐specific effects primarily result from microbiota‐driven modulation of immune development, direct impacts on myelination pathways, or a combination of both remains an important area for future research.

It is worth acknowledging that there are a number of limitations to the current study. Firstly, transcriptomic, metabolomic, and CNPase assessments were performed on the entire PFC. The measurement of CNPase across broad cortical regions, including the PFC, striatum, and other subcortical areas, was intended to parallel the transcriptomic data, allowing comparison of RNA expression with protein levels and oligodendrocyte numbers. While efforts were made to focus on the PFC through careful dissection, some surrounding tissues may have been included in the omics datasets. Immunohistochemical assessments of Olig2 and PDGFR⍺, as well as TEM analysis, were restricted to the mPFC, but subregion differentiation was not performed. We acknowledge this as a limitation, as regional heterogeneity within the PFC and white matter may influence the findings. Furthermore, our study primarily examined early life and young male adult stages (P14–P21 and male P70), which limits our understanding of how microbiota influences myelin remodeling and neural plasticity across the lifespan of each sex. Future studies should include aged male and female animals to evaluate the long‐term impacts on myelin turnover, remodelling, and functional connectivity, providing insight into how microbiota may influence neurodegenerative and neuropsychiatric processes later in life. Studies involving GF zebrafish often include conventionalized controls (CV), in which larvae undergo the same germ‐free derivation protocol but are colonized at hatching or at a later time point after sterilization [60]. These studies have consistently demonstrated that the sterilization process does not have deleterious effects when compared to controls. Based on this evidence, we did not include such a control group in our study. However, the absence of this additional control is a potential limitation.

4. Conclusion

Taken together, our study characterizes the transcriptomic and metabolomic profiles of the PFC that are responsive to microbial signals during early development. Furthermore, we reveal that the maturation of the early‐life gut microbiota coincides with key processes relevant to myelination. We demonstrate that the microbiota influences neural activity and myelination, shaping the connectivity of the developing brain with lasting implications for brain regions crucial for neurocognitive function. By leveraging cross‐species approaches, we expand the current understanding of these key processes and demonstrate for the first time that microbial regulation of myelination is conserved across species, thereby strengthening the robustness and applicability of these findings. Moving forward, further research should focus on directly assessing myelin turnover and maintenance, as well as the regulatory roles of microglia and neuronal activity. To establish the causal role of specific microbial signals during early development, future studies should involve colonizing GF animals with single microbes or defined consortia throughout life. This could reveal whether reintroducing certain bacteria during critical windows rescues transcriptomic and metabolomic signatures related to myelination, microglial function, and neuronal activity. Targeted supplementation, along with tracking microbial molecules over time, could help elucidate their neuroactive potential across species. Additionally, single‐cell time lapse imaging whilst modulating neuronal activity during sensitive periods may clarify the relationship between oligodendrocytes and microglia and whether neuronal activity mediates microbiota effects on myelination. Importantly, these approaches will provide further understanding of sex and cross‐species variability of microbial sensitivity. Improved understanding of these mechanistic underpinnings across diverse species will enhance the translatability of findings to clinical settings and facilitate the development of microbiota‐based therapeutic interventions for disorders associated with altered neuronal function and myelination.

5. Materials and Methods

5.1. Mouse Husbandry

C57/BL6 mice breeding pairs were obtained from Taconic Biosciences, and subsequent generations were used for experiments. Germ‐free and conventionally raised mice were initially housed in breeding pairs for two weeks, after which females were individually housed until birth (designated postnatal day P0). Mice were maintained on a 12‐h light:dark cycle under controlled temperature and humidity conditions (21 ± 1°C, 55–60%), with ad libitum access to standard rodent chow (Special Diet Services, Product Code 801010) and autoclaved water. Experimental analyses included no more than two animals of each sex from any given litter. All experiments adhered to the European Directive 2010/63/EC, met the requirements of the S.I. No. 543 of 2012, and were approved by the Animal Experimentation Ethics Committee of University College Cork and the Health Products Regulatory Authority (HPRA AE19130 P047/118).

5.2. Tissue Collection

Mice designated for fresh culls were sacrificed by decapitation between 10 am and 12 pm. Trunk blood was collected in tubes containing EDTA and centrifuged at 3500 × g for 15 min at 4°C, after which plasma was collected. Brain regions were rapidly dissected on ice. Crude isolation of the whole PFC and surrounding cortical regions was performed by first removing the olfactory bulbs as in previous studies [11, 61]. Briefly, a coronal cut was made at the anterior extent of the brain to define the dissection boundary (approx. Bregma 2.93–1.69 mm). The ventral portion of this section (olfactory bulb, nucleus accumbens, piriform cortex, and other subcortical regions) was carefully discarded. The remaining dorsal cortical tissue, encompassing the prefrontal cortex, orbital cortex, and adjacent cortical areas, was isolated for subsequent analyses. This approach aimed to capture a broad region of the prefrontal cortex relevant to our study while minimizing contamination from ventral and subcortical structures. Notably, this would have included both grey and white matter tissue. The length of the ileum and colon, as well as the weights of the caecum and spleen, were measured, and regions of interest were collected. All biological samples were promptly snap‐frozen on dry ice and stored at −80°C until further analysis.

5.3. RNA‐Sequencing

Young Adulthood: Transcriptomic data from young adult male P70 mice was utilized from a previous study [11], and analyzed using the same bioinformatic analysis discussed below. Early Development: RNA was extracted from the PFC using the RNeasy Plus Universal Mini Kit (Qiagen), according to the manufacturer's instructions. RNA quality and concentrations were then measured using a Nanodrop ND‐1000 (Thermo Scientific). Strand‐Specific mRNA sequencing was conducted by Azenta on the Illumina NovaSeq 6000 (S4 flow cell, 2 x 150 bp Pair‐End (PE) configuration). Quality of sequences was assessed using FastQC v.0.11.8 [62]. Reads were aligned to the mouse genome: GRCm39 [63] using STAR v2.7.10a [64], and sequence reads overlapping genes were counted using HTSeq 2.0.3 [65]. Transcriptomic data are publicly available on the ArrayExpress collection on BioStudies accession number: E‐MTAB‐16319.

Bioinformatics analysis was conducted using R (version 4.3.2) with the RStudio GUI (version 2022.07.2 build 576). Differential expression analysis was performed on normalized gene expression values (TMM + voom) using limma eBayes [66]. This analysis compared gene expression profiles between germ‐free mice and their age‐ and sex‐matched controls. To investigate multivariate, whole compositional differences in the transcriptome, we used the adonis2 PERMANOVA implementation from the vegan library across microbiota, age, and sex categories. To control for multiple comparisons, the Benjamini‐Hochberg post hoc procedure was applied with a false discovery rate (FDR) q‐value of 0.1 as a threshold [67]. For targeted pathway enrichment analysis, we selected relevant terms from the Gene Ontology Database [68] as well as manual annotation and data mining in PubMed (http://www.ncbi.nlm.nih.gov/pubmed) to identify candidate genes. We then evaluated enrichment using the hypergeometric test implemented as phyper in the base R stats package.

5.4. Metabolomics

Metabolomics was conducted by MS‐Omics (Copenhagen). Young adult P70 metabolomic samples were utilized from a previous study [61], and analyzed using the same bioinformatic analysis discussed below. Briefly, samples were acidified using hydrochloric acid, and deuterium‐labelled internal standards were added. All samples were analyzed in a randomized order using a UPLC system (Vanquish, Thermo Fisher Scientific) coupled with a high‐resolution quadrupole‐orbitrap mass spectrometer (Orbitrap Exploris 240 MS, Thermo Fisher Scientific). Ionization was achieved with an electrospray ionization interface operated in positive and negative ionization modes under polarity switching. A QC sample was analyzed. Data were processed using Compound Discoverer 3.3 (ThermoFisher Scientific) and Skyline 22.1 (MacCoss Lab Software). Peaks were quantified using area under the curve (AUC). Further bioinformatics analysis was conducted using R (version 4.3.2) with the RStudio GUI (version 2022.07.2 build 576). Principal component analysis (PCA) was conducted on CLR‐transformed values [69]. PERMANOVA implementation from the vegan library was utilized to identify differentially abundant metabolites per age and sex. To correct for multiple testing, the Benjamini‐Hochberg post hoc procedure was performed with a false discovery rate (FDR) q‐value of 0.1 as a threshold [67]. Untargeted enrichment analysis was performed on differentially abundant metabolites (p < 0.05) using MetaboAnalyst with the murine KEGG library as a reference [70]. Metabolomics data is publicly available via Zenodo (https://doi.org/10.5281/zenodo.18147079).

5.5. Multi‐Omics Integration Analysis

Using Anansi (ANnotation based ANalysis of Specific Interactions), associations between transcriptomics and metabolomics were estimated (q < 0.1) [31]. In summary, all relevant transcripts were converted to KEGG orthologs (KOs), and metabolites were similarly translated into KEGG compounds. Subsequently, linear mixed‐effects models were fitted to estimate the associations between concerted transcript‐metabolite pairs that were adjacent in a KEGG pathway.

5.6. Immunofluorescence Staining

Mice were deeply anaesthetized with sodium pentobarbital (90 mg/kg) and transcardially perfused with PBS followed by 4% PFA. Brains were post‐fixed in PFA for a maximum of 4 h and cryoprotected in a sucrose gradient for 48 h at 4°C before being snap‐frozen on dry ice and stored at −80°C. Brains were sectioned at 20 µm using a Leica CM1900 cryostat and stored in cryoprotectant at −20°C. Sections from throughout the PFC (Bregma 2.93–1.69 mm) were subjected to antigen retrieval by heating in a solution containing 10 mM trisodium citrate, 0.005% Tween at 95°C for 20 min. Following washing in PBS, sections were incubated in a blocking solution (5% NDS, 2.5% BSA, and 0.3% PBS‐T) for 2 h at room temperature. Subsequently, sections were incubated overnight at 4°C with primary antibodies: rabbit anti‐Olig2 (1:250, Merck, ab9610), goat anti‐PDGFR⍺ (1:500, R&D Systems, af1062), and mouse anti‐CNPase (1:2000, Atlas Antibodies, AMAb91072) in an antibody buffer (1% NDS, 1% BSA, and 0.05% PBS‐T). After washing in 0.1% PBS‐T, sections were incubated for 2 h with Alexa 488, 647, and anti‐mouse Alexa Fluor 555 secondary antibodies (1:200, ThermoFisher Scientific). Following washes, sections were counterstained with DAPI (1:10 000) and mounted using DAKO fluorescent mounting medium (Agilent Dako, #S3023). OLIG2/ PDGFR⍺ images were acquired using a confocal laser scanning microscope with a 20x dry objective lens (Olympus FV1000, Olympus UPlanSApo 20x NA:0.75), while CNPase images were captured using an upright light microscope (Olympus BX53, Olympus PlanApo 2x NA:0.08) with a 2x objective.

Images were processed and analyzed using the ImageJ Fiji Software (NIH) [71]. For Olig2/PDGFR⍺ analysis, cell counts within the mPFC were quantified. CNPase analysis of the PFC and surrounding regions involved converting images to 8‐bit. Image quality was enhanced using the unsharp mask function (radius = 3, mask = 0.6). Default thresholding was applied, followed by conversion to binary format. The area of CNPase staining was normalized to the measured PFC area. Image analysis included 6 mice per experimental group, with 4‐6 sections analyzed per mouse.

5.7. Electron Microscopy of Mouse Tissue

Mice were transcardially perfused with 2.5% PFA and 2% glutaraldehyde, followed by microdissection of the mPFC grey matter (prelimbic, infralimbic, and anterior cingulate cortex, (approx. Bregma 2.93–1.69 mm). Tissue was post‐fixed for 2 h at RT followed by 90 min in osmium tetroxide and subsequent dehydration in an ascending series of ethanol, followed by propylene oxide. Each sample was embedded in Araldite resin (Agar Scientific, Essex, UK). Sectioning and random imaging of the entire mPFC was performed by UCD Conway Imaging Core (Ireland). Briefly, samples were sectioned with a Leica UC6 Ultramicrotome. Semi‐thin sections (500 nm) were stained with toluidine blue and examined under a light microscope. Thin sections (80 nm) were taken onto 200 mesh copper grids and double contrasted with 2% uranyl acetate for 20 min and 3% lead citrate for 5 min. Samples were imaged with a Tecnai G2 12 BioTWIN transmission electron microscope, operated at 120 kV. Axon diameter, myelin thickness, and inner tongue thickness were determined from the measured area, assuming circularity, using ImageJ Fiji Software (NIH), as described previously [34]. Diameter = 2 × √(area/π), with analysis performed on a minimum of 50 axons per animal. Analysis did not differentiate between specific subregions of the mPFC and focused primarily on grey matter.

5.8. Zebrafish Husbandry and Transgenic Lines

Adult wild‐type zebrafish (Danio rerio) were housed and maintained under standard procedures approved by the European Directive 2010/63/EC and met the requirements of the S.I No 543 of 2012 and approved by the Animal Experimentation Ethics Committee of University College Cork. Experiments requiring transgenic zebrafish were performed under standard conditions in the Queen's Medical Research Institute BVS Aquatics facility at the University of Edinburgh. Transgenic experiments were performed in compliance with the UK Home Office, according to its regulations under the project license PP5258250. Adult zebrafish were subject to a 14:10 h light:dark cycle. Embryos produced were kept at 28.5°C in embryo medium (EM) [72]. All experiments used zebrafish larvae at 5 dpf on a wildtype (AB/TL), nacre ^−/− [73] or casper (roy ^−/− nacre ^−/−) [74] background. During this stage, sexual differentiation of zebrafish had not yet occurred.

Throughout the text and figures “Tg” refers to a stable, germline‐inserted transgenic line. The following lines were used in this study: Tg(mbp:eGFP‐CAAX) [35], Tg(hspGFF62A:Gal4); Tg(UAS:mem‐Scarlet) [36, 75, 76], Tg(mpeg1:eGFP) [77], and Tg(mbp:mScarlet) and Tg(mbp:memScarlet).

5.9. Germ‐Free Derivation of Zebrafish Larvae

Germ‐free (GF) larvae were generated as previously described [72]. Briefly, after collection, embryos were resuspended in filter‐sterilized (0.2 µm) antibiotic embryo medium (ABEM) containing 100 µg/mL ampicillin, 5 µg/mL kanamycin, 250 ng/mL amphotericin B, and 50 µg/mL gentamicin. Embryos were maintained at a temperature of 28.5°C, with the removal of unfertilized embryos occurring every 2 h. At 50% epiboly, embryos designated for germ‐free derivation were transferred to a tissue culture hood and gently washed three times in sterile EM. Subsequently, embryos were immersed in 0.1% PVP‐I for 2 min, and then immediately washed three times in sterile EM. After washing, the embryos were immersed in 0.003% bleach solution for 20 min before being washed an additional three times in sterile EM. Subsequently, embryos were transferred into sterile tissue culture plates at a density of ∼1 larva/mL. The plates were sealed securely and placed in a 28.5°C incubator on a 14/10 light/dark cycle until 5 days post‐fertilization (dpf), when all tests were performed. Conventionally raised control embryos were also collected, placed in sterile EM, and subjected to the same environmental conditions as GF larvae. From 1–5 dpf hatching, survival, and malformations were recorded for all experimental groups at 10 am.

5.10. Sterility Testing

At 0 dpf and 5 dpf, media sterility was tested by inoculating 10 µl of media from each plate on tryptic soy agar (TSA) (Sigma, #22091), Nutrient Broth (Sigma, #70122), Brain Heart Infusion Broth (Sigma, #53286), or Sabouraud Dextrose Broth (Sigma, #S3306) plates. Subsequently, plates were incubated under aerobic conditions at 26°C for at least 5 days. GF plates that showed visual evidence of growth were excluded from the study.

5.11. Quantitative RT‐PCR

Zebrafish larvae were euthanized via hypothermic shock and immediately snap‐frozen on dry ice. All samples were collected between 9 and 10 am and then stored at −80°C until further processing. Total RNA was extracted from ∼10 whole larvae using the RNeasy Plus Universal Mini Kit (Qiagen) according to the manufacturer's instructions. RNA concentration and quality were determined using a Nanodrop ND‐1000 (Thermo Scientific) and reverse transcribed using the high‐capacity cDNA reverse transcription kit (Thermo Fisher Scientific) in a G‐storm thermocycler. Real‐time PCR amplification was performed on 12.5 ng of cDNA (Power Up SYBR Green Master Mix, Applied Biosystems) to evaluate gene expression levels. Gene expression levels were analyzed in a Lightcycler 480 II (Roche). Expression levels were calculated as the average of three replicates relative to a stably expressed housekeeper, rpl13a. Relative mRNA expression was calculated using the ΔΔCt method [78]. Primer sequences were obtained from the literature and are provided in Table S1.

5.12. Live Imaging of Zebrafish Larvae

Larvae were anesthetized using 600 µM tricaine in EM and embedded in 1.3–1.5% low‐melting‐point agarose on a glass coverslip. This coverslip was then suspended over a microscope slide using high vacuum silicone grease, forming a well containing EM and 600 µM tricaine. Z‐stacks (with an optimal z‐step) were acquired using a Zeiss LSM880 microscope equipped with Airyscan FAST in super‐resolution mode, employing a 63× objective lens (Zeiss C‐Apochromat 63×/1.15 W Korr UV‐VIS‐IR). Data processing utilized default Airyscan processing settings (Zen Black 2.3, Zeiss). Mauthner axon and myelin images were captured from a lateral view of the spinal cord, centered around somite 15.

For automated imaging of the entire larvae, vertebrate automated screening technology (VAST) was performed as described previously [79]. Briefly, larvae were transferred into individual wells of a 96‐well plate containing 600 µM tricaine in EM. Fish were prepared and positioned for imaging using a Large Particle (LP) Sampler and VAST BioImager system (Union Biometrica Inc) equipped with a 600‐µm capillary tube. Larvae were automatically loaded into the capillary, positioned, and imaged using an AxioCam 506m CCD Camera, a CSU‐X1 spinning disk confocal scanner, a 527/54 + 645/60 nm double bandpass emission filter, 1.6x C‐Mount adapter, a PIFOC P‐725.4CD piezo objective scanner, C‐Plan‐Apochromat 10x (NA = 0.5) objective, and an Axio Examiner D1. Z‐stacks covering the depth of the capillary were acquired with a 2‐µm z‐interval, 3 × 3 binning, and 100 ms exposure. Images were acquired using brightfield and the appropriate fluorescent channel. For lateral spinal cord images, the anterior aspect is to the left and the dorsal aspect at the top. Conversely, images of the head regions show a dorsal view, with the anterior and dorsal aspects positioned at the top.

5.13. VAST Image Processing and Analysis

Images acquired via VAST were stitched and processed using ImageJ Fiji software along with customized macro scripts [71, 79]. Semi‐automated counts of oligodendrocytes and microglia, as well as total area measurements, were conducted on the maximum‐intensity projection images. For cell count quantification and total cell volume determination, cell bodies were manually segmented using the brush tool in Arivis Vision4D. This process involved sequential z‐slice masking of clear, in‐focus cell body fluorescence, with overlapping masks subsequently reconstructed into 3D objects for volume measurement. Subsequent analysis included the manual quantification of overlapping cell bodies to investigate the total number of glial contacts.

The detection and quantification of morphological features post‐image acquisition were conducted utilizing FishInspector software [80]. This software employs a deep learning approach, utilizing the MATLAB Neural Network Toolbox and MATLAB Image Processing Toolbox (Mathworks, Natick, MS), allowing for automatic annotation of morphological features. Manual corrections were applied to automated annotations when necessary to ensure accurate detection of visible structures. To mitigate user bias, all images were assessed in a blinded manner.

5.14. Quantification of Axon Morphology

The total number of synaptic boutons per confocal image was manually quantified for each Mauthner axon segment (135 µm) using ImageJ/FIJI software. Subsequently, synaptic boutons were cut from the image to enable further analysis of axon diameter. Axon diameter was assessed using customized scripts [36], developed in the ImageJ Macro Language, and executed within the open‐source image analysis software Fiji [71]. Measurement focused on the axon closest to the imaging objective, following established procedures [36]. The “Axon Trace Tool” was employed to trace the axon's approximate center point along its length, while the “Axon Calibre Tool” measured the average axon diameter based on this trace. This involved calculating the axon's trajectory for each x‐coordinate, plotting a perpendicular line, extracting fluorescence profiles, and determining axon diameter at each location.

5.15. Quantification of Myelin

Quantification of whole‐body myelin from VAST imaging was conducted using maximum intensity projections. In summary, the images were first converted to 8‐bit format. A default thresholding process was applied, followed by noise reduction using the process > noise > despeckle function. Next, the images were converted to binary format, and the total area of myelin was measured. To account for variations in fish size, the myelin area was normalized to the length of the fish's spinal cord tract.

G‐Ratios and myelin sheath thickness from confocal imaging were measured as previously described [36]. Briefly, Z‐stacks were converted to maximum intensity projections. An axon area spanning 50 µm was traced using the polygon selection tool and measured. The average diameter was calculated by dividing the area by the axon length. Similarly, the average diameter of myelin + axon was determined using the same method. Subsequently, g‐ratios were calculated by dividing the axon diameter by the myelin + axon diameter. Myelin thickness was determined by subtracting the axon diameter from the myelin + axon diameter and dividing by two. The total fluorescent intensity of the myelin sheath thickness was calculated by cutting the axon tracing from the image. Following this, the total fluorescence intensity specifically associated with the myelin sheath thickness was calculated using the formula: myelin integrated density—(background area × background mean fluorescence). Subsequently, these values were normalized to the controls for that imaging session.

5.16. Statistical Analysis

All statistical analysis was conducted in SPSS (IBM, SPSS Statistics 28), with the exception of metabolomics and transcriptomics analyses detailed above. A maximum of one outlier per group (if any) was excluded using the Grubbs outlier test. Normality was assessed using the Shapiro–Wilk test, and equality of variances was examined using Levene's test. Nonparametric data were analyzed using the Kruskal–Wallis test followed by post‐hoc Dunn's tests. Parametric data were analyzed using two‐way between‐subjects ANOVAs (with Tukey's HSD post‐hoc comparisons to interrogate significant interactions) or independent‐samples t‐test to interrogate significant interactions between two groups. Data are shown as mean ± SEM. Statistical significance was set at p < 0.05.

Author Contributions

C.M.K.L., J.N., G.C., and J.F.C. conceived and designed the study. C.M.K.L. wrote the manuscript. D.A.L., I.H., J.N., G. C., and J.F.C. critically revised the manuscript. All animal work was performed by C.M.K.L., E.G.K., M.K.C., G.S.S.T., and S.I. Multi‐omics analysis and bioinformatics were performed by T.F.S.B., C.M.K.L., K.T., and interpreted by C.M.K.L. Immunohistochemistry was performed and analyzed by C.M.K.L., S.I., and E.G.K. Electron microscopy samples were embedded and analyzed by C.M.K.L. Gene expression analysis was performed and analyzed by C.M.K.L. and S.I. Zebrafish‐live Imaging was performed by C.M.K.L., D.S., and D.A. and analyzed by D.S. and C.M.K.L. Figures were constructed by C.M.K.L. All authors approved the final version of the manuscript.

Funding

APC Microbiome Ireland is a research centre funded by Science Foundation Ireland (SFI/12/RC/2273_P2). J.F.C. is funded by the Saks Kavanaugh Foundation and the Swiss National Science Foundation (project CRSII5_186346/NMS2068). J.N. was supported by a Government of Ireland Irish Research Council Postdoctoral Fellowship (GOIPD/2019/714) and is currently funded by the Research Ireland Pathway Programme (22/PATH‐S/10876). DAL was supported by a Wellcome Trust Senior Research Fellowship 214244/Z/18/Z and an MS Society, UK Edinburgh Research Centre award (133). DA was supported by a Wellcome Trust, Four‐Year PhD Program in Tissue Repair 108906/Z/15/Z.

Conflicts of Interest

J.F.C. has received research funding from Dupont/IFF, Reckitt, and Nutricia and has been an invited speaker at conferences organised by Yakult, Bromotech & Nestle. G.C. has received honoraria from Janssen, Probi, and Apsen as an invited speaker; research funding from Pharmavite and Fonterra; and serves as a paid consultant for Yakult, Zentiva, and Heel Pharmaceuticals. I.H. has received honoraria from Lundbeck as an invited speaker. J.N. has received research funding from Reckitt and BioGaia, and has received honorarium from Nestle as an invited speaker. All other authors report no potential conflicts of interest.

Supporting information

Supporting File 1: advs73599‐sup‐0001‐SuppMat.docx.

Supporting Table 1: advs73599‐sup‐0002‐Supplementary Table 1.xlsx.

Supporting Table 2: advs73599‐sup‐0003‐Supplementary Table 2.xlsx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.