Nuanced contribution of gut microbiome in the early brain development of mice

ABSTRACT

The complex symbiotic relationship between the mammalian body and gut microbiome plays a critical role in the health outcomes of offspring later in life. The gut microbiome modulates virtually all physiological functions through direct or indirect interactions to maintain physiological homeostasis. Previous studies indicate a link between maternal/early-life gut microbiome, brain development, and behavioral outcomes relating to social cognition. Here we present direct evidence of the role of the gut microbiome in brain development. Through magnetic resonance imaging (MRI), we investigated the impact of the gut microbiome on brain organization and structure using germ-free (GF) mice and conventionalized mice, with the gut microbiome reintroduced after weaning. We found broad changes in brain volume in GF mice that persist despite the reintroduction of gut microbes at weaning. These data suggest a direct link between the maternal gut or early-postnatal microbe and their impact on brain developmental programming.

KEYWORDS:

• Germ-free mice
• gut microbiome
• early-life alteration
• brain development
• MRI

Introduction

The mammalian body is inhabited by a myriad of microorganisms originating from the vertical transfer of vaginal-perianal microbes during delivery.^1,2 The complex symbiotic relationship rapidly established after birth between the offspring and the microorganisms hosted,³ particularly within the gastrointestinal system, is critical for the establishment and maturation of immune functions,⁴ metabolism,^5,6 and the modulation of physiological homeostasis.^7,8 Through direct anatomical connection, indirect endocrine means, or the mobilization of the immune system, microbial metabolites from the GI tract can modulate a broad range of physiological functions.^9,10

It is increasingly apparent that the gut microbiota is a direct, modifiable factor of neurodevelopment and neurological outcomes in humans. Early exposure to antibiotics, which alters microbiome diversity,^11,12 is associated with an increased risk of cognitive and neurodevelopmental disorders such as attention-deficit hyperactivity disorder and autism spectrum disorder.^13,14 Fecal transplantation performed on germ-free (GF) dams using human maternal microbes from preterm mothers suggests poor growth outcomes.¹⁵ Interestingly, there is a correlation between the time and duration of antibiotic exposure and the severity of cognitive deficits later in life,¹⁶ and the levels of gut microbe by-products 3,4-dihydroxyphenylacetic acid and sodium butyrate are positively associated with mental health^17,18 and adult neurogenesis.¹⁹

Experiments using the offspring of germ-free GF and antibiotic-treated dams further revealed that planar indoles and amine oxides produced by gut microbes regulate the expression of genes involved neurogenesis, neuronal differentiation, and axonogenesis.^20,21 Alternate gut bacteria-derived metabolites and by-products, such as the short-chain fatty acids and peptidoglycan is a direct modulator of the blood-brain barrier structure, neurogenesis, and neuronal maturation pathways.^22–26 In relation, Sven Pettersson and others observed that altered gut microbiome colonization and composition significantly impact cognitive behavior in mice,^27–31 with a preferential effect on social behavior. Although pieces of information point to the involvement of gut microbiota in the modulation of brain development and function,³² there is a lack of direct evidence of gut microbiome-dependent change in brain development.

Preclinical magnetic resonance imaging (MRI) methods provide a unique opportunity for a complete brain structure examination under the defined absence of gut microbiome. In this study, we examined the impact of early life absence of gut microbiome on brain development through structural MRI between control, specific-pathogen-free mice carrying normal microbiome population and the GF mice (Figure 4). Alterations in isocortex and olfactory cortex volumes associated with higher-order cognitive control ensued from the complete depletion of microorganisms sustained through a restriction in environmental exposure. Additionally, the conventionalization of young adult GF mice failed to normalize the brain changes observed. These observations provide a premise for future functional and neurocircuitry studies to explore the structural determinants of cognitive deficits associated with early-life disturbance in the gut microbiome.

Results

GF mice exhibit distinct brain volume changes compared with age-matched SPF mice

Associations between early-life microbiome dysbiosis and later-stage cognitive deficits suggest neurodevelopmental deviations under gnotobiotic conditions.^16,33,34 To identify brain regions modulated by commensal bacteria, volumetric analysis of 6 to 8 weeks old SPF and GF mouse brains was achieved with MRI. Corrected statistical maps depicting significant group differences (p corrected < 0.05) are shown with an anatomical underlay. We observed a widespread alteration in brain volume in the GF versus the SPF mouse brain (Figure 1a). Notably, the absence of microbiome results in a significant bilateral increase in the olfactory cortex and isocortical brain volumes (AP + 1.98 to −3.28, orange-red sections) and a decrease in the sensory-related superior colliculus (SCs; AP −4.96), thalamus (TH, AP −2.8), and hypothalamus (AP −0.94; blue-green sections) in the GF compared with SPF mouse (Figure 1a). Specifically, there is a significant increase in the volumes of the somatomotor (MO, includes MOp and MOs, 0.044 ± 0.006, p =.016), somatosensory (SSp, includes SSp-bfd, SSp-ul, SSp-ll, SSp-m, SSp-tr and SSp-n, 0.073 ± 0.014, p =.016; SSs 0.073 ± 0.010, p =.008), anterior cingulate cortex (ACA, including ACAd and ACAv, p =.016), prelimbic area (PL 0.105 ± 0.026, p =.032), ectorhinal area (ECT 0.068 ± 0.009, p =.008), frontal pole of cerebral cortex (FRP 0.074 ± 0.010, p =.008), temporal association areas (TEa 0.063 ± 0.017, p =.016) and the perihinal area (PERI 0.072 ± 0.014, p =.016) of the isocortex, and the anterior olfactory nucleus (AON 0.094 ± 0.017, p =.032) and taenia tecta (TT, includes TTv and TTd, 0.101 ± 0.015, p =.008) of the olfactory cortex in GF mice (Figure 1b and Supplementary Figure S1).

Figure 1. Widespread alterations in the subcortical brain volume of germ free (GF) mice.

(a) Coronal view of brain regions exhibiting reduced (blue-green) or increased (yellow-red) brain volume in GF compared to age-matched SPF subjects (control). The AP coordinates (in mm) of the exact coronal sections location in the mouse brain are provided on top of each of the presented statistical maps. MO somatomotor cortex, SSp primary somatosensory cortex, OLF olfactory cortex, CP caudoputamen, LPO lateral preoptic area, AUD auditory area, TEa temporal association area, LHA lateral hypothalamic area, TH thalamus, HPF hippocampal formation, PAG periaqueductal gray, SCs sensory-related superior colliculus, P pons, CB cerebellum. (b) Relative brain volume changes of age-matched SPF compared with GF, normalized against the average brain volume from the SPF subjects. Orange, isocortex; Blue, olfactory cortex. (c) Relative subcortical region, hippocampal formation, cerebral nuclei, and the midbrain, hindbrain, and brain stem volume changes of age-matched SPF compared with GF normalized against the average brain volume from the SPF subjects. The precise names of brain regions can be found in the list of abbreviations in supplementary materials. Brain regions are arranged based on their anatomical location. SPF n = 5, GF n = 5. Data presented as mean ± SEM. *, p < .05; **, p < .01.

Conventionalization post-weaning does not reverse major brain volume changes observed in GF mice

To further understand the brain volume differences observed and examine the impact of the absence of gut microbiome on brain volume changes, we reviewed the effect of reintroducing gut microbiome post-weaning (conventionalization) on brain volumes of GF mice (CONV) compared against age-matched SPF mice (Figure 2a). Despite having a fecal bacterial population resembling the donor SPF mice post conventionalization,³⁵ brain areas altered in CONV mice are similar to that of GF mice compared against SPF mice (Figure 2b; MO 0.054 ± 0.008, p =.016; SSp 0.069 ± 0.006, p =.008; SSs 0.052 ± 0.015, p =.008; ACA 0.090 ± 0.011, p =.008; PL 0.145 ± 0.032, p =.008; ECT 0.064 ± 0.006; p =.008; FRP 0.116 ± 0.025, p =.032; TEa 0.044 ± 0.011, p =.032; PERI 0.079 ± 0.004, p =.008; AON 0.125 ± 0.011, p =.008; TT 0.107 ± 0.017, p =.008), and that volumes of these regions are comparable between CONV and GF mice (Figure 3b, Supplementary Figure S1; MO 0.009 ± 0.008, p =.421; SSp −0.004 ± 0.006, p >.999; SSs −0.020 ± 0.014, p =.421; ACA 0.011 ± 0.010, p >.999; PL 0 ± 0.010, p =.548; ECT 0 ± 0.008, p =.691; FRP 0 ± 0.009; p =.151; TEa −0.018 ± 0.010, p =.421; PERI 0 ± 0.013, p =.841; AON 0.028 ± 0.011, p =.222; TT 0.005 ± 0.016, p >.999).

Figure 2. Conventionalisation of GF mice post-weaning does not reverse majority of the brain volume changes in GF mice.

(a) Coronal view of brain regions exhibiting reduced (blue-green) or increased (yellow-red) brain volume in post-weaning conventionalized GF mice (CONV) compared to age-matched SPF subjects (control). The AP coordinates (in mm) of the exact coronal sections location in the mouse brain are provided on top of each of the presented statistical maps. MO somatomotor cortex, SSp primary somatosensory cortex, OLF olfactory cortex, CP caudoputamen, LPO lateral preoptic area, AUD auditory area, TEa temporal association area, LHA lateral hypothalamic area, TH thalamus, HPF hippocampal formation, PAG periaqueductal gray, SCs sensory-related superior colliculus, P pons, CB cerebellum. (b) Relative brain volume changes of age-matched SPF compared with CONV, normalized against the average brain volume from the SPF subjects. Orange, isocortex; Blue, olfactory cortex. (c) Relative subcortical region, hippocampal formation, cerebral nuclei, and the midbrain, hindbrain, and brain stem volume changes of age-matched SPF compared with CONV normalized against the average brain volume from the SPF subjects. The precise names of brain regions can be found in the list of abbreviations in supplementary materials. Brain regions are arranged based on their anatomical location. SPF n = 5, CONV n = 5. Data presented as mean ± SEM. *, p < .05; **, p < .01.

Figure 3. Brain regions sensitive toward gut microbe-dependent biochemical changes are affected by the reintroduction of gut microbiome.

(a) Coronal view of brain regions exhibiting reduced (blue-green) or increased (yellow-red) brain volume in GF compared to age-matched post-weaning conventionalized GF mice (CONV) subjects. The AP coordinates (in mm) of the exact coronal sections location in the mouse brain are provided on top of each of the presented statistical maps. MO somatomotor cortex, SSp primary somatosensory cortex, OLF olfactory cortex, CP caudoputamen, LPO lateral preoptic area, AUD auditory area, TEa temporal association area, LHA lateral hypothalamic area, TH thalamus, HPF hippocampal formation, PAG periaqueductal gray, SCs sensory-related superior colliculus, P pons, CB cerebellum. (b) Relative brain volume changes of age-matched GF compared with CONV, normalized against the average brain volume from the SPF subjects. Orange, isocortex; Blue, olfactory cortex. (c) Relative subcortical region, hippocampal formation, cerebral nuclei, and the midbrain, hindbrain, and brain stem volume changes of age-matched GF compared with CONV normalized against the average brain volume from the SPF subjects. The precise names of brain regions can be found in the list of abbreviations in supplementary materials. Brain regions are arranged based on their anatomical location. SPF n = 5, CONV n = 5. Data presented as mean ± SEM. *, p < .05; **, p < .01.

The reintroduction of gut microbiome influenced brain regions sensitive towards gut microbe-dependent biochemical perturbations

Gut microbiome disturbances can induce permanent alterations in key neurological processes and brain development or transient, reversible fluctuations in brain volume through neurological effects by changes in the systemic level of bacteria-dependent metabolites,³⁶ neurotransmitters,^37,38 cytokine,³⁹ and endocrinal signals.⁴⁰ .The reintroduction of microorganisms postnatally is sufficient to normalize the subcortical claustrum (CLA SPF vs GF 0.060 ± 0.008, p =.008; SPF vs CONV 0.029 ± 0.011, p =.151), the isocortex infralimbic area (ILA SPF vs GF 0.096 ± 0.023, p = .016; SPF vs CONV 0.071 ± 0.04, p = .222), auditory area (AUD SPF vs GF 0.076 ± 0.016, p =.032; SPF vs CONV 0.045 ± 0.014, p =.095), hippocampal formation CA1 region (CA1slm SPF vs GF −0.043 ± 0.012, p =.008; SPF vs CONV 0.010 ± 0.015, p =.056), the pons (P SPF vs GF −0.084 ± 0.014, p =.016; SPF vs CONV −0.009 ± 0.020, p =.548), and the brain stem superior colliculus (SCs SPF vs GF −0.093 ± 0.012, p =.008; SPF vs CONV −0.065 ± 0.018, p =.056) (Figure 2b, c). Surprisingly, we observed a significant alteration in the thalamic nuclei (AP-0.34 to −2.8) and pons (AP −4.96) of the CONV mouse versus GF mouse brain (Figure 3a) and alterations in previous unaffected dorsal agranular insular area (Figures 1b, 2 b and 3b; Aid), posterior parietal association area (PTLp), and lateral entorhinal area (Figures 1c, 2 c and 3c; ENTl), endopiriform nucleus (EP consisting of EPd and EPv), dentate gyrus (DG), and the ventral pallidum (PALv). The fiber tract and ventricular system are unaffected by the altered gut microbiome composition (Supplementary Figure S2).

Discussion

Epidemiological studies have reported a possible correlation between alterations in the composition of the gut microbiome and increased risk of neurodevelopmental and neuropsychiatric disorders.^13,41 Preclinical animal studies further showed that exposure to gut microbial products can alter key neurodevelopmental processes.^20,21,42,43 However, the direct contribution of the gut microbiota and neurocognitive outcomes remains inconclusive due to a lack of understanding of the underlying mechanisms.^44–46 Here, we present evidence for an association between structural alterations in the olfactory cortex and isocortex of the GF mice (Figure 4) and the neonatal gut microbe’s regulated developmental structural changes.

Figure 4. Ablation of gut microbiota persistently affects brain development in young adult mice.

An increase in the volume of olfactory and somatosensory cortices observed in the absence of microorganisms in the GF mice is irreversible with the reintroduction of the normal gut microbiome from the SPF mice in early postnatal life. As such, the resulting brain volume changes in CONV mice are similar to that of GF mice compared to SPF mice. Alterations in development and the neuronal circuitry of these brain regions are often associated with the development of neuropsychiatric conditions such as attention deficit hyperactivity disorder (ADHD), anxiety, and depression.

The increase of interest in the effects of intestinal microbiota on human health has led to extensive metagenome sequencing efforts to understand the variation and diversity between organisms and health conditions^47–50 and the generation of preclinical models to elucidate the gut microbiome mechanism of action.^51–53 Further research has clarified the contribution of the gut in the brain function of adults through the direct activation of the vagus nerve,⁵⁴ the transfer of bacteria metabolites into the nervous system⁵⁵, or the regulation of immune activation.⁵⁶ The use of GF mice provides the opportunity to decipher the role of the gut microbiota and regulation of brain development and function, relatively less explored.

Brain volume change is a powerful correlate of neurodevelopmental and neurocognitive outcomes of individuals,^57,58 and MRI is a noninvasive method that allows for the complete examination of brain structure changes in a single experimental run. We observed broad structural volume alteration specific to the somatosensory in GF mice, critical for general and higher-order sensory processing and integration^59–63 (Figures 1b, Figure 3b, and Supplementary Figure S1). This is consistent with previous reports detailing defects in sensory processing in neuropsychiatric conditions associated with a disturbed gut microbiome.^13,64,65 In addition, changes in the ILA and ACA associates with contextual fear conditioning⁶⁶ and the modulation of fear and anxiety^67–69 correspond to altered anxiety behavior commonly observed in GF mice.^27,28,70 Previous research on similar-aged GF mice suggests that a combination of alterations in microglia colonization, cell survival,⁷¹ and white matter development^72,73 contributes to gut microbe-dependent brain morphological development and maturation.

Furthermore, there is a directed increase in the volumes of the AON and TT of the olfactory cortex (Figures 1b, 3b, and Supplementary Figure S1), which function as areas for olfactory information processing and integration.⁷⁴ Although relatively less explored, olfactory compromises are often observed in autism spectrum disorder patients^75,76 and correlates with the development of depression⁷⁷ and schizophrenia.⁷⁸ The olfactory nerve undergoes a distinct developmental process,⁷⁹ thought to involve the function of yet unknown odorant receptors,⁸⁰ which may be directly responsive to gut microbiome-produced metabolites such as short-chain fatty acids and volatile amines that functions as odorants.^81,82 A deep examination of the interactions between gut microbiome activities and the olfactory system is required to understand the relevance of microbiome disturbance in the modulation of olfactory development and behavior.

Gut microbe metabolites were shown to regulate the expression of genes and processes involved in neural stem cell differentiation, neurogenesis, and development.^20,21,83 Changes in cerebral metabolism,⁸⁴ expression of plasticity,²⁷ growth factor¹⁵ or myelin-related genes,⁸⁵ and activity-related transcriptional pathways⁸⁶ were related to changes in the amount of brain structure in the prefrontal cortex observed in GF mice. Also, gut-derived neuroactive substances such as tryptophan and serotonin modulate the inflammatory status and function of the microglia in the central nervous system (CNS).^87,88 Due to the time-sensitivity of the production of morphogens,⁸⁹ and the vulnerability of the neurological system to changes during key neurodevelopmental stages, neurological deviations can lead to persistent structural, functional, or cognitive disability. Specifically, the neurological impact of microbiome disturbance early in life correlates with later cognitive outcomes in alternate GF mouse models⁷¹ and human infants,⁹⁰ similar to our observations that the recolonization of gut microbiome does not reverse detected brain changes. In rodents, key neurodevelopmental milestones are concluded by a month of age (postnatal),⁹¹ corresponding to the timepoint of gut microbiome reintroduction to the GF mice in this study. Notably, reintroducing the gut microbiome through conventionalization at 3 to 4 weeks of age cannot reverse the majority of observed brain alterations (Figure 2b, c), in line with a previous report suggesting that behavioral changes in NMRI mice can only be reversed with postnatal colonization of microbes from females transplanted with microbes before pregnancy.²⁷ The conventionalization of GF mice is nonetheless sufficient to evoke changes in the colliculus superior (SCs) and hippocampus (Supplementary Figure S2), reflecting their sensitivity toward metabolic alterations.^92–94

We recognize that brain volume changes are dynamic and can be affected by short-term deviations in neurophysiological conditions.^95–97 Furthermore, it is not possible to distinguish the role of maternal gestational microbiota and early postnatal microbiome in brain volume changes observed, and the brain volume alterations do not reveal the underlying structural vulnerability responsible for the cognitive correlates observed in individuals with gut microbiome disturbance. It is unclear how the difference in weight and body fat content of GF mice from an altered gut microbiome-dependent metabolic profile may contribute to changes in brain development.⁹⁸ The involvement of sex-dependent hormones in shaping gut microbiome composition alongside neurodevelopment and brain structure maturation requires further evaluation.^72,99,100 Treatment of neurodevelopmental disorders in later stages of life is a possible concept for the normalization of neuronal network function without impacting the underlying defect in consideration of adult neurological plasticity. Further work is required to investigate the mechanisms involved in the observations made in the current report.

Materials and methods

Animal welfare

All animal procedures were carried out per institutional guidelines by Nanyang Technological University and approved by Regional Animal Research Ethical Board, Institutional Animal Care and Use Committee, Singapore. Male NMRI (Naval Medical Research Institute) mice, 6–8 weeks old, are used for MRI analysis in this study. The Germ-free mouse used was provided by Professor Sven Pettersson. The mouse colony is initially obtained from the Core Facility for Germ-free Research, Karolinska Institutet, and maintained in the gnotobiotic facility of the Nanyang Technological University before their use in experiments. All mice were kept under a 12-h light/dark cycle and provided with water and food (autoclaved R36 Lactamin) ad libitum. GF mice were placed and raised in special sterile plastic isolators. The isolator sterility was analyzed weekly through the bacterial and fungal culture of fecal homogenates of GF mice. Part of the GF mice population was conventionalized (CONV) after weaning (between 3 to 4 weeks of age) by gavage of 100 μl of specific pathogen-free (SPF) mouse fecal homogenate (dissolving two fecal pellets from SPF mice in 1 ml of phosphate-buffered saline) and cohoused with SPF mice for 21 days before being euthanized. 5 mice from each group are used in this study. No subjects were excluded from the analysis.

MRI acquisition and data processing

Mouse brains are imaged postmortem in situ.¹⁰¹ Data were acquired with a Bruker 11/16 Biospec spectrometer (Bruker BioSpin MRI, Ettlingen, Germany) operating at 11.75 T, equipped with a BGA-S gradient system, a linear volume resonator coil for transmission, and a two-by-two phased-array cryocoil surface receiver coil. Images were acquired using Paravision 6 software. Localizer images were acquired to ensure the correct positioning of mice to the coil and the magnet isocentre. 3D volumetric scanning occurred for 8h30 min, with scans acquired using spin echo RARE sequence¹⁰² with a field of view saturation slice positioned on the inferior portion of the head and acquisition parameters: repetition time = 2500 ms, echo time = 5.1 ms, RARE factor = 10, matrix size 250 × 225 × 250, and field of view = 20 × 18 × 20 mm³, average = 2, achieving a resolution of 0.08 mm³.

Anatomical scans were processed through voxel-based morphometry using ANTS (Advanced Normalization Tools 2.4.0, picsl.upenn.edu/software/ants). Briefly, images were first corrected for intensity field inhomogeneity using the N4 algorithm. Second, a study template was estimated by realigning individual volumes to each other over 3 iterative steps of non-linear transformation. Third, the individual volumes underwent SyN diffeomorphic image registration to the study template. Fourth, the Jacobian determinants were estimated and voxel-wise comparisons between groups were carried out with a two-sample t-test using FSL 6.0.5 (fsl_glm) and corrected for multiple hypothesis testing using the easy threshold function.

Statistical analysis

Jacobian determinants were extracted using the DSURQE mouse brain atlas.¹⁰³ All data are expressed as mean ± standard error mean (SEM) and ported to GraphPad Prism 7.04 (GraphPad Software) for analysis. Datasets were tested for normal distribution with the Shapiro-Wilk test. Further analysis was conducted with the Mann-Whitney U test between SPF, GF, and CONV datasets. P values <.05 were considered statistically significant.

Authors’ contributions

Conceptualisation, X.Y.Y., S.P., and S.J.; data acquisition and analysis, J.G., X.Y.Y., and H.U.L., B.H.; writing – original draft preparation. X.Y.Y., W.R.C., and S.J.; writing – review and editing, X.Y.Y., S.P., J.G., W.H., and S.J.; supervision, S.J.; funding acquisition S.P., W.H., and S.J.; critical revision of manuscript, S.J., S.P., W.H., and J.G. All authors have read and agreed to the published version of the manuscript.

Availability of data and materials

The raw datasets generated for this study are available in http://doi:10.18112/openneuro.ds004254.v1.0.0. The programming code used for data preprocessing and analysis can be found at https://gitlab.socsci.ru.nl/preclinical-neuroimaging/germfree. Further inquiries can be directed to the corresponding authors.

Ethics approval

All animal procedures were carried out per institutional guidelines by Nanyang Technological University and approved by Regional Animal Research Ethical Board, Institutional Animal Care and Use Committee, Singapore.

Disclosure statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2023.2283911

Additional information

Funding

This work is supported by the Joint Council Office grant (BMSI/15-800003-SBIC-00E) from Agency for Science, Technology and Research (A*STAR), Singapore. S.P. is supported by the ASEAN Microbiome Nutrition Centre (AMNC), Singapore, UK Dementia group, and National Neuroscience Institute (NNI), SingHealth, Singapore. W.H. is supported by A*STAR Intramural funding, A*STAR Strategic Research Program (the Brain-Body Initiative, iGrants call ID #21718), the A*STAR Central Research Fund, and the A*STAR JCO-VIP award. S.J. is supported by the industry academic cooperation foundation, CHA University grant (CHA-202300230001).