Aberrant gut-microbiota-immune-brain axis development in premature neonates with brain damage

Abstract

Premature infants are at substantial risk for suffering from perinatal white matter injury. Though the gut microbiota has been implicated in early-life development, a detailed understanding of the gut-microbiota-immune-brain axis in premature neonates is lacking. Here, we profiled the gut microbiota, immunological, and neurophysiological development of 60 extremely premature infants, which received standard hospital care including antibiotics and probiotics. We found that maturation of electrocortical activity is suppressed in infants with severe brain damage. This is accompanied by elevated γδ T cell levels and increased T cell secretion of vascular endothelial growth factor and reduced secretion of neuroprotectants. Notably, Klebsiella overgrowth in the gut is highly predictive for brain damage and is associated with a pro-inflammatory immunological tone. These results suggest that aberrant development of the gut-microbiota-immune-brain axis may drive or exacerbate brain injury in extremely premature neonates and represents a promising target for novel intervention strategies.

Keywords: Klebsiella; T cell ontogeny; early-life development; extremely premature infants; gut-immune-brain axis; microbiome; neurophysiology; perinatal white matter injury.

Graphical abstract

Figure 1

Neurophysiological development of extremely premature infants with and without severe brain injuries Blue color for age-adequate cranial magnetic resonance imaging (cMRI) results or mild brain injury (BI). Red color for severe BI. (A) Schematic illustration of study design. Solid arrows indicate sampling time points with blood drawings on day 3, day 7, day 28 post-delivery, 32 weeks gestational age, and at term-equivalent age. Dashed lines indicate time points for stool sampling and neurophysiological assessment (day 3, day 7, and day 14, followed by biweekly sampling until discharge). (B) Representative cMRI images at term-equivalent age for IVH > 2 (left) and PVL (right). (C) Comparison of Kidokoro scores between infants with and without severe BI. (D) Comparison of aEEG background patterns (1) continuous voltage (CV), (2) discontinuous voltage (DV), (3) burst suppression (BS). (E) Comparison of the variability of cranial oxygen supply (VAR-cSO₂) (fraction of deviation beyond 10% from the mean of the total NIRS measurement) between infants with and without BI. (F) Receiver operating characteristic (ROC) curve visualizing the predictive potential of cranial oxygen fluctuation (VAR-cSO₂) 4 weeks post-delivery, and BS signatures 2 weeks post-delivery for brain damage. (G) Summary of area under the curve (AUC) scores of neurophysiological parameters at given time points. Color gradient ranges from dark gray (low AUC score) to dark yellow (high AUC score). Box plots show group median and interquartile range. Smoothed lines result from locally estimated scatterplot smoothing (LOESS) and indicate trends of neurophysiological development. Asterisks represent p values: ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001.

Figure 2

T cell ontogenesis in extremely premature infants with and without severe brain injury (A) Principal coordinates analysis (PCoA)—biplot of sequestered cytokine and chemokine composition. Silhouette scoring identified 3 main cyto-Clusters (as indicated by different symbols). The shade of the symbols (gray to black) is determined by days post-delivery—the older the infant the darker the symbol. The size of the symbols is determined by the Kidokoro Score as assessed at term-equivalent age via cMRI—the higher the score the larger the symbol. Significantly correlated (p < 0.05) cytokines/chemokines are plotted as arrows. In addition, box plots next to PCoA show the range of combined values for burst suppression and variance of cranial oxygenation (BS/100 × VAR-cSO₂), as well as the range of Kidokoro Scores in the respective cyto-clusters (cyto-cluster 1, CC1; cyto-cluster 2, CC2; cyto-cluster 3, CC3). (B) Blood cytokine/chemokine concentrations in infants with (red) and without (blue) severe BI (3 days post-delivery, d3; 7 days post-delivery, d7; 28 days post-delivery, d28; 32 weeks of gestational age, w32; term-equivalent age = term). (C) Representative images illustrating the gating strategy for FACS analysis. Differently colored gates mark gating for respective cell populations. Magenta, untargeted; blue, T helper cells; orange, cytotoxic T cells; green, T regulatory cells; pink, γδ T cells. For box plots in (D) and (F–J), darker shade represents data from infants with, and lighter shade from infants without, severe BI. (D) Proportion of CD3⁺ cells during hospitalization. (E) Relative distribution of T regulatory (green), T helper (blue), γδ T (pink), and cytotoxic T cells (orange). (F–I) Box plots showing the expression of T cell receptors relative to the given parental population throughout hospitalization. (J) Box plots showing the expression of CCR6+/T regulatory cells with the addition of smoothed lines resulting from locally estimated scatterplot smoothing (LOESS) to indicate trends of development. (K) Spearman correlation analysis to reveal associations between T cell populations and selected sequestered cytokines/chemokines at respective time points (p < 0.001). Red color indicates positive correlation, blue color indicates negative correlation. Box plots show group median and interquartile range. Asterisks represent p values: ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001.

Figure 3

The microbiome of extremely premature infants and its diversification corresponding to brain injury (A) Total bacterial cell counts per gram feces over time. (B) Absolute abundances of the most abundant bacterial genera in infants with (red) and without (blue) severe BI. (C) T-distributed stochastic neighbor embedding (tSNE) ordination of all 16S rRNA gene amplicon libraries corrected by absolute abundances and 16S rRNA gene copy number. Hierarchical clustering (Ward’s method) identified 3 main clusters (green, orange, and purple) and 8 sub-clusters (symbols). (D) Area plot illustrating the percentage distribution of sub-clusters throughout time. (E) Absolute abundance of bacterial genera in respective main clusters. (F) Network illustrating the trajectories of microbiome succession in extremely premature infants. Circles correspond to tSNE sub-clusters. Circle size indicates the number of patients found in a sub-cluster for at least one sampling time. Arrows represent transitions between clusters, whereby the thickness and shade of the arrow indicates the probability of the transition. Recursive arrows represent the percentage of consecutive observations within the same cluster. Pie charts show the distribution of patients in respective clusters diagnosed as healthy (blue, lacking any diagnosis listed in Table 1), with severe BI (red), or other pathologies (other DMG, see Table 1; ochre). Sub-clusters are arranged according to the average patient age of samples in the cluster, ordered from youngest (left) to oldest (right). Box plots next to circles show the absolute abundance of Klebsiella (bright green) and Escherichia-Shigella (dark blue) in the associated sub-cluster. Box plots show group median and interquartile range. Smoothed lines result from locally estimated scatterplot smoothing (LOESS) and indicate trends of microbiome development. Asterisks represent p values: ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001.

Figure 4

SCFAs in stool of extremely premature infants (A) Enteral feeding over time for infants with (red) and without severe BI (blue). (B) Fecal SCFAs in nmol g⁻¹ feces throughout hospitalization of premature infants with (red) and without (blue) severe BI. (C) Weighted contour plot showing the association of SCFAs with percentage of enteral feeding and days post-delivery. Brighter color indicates higher SCFA concentrations. (D–F) (D) Heatmap of Pearson correlations of bacterial genus abundance with SCFA levels [nmol] before 28 days post-delivery and (E) at or after 28 days post-delivery (F) SCFA concentrations in infants with brain injuries (red) and infants without (blue) before 28 days post-delivery (< d28) and at or after (> d28). (G) Differences in overall fermentation measured as nM per Carbon (nM_C) between infants with (red) and without severe BI (blue) before 28 days post-delivery (< d28) and at after (> d28). Squares show group median, error bars show standard deviation. Smoothed lines result from LOESS and indicate trends of microbiome development. Asterisks represent p values: ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001.

Figure 5

The gut-microbiota-immune-brain axis of extremely premature infants and biomarkers of brain injury (A) Data integration analysis for biomarker discovery using Latent variable approaches (Diablo). Integration of microbiome (yellow), metabolite (pink), T cell (gray), cytokine and chemokine secretion (orange), and neurophysiological/clinical (blue) data (n = 84). Lines within the circle represent positive correlations in green, and negative correlations in black (canonical correlation cut-off: +/− 0.65). Lines outside of the circle indicate the predictive potential of corresponding variables for age-adequate or mild (blue) or severe (red) BI. (B) Correlation circle plot showing associations between variables as measured via canonical correlations. Neighboring of variables indicates correlations between variables.