Infants < 90 days of age with late-onset sepsis display disturbances of the microbiome-immunity interplay

Abstract

Objective: We hypothesized that previously healthy infants < 90 days of age with late-onset sepsis (LOS) have disturbances of the gut microbiome with yet undefined specific immunological patterns.

Methods: We performed a prospective single-center convenience sample study between January 2019 and July 2021 in a case-control design. Routine diagnostics included conventional cultures (blood, cerebrospinal fluid, urine), PCRs and inflammatory markers in infants aged < 90 days with clinical LOS. We additionally analyzed blood lymphocyte subsets including CD4 + CD25 + forkhead box protein (FoxP3)⁺ Tregs and performed 16 S rRNA sequencing of stool samples, both compared to age-matched healthy controls. Results were adjusted for potential confounders that may influence microbial composition.

Results: 51 infants with fever and clinical LOS were enrolled. Bacterial sepsis was diagnosed in n = 24 (47.1%) and viral infection in n = 13 (25.5%) infants, whereas in 14 (27.3%) infants the cause of fever remained undetermined. When compared to healthy controls, the gut microbiome of LOS infants at disease onset was characterized by a shift in community composition, specifically, decreased abundance of B. longum and an increase of Bacteroidia spp. Intriguingly, the abundance of B. longum negatively correlated with the frequency of blood CD4-positive cells in healthy controls but not in infants with LOS. At one year of age, we observed microbiome differences in infants with history of LOS when compared to healthy controls, such as an increased gut microbial diversity.

Conclusion: Our data suggest potential signatures of the microbiome-immunity interplay in infants with LOS, which should be investigated further as possible targets for prevention.

Keywords: Bifidobacteria; Infants < 90 days; Microbiome; Neonatal immunity; Sepsis.

Fig. 1

Gut microbial composition at onset of sepsis differs from healthy, same aged children. On class level, Actinobacteria are diminished in sepsis infants, while Bacteroidia are increased instead (A). Genus level assignment shows that the decrease in relative abundance of Actinobacteria is largely due to less of genus Bifidobacterium, while the genus Bacteroides is increased in septic children. On species level, the differences between sepsis and control group are depicted by B. longum and B. adolescentis B alongside unclassified Bifidobacteria (increased in controls) and unclassified Bacteroides sequences as well as B. fragilis (in sepsis infants) (C). While measures of alpha-diversity are similar between the groups (D-E), constrained correspondence analysis depicts the global changes of both groups to be significant (F)

Fig. 2

LEfSe-analysis supports the notion of the genera Bifidobacterium and Bacteroides defining hallmarks of control vs. sepsis children. Significant associations of the genus Bifidobacterium with the control group and Bacteroides with the LOS group and their respective upper taxonomic ranks is shown by linear discriminant effect size analysis (LEfSe). Additional taxa are assigned in this analysis to both of the groups

Fig. 3

Sustained deviation in microbial composition following treatment for LOS. At age of 12 months infants’ microbiota has developed to a largely Clostridia-dominated community (A) with typical genera/species expected in the gut (B-C). Assessment of Shannon´s diversity index (D) and species richness (E) as well as constrained correspondence analysis (F) show significant differences between sepsis and controls

Fig. 4

Counts of CD4 + CD25- immune cells at the age of < 90 days correlate with relative abundance of Bifidobacterium genus and specifically B. longum in multiple linear regression analysis. Genus Bifidobacterium is significantly negatively correlated with the abundance of CD4 + CD25- cells (A). However, when stratifying the analysis by LOS status the negative correlation remains significant only for the control group (B). This deviation between the groups is even more pronounced for the species level assignment of B. longum, where the negative correlation is present for the control group but completely missing for the sepsis group