Rare and Intermediate Taxa Shape the Gut Bacterial Structure in Neonates and Preterm Infants with Necrotizing Enterocolitis

Abstract

Necrotizing enterocolitis (NEC) is a common neonatal gastrointestinal disease with high morbidity and mortality, especially in premature infants. In a prospective case-control study, we aimed to investigate the dynamic changes in the gut microbiota of preterm infants with NEC. Infants diagnosed with NEC and preterm neonates were enrolled in this study, while normal neonates were selected as the control group. The collected samples were divided into three groups: the control group (NC), the neonatal NEC group (NEC), and the premature delivery NEC group (pdNEC). Along with basic clinical data, fecal samples from the infants (n = 39) were collected at the time of the first diagnosis of NEC for 16S rRNA gene sequencing. Analysis of the gut microbiota revealed no significant difference in α-diversity between infants with NEC and controls, regardless of preterm birth. The significant difference in β-diversity was primarily driven by the rare and intermediate subgroups. The rare gut subgroup found in premature infants with NEC played a crucial role in the deterministic process and specialized functionality of the microbiota, ultimately forming a sparse association network structure. Finally, multiple biomarkers of Enterococcus from the Firmicutes phylum were identified, providing a theoretical basis for diagnosing NEC in premature infants.

Keywords: Necrotizing enterocolitis; association network; biomarkers; gut microbiota; premature infants.

Fig. 1. Comparison of intestinal microbiota structure at the amplicon sequence variant (ASV) level in NC, NEC, and pdNEC.

The ASVs were classified into rare (≤ 0.01%), rich (≥ 0.1%), and intermediate (between 0.01% and 0.1%) subgroups based on their relative abundance, and the corresponding alpha diversity indices were calculated. ANOVA and Tukey’s multiple comparison tests were used for comparisons and different lowercase letters represented significant differences.

Fig. 2. Comparisons to the β-diversity distance and dissimilarity index of intestinal microbiota in different groups (NC, NEC, and pdNEC).

Significant changes in beta diversity were calculated using PERMANOVA on the Bray-Curtis distance matrix. Different lowercase letter represented significant differences of dissimilarity index between groups, based on Kruskal-Wallis test combined Wilcoxon tests.

Fig. 3. Comparison of tNST, βNTI, and niche breadth of intestinal microbiota in NC, NEC, and pdNEC groups.

(A) tNST values for bacterial communities for NC, NEC, and pdNEC groups. Light yellow and light blue background color indicated tNST thresholds of 0.5. (B) βNTI values for bacterial communities for different groups. Light-yellow and light-blue background color indicated betaNTI thresholds of +2 and −2. (C) Proportion of community assembly processes in different groups, and the proportion of 0.00% is deleted. Significant changes in tNST, βNTI, and niche breadth were calculated using one-way ANOVA followed by Tukey HSD tests, and significant differences were indicated by horizontal bars with corresponding p-values.

Fig. 4. Relative abundance of the bacterial phyla and the distribution of species of different subgroups in each clinical group.

(A) Relative abundance of ASVs at phylum level between groups. (B) Distribution of phyla in different subgroups. (C) Clustering heatmap of ASVs at the phylum level.

Fig. 5. Relative abundance and association networks of the intestinal genera in different NEC groups.

Spearman rank correlation analysis was performed to calculate the correlation coefficient and displayed ASVs with a correlation >0.6 and a p-value of <0.05. The red side indicated a positive correlation, and the blue side represented a negative correlation. The size of the nodes reflected the average relative abundance of ASVs, and nodes belonging to a cluster had the same color.

Fig. 6. Identification of bacterial biomarkers of NECs by random forest models.

(A) A 10-fold cross-validation on a random forest model was performed to detect unique amplicon sequence variant (ASV)-based markers. (B) The mean decreased accuracy and importance of the top 30 biomarkers. (C) Relative abundance distribution of biomarkers at phyla level, genus level, and different subgroups.