Metagenomic Sequencing with Strain-Level Resolution Implicates Uropathogenic E. coli in Necrotizing Enterocolitis and Mortality in Preterm Infants

Abstract

Necrotizing enterocolitis (NEC) afflicts approximately 10% of extremely preterm infants with high fatality. Inappropriate bacterial colonization with Enterobacteriaceae is implicated, but no specific pathogen has been identified. We identify uropathogenic E. coli (UPEC) colonization as a significant risk factor for the development of NEC and subsequent mortality. We describe a large-scale deep shotgun metagenomic sequence analysis of the early intestinal microbiome of 144 preterm and 22 term infants. Using a pan-genomic approach to functionally subtype the E. coli, we identify genes associated with NEC and mortality that indicate colonization by UPEC. Metagenomic multilocus sequence typing analysis further defined NEC-associated strains as sequence types often associated with urinary tract infections, including ST69, ST73, ST95, ST127, ST131, and ST144. Although other factors associated with prematurity may also contribute, this report suggests a link between UPEC and NEC and indicates that further attention to these sequence types as potential causal agents is needed.

Figure 1. Species Composition and Relative Abundance in the Infant Gut

(A) Days 3–9 postpartum. (B) Days 10–16 postpartum. (C) Days 17–22 postparum. Species are ranked top to bottom by average relative abundance across samples. Species presented exhibit a minimum average relative abundance of 0.2% or achieve at least 5% relative abundance in at least one sample (complete abundance data in Table S1). Samples were hierarchically clustered with Bray-Curtis dissimilarity. Color indicates relative abundance on a logarithmic scale.

Figure 2. Relative Abundance and Prevalence of the Most Frequently Occurring Species in the Early Preterm Infant Microbiome

Fourteen of the most prevalent species observed in the preterm infant microbiome are presented. E.coli is an outlier with a median relative abundance exceeding 0.80 in each of the three collection windows presented. Relative abundance calculations include only samples where the infant carried the organism with at least 1% relative abundance. Prevalence is indicated as red circle, mean relative abundance as a blue diamond, and median relative abundance as box and whisker plot.

Figure 3. Functional and Metagenomic MLST Sub-typing of E. coli Strains Present in Metagenomic Samples Indicates UPEC Lineages Are Associated with NEC Risk

E. coli accessory gene-content information (Table S2) was used to hierarchically cluster infant strains and indicated three distinct functional clades. Approximately Unbiased (AU) p value, a multiscale bootstrap resampling method (Shimodaira, 2004), indicates the strength of support for clusters. Clades 1 (red) and 2 (green) contain NEC cases. Metagenomically identified MLST IDs are indicated, and the corresponding phylogroup inferred from literature and available resources (http://mlst.warwick.ac.uk/mlst/). Consecutively numbered and shaded infant IDs indicate fraternal twin pairs. Boxed infant IDs indicate two strains of E.coli identified in temporally separated samples from a single infant (patient ID 21461).