Specific microbiome changes in a mouse model of parenteral nutrition associated liver injury and intestinal inflammation

Abstract

Background: Parenteral nutrition (PN) has been a life-saving treatment in infants intolerant of enteral feedings. However, PN is associated with liver injury (PN Associated Liver Injury: PNALI) in a significant number of PN-dependent infants. We have previously reported a novel PNALI mouse model in which PN infusion combined with intestinal injury results in liver injury. In this model, lipopolysaccharide activation of toll-like receptor 4 signaling, soy oil-derived plant sterols, and pro-inflammatory activation of Kupffer cells (KCs) played key roles. The objective of this study was to explore changes in the intestinal microbiome associated with PNALI.

Methodology and principal findings: Microbiome analysis in the PNALI mouse identified specific alterations within colonic microbiota associated with PNALI and further association of these communities with the lipid composition of the PN solution. Intestinal inflammation or soy oil-based PN infusion alone (in the absence of enteral feeds) caused shifts within the gut microbiota. However, the combination resulted in accumulation of a specific taxon, Erysipelotrichaceae (23.8% vs. 1.7% in saline infused controls), in PNALI mice. Moreover, PNALI was markedly attenuated by enteral antibiotic treatment, which also was associated with significant reduction of Erysipelotrichaceae (0.6%) and a Gram-negative constituent, the S24-7 lineage of Bacteroidetes (53.5% in PNALI vs. 0.8%). Importantly, removal of soy oil based-lipid emulsion from the PN solution resulted in significant reduction of Erysipelotrichaceae as well as attenuation of PNALI. Finally, addition of soy-derived plant sterol (stigmasterol) to fish oil-based PN restored Erysipelotrichaceae abundance and PNALI.

Conclusions: Soy oil-derived plant sterols and the associated specific bacterial groups in the colonic microbiota are associated with PNALI. Products from these bacteria may directly trigger activation of KCs and promote PNALI. Furthermore, the results indicate that lipid modification of PN solutions may alter specific intestinal bacterial species associated with PNALI, and thus suggest strategies for management of PNALI.

Figure 1. Identification of specific taxa associated with PNALI from comparison of PN/DSS and NS/DSS mice.

A. Manhattan Plot showing the -log₁₀ p-values from the two-part statistical test comparing 46 taxa identified between PNALI mice (PN/DSS) and controls (NS/DSS). The two-part statistic compares the proportion of samples in each group that contained a specific taxon, and the median relative abundance between groups. Taxa that were significantly different are numbered (p<0.05 and <0.01 are marked with a solid line). The numbered peaks correspond to Clostridiales (1), Lachnospiraceae (2), Anaerotruncus (3), Ruminococcus (4) and Erysipelotrichaceae (5). B. Change in relative abundance and proportion of animals positive for each taxon between groups is plotted. This analysis was limited to taxa present in ≥50% of individuals in either group. Taxa labels are given for groups present in >1% relative abundance. Names are shaded to correspond with statistical significant taxa (black) versus those that did not achieve statistical significance (grey). The purpose of this plot was to assess the biological relevance of each taxon identified irrespective of statistical significance. The quadrants that are labeled show where both relative abundance and proportion are higher for the group indicated. C. Manhattan plot of two-part analysis for comparison of PN/DSS versus chow. The four peaks correspond to Bacteroidetes (1), Ruminococcus (2), Erysipelotrichaceae (3) and Turicibacter (4). D. Change in relative abundance and proportion between PN/DSS and chow mice as described in B.

Figure 2. Box-whisker plots are shown comparing relative abundance across groups of mice for prominent taxa identified in two-part analyses.

Statistically significant differences are marked for comparison of other groups to the PN/DSS mice (bolded; *p<0.05, **p<0.01, ***p<0.001). A. Erysipelotrichaceae B. S24-7 C. Lachnospiraceae. Box represents median and 25 and 75% ile (interquartile range, IQR) and whiskers represent 1.5x IQR. Erysipelotrichaceae increases in relative abundance in groups that do not have enteral feedings (PN) and those that exhibit abnormal liver function (PN/DSS). The S24-7 group is the primary Gram-negative group observed, and is eradicated by antibiotic treatment. This taxon is the most likely candidate for TLR4 activation via LPS.

Figure 3. Effect of antibiotic (Abx) treatment on the mouse fecal microbiome in PN/DSS mice with PNALI.

A. Manhattan plot of two part analysis of PN/DSS versus PN/DSS/Abx mice. Peaks correspond to Propionibacterium (1), S24-7 (2), Staphylococcus (3), Anaerotruncus (4) and Anaeroplasma (5). The effect of antibiotics was to depress several of the more prominent groups allowing detection of a large number of rarer taxa. B. Change in relative abundance and proportion between PN/DSS and PN/DSS/Abx mice as described in Figure 1B. Several taxa exhibited large changes in relative abundance without achieving significance (Lachnospiraceae, Ruminococcaceae, Erysipelotrichaceae, Clostridiales and Enterococcus).

Figure 4. Principal component analysis of all groups of mice.

A. Principal component (PC) 1 versus PC2. There is a strong signal separating groups exposed to plant material (chow or soy-based PN) from the Omegaven and no lipid control groups along PC1. This includes the two groups of animals given the stigmasterol spiked Omegaven PN formula. The taxa most influential in PC1 and PC2 are shown as vectors of the PC loadings. B. PC3 versus PC2 is shown to better distinguish between the groups exposed to plant based nutrition. Important taxa are shown as PC loading vectors as in panel A, Esch. = Escherichia and Erysip. = Erysipelotrichaceae.

Figure 5. Proposed pathways involved in the pathogenesis of PNALI.

Administration of PN and the associated absence of enteral feedings are the primary drivers of intestinal dysbiosis, in addition to small bowel bacterial overgrowth caused by impaired intestinal motility. In the presence of increased intestinal permeability secondary to injury and inflammation, microbe associated molecular patterns (MAMPs) derived from specific bacterial populations that thrive in the dysbiotic environment are absorbed into the portal circulation and promote TLR signaling and activation of hepatic Kupffer cells (KCs.) Soy lipid-derived plant sterols excreted into the gut via the enterohepatic circulation system in turn promote dysbiosis in addition to their effect on activation of KCs and suppression of hepatocyte bile secretion. These two latter processes culminate in PNALI and cholestasis.