Levofloxacin prophylaxis and parenteral nutrition have a detrimental effect on intestinal microbial networks in pediatric patients undergoing HSCT

Abstract

The gut microbiome (GM) has shown to influence hematopoietic stem cell transplantation (HSCT) outcome. Evidence on levofloxacin (LVX) prophylaxis usefulness before HSCT in pediatric patients is controversial and its impact on GM is poorly characterized. Post-HSCT parenteral nutrition (PN) is oftentimes the first-line nutritional support in the neutropenic phase, despite the emerging benefits of enteral nutrition (EN). In this exploratory work, we used a global-to-local networking approach to obtain a high-resolution longitudinal characterization of the GM in 30 pediatric HSCT patients receiving PN combined with LVX prophylaxis or PN alone or EN alone. By evaluating the network topology, we found that PN, especially preceded by LVX prophylaxis, resulted in a detrimental effect over the GM, with low modularity, poor cohesion, a shift in keystone species and the emergence of modules comprising several pathobionts, such as Klebsiella spp., [Ruminococcus] gnavus, Flavonifractor plautii and Enterococcus faecium. Our pilot findings on LVX prophylaxis and PN-related disruption of GM networks should be considered in patient management, to possibly facilitate prompt recovery/maintenance of a healthy and well-wired GM. However, the impact of LVX prophylaxis and nutritional support on short- to long-term post-HSCT clinical outcomes has yet to be elucidated.

Fig. 1. A different gut microbiome network structure is found in pediatric HSCT patients in association with EN, PN, and LVX prophylaxis.

Gut microbiome networks are shown for each treatment group (EN in yellow, (a); PN LVX (–) in pink, (b); PN LVX (+) in cyan, (c); n = 30 for each group), including all timepoints (i.e., before (T0) and after HSCT (T1 and T2)). Node sizes are proportional to the mean relative abundance of the species, and the network layout, derived from spin-glass clustering into modules, is maintained over the three plots. Green edges represent positive correlations, while red edges are indicative of negative correlations. Only edges with Spearman’s correlation FDR-corrected p < 0.05 and (rho) ρ < −0.3 or ρ > 0.3 are shown. Edge line width is proportional to Spearman’s ρ. To simplify the reading of the networks, only the nodes with an abundance >0.2% were labeled; bold labels correspond to species with an abundance higher than 1%. To construct the graph, only species with relative abundance >0.3% in at least 5% of the samples were considered.

Fig. 2. Levofloxacin prophylaxis results in a sustained reduction in alpha diversity.

Boxplots representing the distribution of alpha diversity estimated with the Shannon index for each patient group (EN, PN LVX (–), PN LVX (+)) before (T0) and at 2 time points after HSCT (T1 and T2). Significant differences between groups are shown (Wilcoxon rank sum test with FDR correction, *** p < 0.001, ** p < 0.01, * p < 0.05). n = 10 biologically independent samples for each timepoint. Whiskers represent range between the first quartile (Q1) and the third quartile (Q3). Data points outside the boundary of the whiskers are plotted as outliers.

Fig. 3. Functional potential of the components of the gut microbiome networks in pediatric HSCT patients.

Heatmap representing the Kendall correlation between species-level relative abundances and pathway CPMs (copies per million, see Methods) at all three timepoints. Species considered are the same used for network construction in Fig. 1. Only pathways that were differentially represented among groups (p < 0.05, Kruskal–Wallis test, grouping samples by feeding route, LVX administration, and timepoint) were considered. Pathway names (following the MetaCyc annotation) were shortened as follows: SPWY super pathway, bsy biosynthesis. n = 90 samples from 30 individuals.

Fig. 4. The gut microbiome profiles of pediatric HSCT patients receiving PN and LVX prophylaxis are associated with altered carbohydrate and xenobiotic metabolic potential.

On top of each figure, bar plots show the KEGG pathway classification of identified Kos divided by amino acid (a), carbohydrate (b), lipid (c), and xenobiotic (d) metabolism, represented as the mean relative contribution of each pathway to the total CPMs (copies per million) assigned to a given metabolism. Below each bar, the average number of normalized reads assigned to each metabolism is represented (CPM ± standard error of the mean). Data are shown for each patient group (PN combined with LVX prophylaxis—PN LVX (+) vs PN alone—PN LVX (–) vs EN alone—EN) before (T0) and at 2-time points after HSCT (T1 and T2). Significant differences among groups are shown (Kruskal–Wallis test with FDR correction). n = 10 biologically independent samples for each time point.

Fig. 5. Network parameter analysis showed lower gut microbiome ability to withstand stress after LVX prophylaxis and PN in pediatric patients undergoing HSCT.

Each bar refers to a single network that was computed for the gut microbiome of each patient group (PN combined with LVX prophylaxis—PN LVX (+) vs PN alone—PN LVX (–) vs EN alone—EN) before (T0) and at 2 timepoints after HSCT (T1 and T2), leading to the derivation of a single parameter value for each network. The following network parameters were considered: modularity (i.e., the measure of connections between and within modules), total cohesion (TC, i.e., quantification of connectivity in terms of positive and negative interactions), and the ratio of negative to positive cohesion (N:P cohesion ratio). On top of each series of values a bell-curve represents the one-sample t test performed and the corresponding one-sided p value. See also Methods. n = 10 biologically independent samples for each timepoint.