Contribution of the Oral and Gastrointestinal Microbiomes to Bloodstream Infections in Leukemia Patients

Abstract

Bloodstream infections (BSIs) pose a significant mortality risk for acute myeloid leukemia (AML) patients. It has been previously reported that intestinal domination (>30% relative abundance [RA] attributed to a single taxon) with the infecting taxa often precedes BSI in stem cell transplant patients. Using 16S rRNA amplicon sequencing, we analyzed oral and stool samples from 63 AML patients with BSIs to determine the correlation between the infectious agent and microbiome composition. Whole-genome sequencing and antimicrobial susceptibilities were performed on all BSI isolates. Species-level detection of the infectious agent and presence of antibiotic resistance determinants in the stool (bla_CTX-M-15, bla_CTX-M-14, cfrA, and vanA) were confirmed via digital droplet PCR (ddPCR). Individuals with Escherichia coli (stool P < 0.001), Pseudomonas aeruginosa (oral P = 0.004, stool P < 0.001), and viridans group streptococci (VGS) (oral P = 0.001) bacteremia had a significantly higher relative abundance of those respective genera than other BSI patients, which appeared to be site specific. Although 78% of patients showed presence of the infectious genera in the stool and/or saliva, only 7 exhibited microbiome domination. ddPCR confirmed species specificity of the 16S data and detected the antibiotic resistance determinants found in the BSI isolates within concurrent stools. Although gastrointestinal (GI) domination by an infecting organism was not present at the time of most BSIs in AML, the pathogens, along with AMR elements, were detectable in the majority of patients. Thus, rapid genetic assessment of oral and stool samples for the presence of potential pathogens and AMR determinants might inform personalized therapeutic approaches in immunocompromised patients with suspected infection. IMPORTANCE A major cause of mortality in hematologic malignancy patients is BSI. Previous studies have demonstrated that bacterial translocation from the GI microbiome is a major source of BSIs and is often preceded by increased levels of the infectious taxa in the GI (>30% abundance by 16S rRNA sequencing). In this study, we sought to better understand how domination and abundance levels of the oral and gut microbiome relate to bacteremia occurrence in acute myeloid leukemia patients. We conclude that analyses of both oral and stool samples can help identify BSI and antimicrobial resistance determinants, thus potentially improving the timing and tailoring of antibiotic treatment strategies for high-risk patients.

Keywords: bacteremia; colonization; intestinal domination; leukemia; oral microbiome.

FIG 1

Linear discriminant analysis for effect size (LEfSe) analysis shows enrichment of taxa that cause respective infections. LEfSe was performed on stool and oral samples for each infectious agent designated to determine differences in the microbiome by site and infectious species. Organisms with a green bar are enriched in patients with the designated infection, while organisms with a red bar are enriched in patients who do not have that designated infection. Only the graphs in which the site was enriched with the etiological agent are shown.

FIG 2

Only patients with Escherichia coli infections showed differences in relative abundance between their stool and oral samples. Wilcoxon testing was performed on infectious agents of interest to determine if patients showed differing levels of abundance of their etiological agent between their stool and oral microbiomes. This was done by grouping patients by each infectious agent and plotting the relative abundance of the infectious agent in stool samples versus the relative abundance of the infectious agent in oral samples. The only set of patients that showed a statistically significant difference between samples were patients with E. coli infections.

FIG 3

Graphical depiction of percent positive droplets by ddPCR compared to relative abundance via 16S rRNA sequencing. Digital droplet PCR was performed on DNA extracted from stool samples of patients who were infected by Escherichia coli (A), Enterococcus spp. (B), Streptococcus mitis (C), Streptococcus oralis (D), Staphylococcus epidermidis (E), Klebsiella pneumoniae (F), Pseudomonas aeruginosa (G), and Staphylococcus aureus (H). ddPCR-positive percentages were determined by dividing the number of positive droplets by the total number of droplets. Those values were plotted against the relative abundance (RA) values gathered from 16S rRNA gene sequencing. The Pearson test was used to determine the correlation coefficient (r) and P values for all graphs. Simple linear regression was used to plot a line of best fit (red line) on each graph, where each dot represents an individual patient. A ROUT analysis was used to remove outliers from each data set prior to preforming the analysis.

FIG 4

Presence of acquired resistance determinants among bloodstream isolates. A binary heatmap is shown based on the presence or absence of specific acquired resistance genes present among all infectious isolates. The x axis is organized alphabetically by species, and the y axis is organized according to the antibiotic resistance conferred by each gene. The heatmap is colored from pale red, indicating a gene is not present, to bright red, indicating that a gene is present in that isolate. Species are abbreviated as follows: Escherichia coli, EC; Enterococcus faecium, Efm; Enterococcus faecalis, Efa; Pseudomonas aeruginosa, PA; Klebsiella pneumoniae, KP; Staphylococcus aureus, SA; Staphylococcus epidermidis, SE; and viridans group streptococci, VGS.

FIG 5

Presence of acquired resistance genes verified by ddPCR. This two-axis graphical depiction shows the 16S rRNA abundance of the genera of interest labeled on the right axis and percentage of positive droplets for the tested antibiotic resistance genes on the left axis. The x axis depicts each sample as the patient number followed by the pathogen being tested.