Abstract
Introduction
The gastrointestinal tract is the predicted reservoir for most bloodstream infections (BSIs) after hematopoietic stem cell transplantation (HSCT). Whole-genome sequencing and comparative genomics have the potential to improve our understanding of the dynamics of gut colonization that precede BSI in HSCT recipients.
Methods
Within a prospective cohort study of children (<18 years) undergoing HSCT, 9 subjects met criteria for mucosal barrier injury BSI. We performed whole-genome sequencing of the blood culture isolate and weekly fecal samples preceding the BSI to compare the genetic similarity of BSI isolates to fecal strains. We evaluated temporal associations between antibiotic exposures and the abundances of BSI strains in the gut microbiota and correlated detection of antibiotic resistance genes with the phenotypic antibiotic resistance of these strains.
Results
Median age was 2.6 years, and 78% were male. BSIs were caused by Escherichia coli (n=5), Enterococcus faecium (n=2), Enterobacter cloacae (n=1), and Rothia mucilaginosa (n=1). In the 6 BSI episodes with evaluable comparative genomics, the fecal strains were identical to the blood culture isolate (>99.99% genetic similarity). Gut domination by these strains preceded only 4 of 7 E. coli or E. faecium BSIs by a median (range) of 17 (6–21) days. Increasing abundances of the resulting BSI strains in the gut microbiota were frequently associated with specific antibiotic exposures. E. cloacae and R. mucilaginosa were not highly abundant in fecal samples preceding BSIs caused by these species. The detection of antibiotic resistance genes for beta-lactam antibiotics and vancomycin predicted phenotypic resistance in BSI strains.
Conclusions
Bacterial strains causing mucosal barrier injury BSI in pediatric HSCT recipients were observed in the gut microbiota prior to BSI onset, and changes in the abundances of these strains within the gut preceded most BSI episodes. However, frequent sampling of the gut microbiota and sampling of other ecological niches is likely to be necessary to effectively predict BSI in HSCT recipients.
Keywords: bloodstream infection, hematopoietic stem cell transplantation, microbiome
INTRODUCTION
Hematopoietic stem cell transplantation (HSCT) is used to treat a growing number of malignant and non-malignant conditions in children and adults. The conditioning chemotherapy that patients receive in preparation for HSCT results in prolonged neutropenia and injury to mucosal surfaces, facilitating translocation of microbes across these barriers and predisposing to bloodstream infection (BSI). The most frequent causes of BSI in HSCT recipients are Enterobacteriaceae, enterococci, and viridans group streptococci [1, 2] These infections are associated with substantial morbidity and mortality, with case fatality rates of 8–67% depending on the causative organism [2, 3]. The gastrointestinal tract is the purported source of most of these infections, which suggests that serial monitoring of the gut microbiota could identify patients who are high risk for BSI after HSCT. However, prior studies were limited by infrequent fecal sampling or use of low-resolution methods (e.g. pulsed-field gel electrophoresis) to compare the genetic similarity of blood and fecal strains [4, 5, 6, 7]. Whole-genome sequencing recently emerged as a tool that provides unparalleled resolution for comparative microbial genomics and enables confident determination of clonal relationships [8].
In this study, we performed whole-genome sequencing of fecal samples and the blood culture isolate for BSI episodes occurring in a cohort of pediatric HSCT recipients. We compared the genetic similarity of fecal strains to the BSI strain and determined if shifts in the relative abundances of these strains in the gut microbiota preceded BSI. As secondary objectives, we evaluated associations between antibiotic exposures and the abundances of BSI strains in the gut microbiota and correlated detection of antibiotic resistance genes with the phenotypic antibiotic resistance of these strains.
MATERIALS AND METHODS
Study Population
Eligible subjects were less than 18 years of age and undergoing evaluation for their first allogeneic or autologous HSCT at Duke University between October 2015 and February 2018. One hundred subjects were enrolled and followed from prior to HSCT until 100 days after HSCT. For Pneumocystis jirovecii prophylaxis, subjects received trimethoprim-sulfamethoxazole starting at hospital admission and continuing until two days before HSCT, followed by inhaled or intravenous pentamidine starting 30 days after HSCT. Routine antibacterial prophylaxis was not used. The analyses presented herein were limited to subjects who developed a mucosal barrier injury laboratory-confirmed bloodstream infection, as defined by the U.S. National Healthcare Safety Network [22].
Data and specimen collection
Sociodemographic and clinical data were collected from a caregiver questionnaire and medical record review. Fecal samples were collected into vials containing RNAlater solution and placed immediately into a 4°C refrigerator. Study staff transported these samples daily (Monday through Friday) from clinical sites to the laboratory for storage in a −80°C freezer. Blood culture isolates were obtained from the Duke University Clinical Microbiology Laboratory. The laboratory routinely stores bacterial isolates, in glycerol at −80°C, from blood cultures for which there is monomicrobial growth. Bacterial isolates are tested for susceptibility to antimicrobials using the MicroScan WalkAway (Siemens Healthcare Diagnostics, Berkeley, CA), and breakpoints are provided by the Clinical and Laboratory Standards Institute [23]. For BSI episodes occurring in enrolled subjects, blood culture isolates were plated on blood agar and single colonies were selected for whole-genome sequencing.
DNA extraction and whole-genome sequencing
We selected weekly fecal samples from study enrollment to the BSI episode for DNA extraction and whole-genome sequencing. Total genomic DNA was extracted from fecal samples and blood culture isolates using a modified version of the Earth Microbiome Project protocol, as previously described [24, 25]. DNA sequencing libraries were constructed using the Nextera XT DNA Library Prep Kit (Illumina Inc., San Diego, CA) and sequenced on a NextSeq500 Sequencing System as 150-base paired-end reads. Sequences were trimmed and removed of host contamination using Trimmomatic and Bowtie [26, 27]. Host-decontaminated reads were then profiled for bacterial species abundances using MetaPhlAn2 [28, 29]. Abundances of antibiotic resistance genes were determined by mapping trimmed and host-decontaminated reads to the Comprehensive Antibiotic Resistance Database (CARD) using the ShortBRED bioinformatics pipeline [30, 31]. Normalized taxonomic and resistance gene abundances were then used for downstream analyses and visualization in R v3.4.2 (R Foundation for Statistical Computing, Vienna, Austria).
Strain variant identification
Genomic assemblies were constructed for blood culture isolates using SPAdes-3.11 assembler [32]. Multilocus sequence typing assignments were determined from the assembled contigs using mlst (https://github.com/tseemann/mlst). Based on the assignments, representative references were chosen for each species (Escherichia coli, accession HG941718.1; Enterococcus faecium, accession NC_017960.1). Paired-read data from both the blood culture isolate and fecal samples were aligned to these references using Bowtie2 and SAMtools, and single nucleotide variants were detected with Pilon [27, 33, 34]. Samples where aligned reads obtained passing coverage of at least 80% of the total reference genome were included in subsequent strain-level analyses. This resulted in comparison of E. coli strains over more than 3.8 Mbp and E. faecium strains over more than 2.4 Mbp, comprising 75% and 90% of the genomes for these species, respectively. Sequencing reads from fecal samples were used to draw a consensus conclusion about the genomic sequence of the dominant strain in the gut microbiota. Custom scripts were applied to determine conserved core nucleotide positions and produce multi-FASTA alignments. Core single nucleotide variant positions were determined from a multi-sequence alignment file of all core genome sequences using SNP-sites [35]. RAxML v8.2.10 and FigTree v1.4.3 were used to generate and format the phylogenetic trees [36, 37].
RESULTS
Study population
During the study period, 14 mucosal barrier injury BSIs occurred among the study population. For 3 infections, the blood culture isolate was not frozen by the microbiology laboratory or did not grow from the stored sample. No fecal samples were available preceding two additional BSI episodes. The final dataset was comprised of 9 mucosal barrier injury BSI episodes. Characteristics of these subjects and BSI episodes are shown in Table 1. Most subjects were male and had a malignant HSCT indication, while nearly half were recipients of an umbilical cord blood transplant. The most frequent BSI organisms identified were E. coli (n=5) and E. faecium (n=2).
Table 1.
Characteristics of the study population (n=9)
| N | % | |||
|---|---|---|---|---|
| Age in years, median (IQR) | 2.6 | (2.0–6.9) | ||
| Sex | ||||
| Female | 2 | 22% | ||
| Male | 7 | 78% | ||
| Race/ethnicity | ||||
| Black or African-American | 1 | 11% | ||
| Non-Hispanic white | 4 | 44% | ||
| Middle Eastern or Arab-American | 3 | 33% | ||
| Native American or Alaskan Native | 1 | 11% | ||
| HSCT indication | ||||
| Hematological malignancy | 2 | 22% | ||
| Genetic or metabolic disorder | 2 | 22% | ||
| Non-malignant hematological disorder | 2 | 22% | ||
| Solid tumor | 3 | 33% | ||
| HSCT type and source | ||||
| Allogeneic | ||||
| Matched, related bone marrow | 1 | 11% | ||
| Umbilical cord blood | 4 | 44% | ||
| Autologous | 4 | 44% | ||
| BSI organism | ||||
| Enterobacter cloacae | 1 | 11% | ||
| Enterococcus faecium | 2 | 22% | ||
| Escherichia coli | 5 | 56% | ||
| Rothia mucilaginosa | 1 | 11% | ||
IQR, interquartile range; HSCT, hematopoietic stem cell transplantation; BSI, bloodstream infection
Whole-genome sequencing
A total of 229,427,210 high-quality bacterial sequencing reads (mean of 4,498,573 sequences per sample) were obtained from the 42 fecal samples and 9 blood culture isolates included in these analyses. The median (interquartile range) number of reads per fecal sample was 2,715,388 (815,979–5,006,084). The median (interquartile range) number of reads per BSI isolate was 11,195,760 (6,667,320–14,876,920), with a median of 177-fold genome coverage depth. Sequences were assigned to 199 bacterial species, representing 74 genera from 6 phyla. The relative abundances of frequently occurring species in fecal samples are shown in Supplemental Figure 1.
Fecal strains are identical to subsequent BSI isolates
The genetic relatedness of fecal strains to blood culture isolates is shown in Table 2. For all BSI episodes for which comparative genomics was possible, the dominant strain within one or more fecal samples was identical (>99.99% genetic similarity) to the blood culture isolate. Moreover, within an individual patient, strains from fecal samples collected closer to the BSI episode had fewer nucleotide differences relative to the BSI isolate (Patients 2 and 7). Phylogenetic trees derived from core genomic sequences of E. coli and E. faecium are shown in Figure 1. Blood culture isolates and fecal strains clustered by patient. The lone exception was an E. coli strain identified in a fecal sample obtained 34 days prior to the BSI episode (−34d) from Patient 7, which suggests that the BSI strain replaced the prior E. coli strain in the gut of this patient between −34d and −25d.
Table 2.
Genomic concordance of fecal strains to blood culture isolates
| Patient | BSI Organism | Day of Fecal Sample | SNVs |
|---|---|---|---|
| 2 | Enterococcus faecium | −7 | 46 |
| −4 | 40 | ||
| 3 | E. faecium | −11 | 12 |
| 5 | Escherichia coli | −20 | 0 |
| −1 | 0 | ||
| 6 | E. coli | −4 | 5 |
| 7 | E. coli | −34 | 826 |
| −25 | 1 | ||
| −16 | 0 | ||
| −7 | 0 | ||
| −1 | 0 | ||
| 8 | E. coli | −6 | 2 |
BSI, bloodstream infection; SNVs, single-nucleotide variants For E. faecium isolates, a total of 2,436,174 core bases were compared to the reference genome. For E. coli isolates, 3,848,800 core bases were compared to the reference genome.
Figure 1. Phylogenetic trees for E. coli and E. faecium strains.
Core, single nucleotide variant phylogeny of E. coli and E. faecium is presented for blood culture isolates and corresponding fecal strains. Strains are colored according to patient and the sample day is presented relative to the BSI episode. Strains from a single patient are more closely phylogenetically related than are strains from different individuals.
Relative abundances of strains in the gut microbiota increase prior to BSI onset
Supplemental Figure 2 shows the Shannon Diversity Index of fecal samples relative to the timing of the BSI. In most patients, the BSI episode was preceded by a loss of gut microbial diversity, which often corresponded with a high or rising relative abundance of the BSI species (Figure 2). We observed several patterns of pathogen presence in the gut microbiota preceding BSI. First, gut domination by the BSI species, defined as a relative abundance of at least 30% in the gut microbiota, preceded 4 of 7 BSI episodes caused by E. coli or E. faecium for a median (range) of 17 (6–21) days [6]. Second, gut domination did not occur in Patients 4 and 6, but both BSI episodes were preceded closely by rapid increases in the relative abundance of the BSI species in the gut microbiota. The abundance of E. coli rose from 0.0% (−4d) to 29% (−1d) in Patient 4 and from 0.2% (−10d) to 11% (−2d) in Patient 6. Third, Patient 8 developed an E. coli BSI following a stable, moderate abundance (4–12%) of this species in the gut microbiota between −18d and −1d. Finally, the BSIs caused by E. cloacae and R. mucilaginosa were preceded by low abundances of these species in the gut microbiota. In Patient 1, the E. cloacae relative abundance was <2% in serial fecal samples obtained between −20d and −1d. R. mucilaginosa was not detected in fecal samples obtained from Patient 9 on −15d and −4d.
Figure 2. Relative abundances of bacterial species in the gut microbiota preceding BSI onset.
Bacterial reads were classified at the species level, and the relative abundance of the BSI species in fecal samples is shown preceding the BSI episode. Each darkened vertical bar corresponds to the relative abundance of the BSI species in a single fecal sample. Shading between each bar is provided for visualization of changes in relative abundance between fecal samples. The blood culture isolate is shown as a vertical bar and black circle on day 0. Specific antibiotic exposures are depicted as horizontal bars above each plot, with the position and length of the bar corresponding to the timing and duration of the antibiotic exposure. TZP, piperacillin-tazobactam; SXT, trimethoprim-sulfamethoxazole.
Detection of antibiotic resistance genes and the phenotypic resistance of BSI isolates
The results of antibiotic susceptibility testing of blood culture isolates are shown in Table 3. Both E. faecium isolates were resistant to vancomycin. The E. cloacae strain and 3 E. coli isolates (Patients 5, 7, and 8) were resistant to third-generation cephalosporins. The detection of antibiotic resistance genes in blood culture isolates is shown in Figure 3. The vanA gene cluster was detected from both vancomycin-resistant E. faecium (VRE) isolates. Genes encoding extended-spectrum beta lactamases (TEM-211, CTX-M-72, CTX-M-100, CTX-M-110, and OXA-1) were detected from the E. cloacae and E. coli isolates with resistance to third-generation cephalosporins. Resistance genes for aminoglycosides, fluoroquinolones, tetracyclines, and trimethoprim-sulfamethoxazole were detected frequently but did not strongly correlate with phenotypic resistance in the E. cloacae and E. coli isolates.
Table 3.
Antibiotic susceptibility testing of blood culture isolates
| Patient | BSI Organism | AMP | CRO | CEF | TZP | MEM | TOB | CIP | SXT | TCN | VAN |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Enterobacter cloacae | R | R | S | S | S | S | S | R | S | - |
| 2 | Enterococcus faecium | R | - | - | - | - | - | - | - | - | R |
| 3 | E. faecium | R | - | - | - | - | - | - | - | - | R |
| 4 | Escherichia coli | S | S | S | S | S | S | S | S | S | - |
| 5 | E. coli | R | R | R | S | S | R | R | R | S | - |
| 6 | E. coli | R | S | S | S | S | S | S | R | S | - |
| 7 | E. coli | R | R | R | R | S | R | R | S | S | - |
| 8 | E. coli | R | R | R | S | S | I | R | R | S | - |
| 9 | Rothia mucilaginosa | S | - | - | - | - | - | R | R | - | S |
BSI, bloodstream infection; AMP, ampicillin; CRO, ceftriaxone; CEF, cefepime; TZP, piperacillin-tazobactam; MEM, meropenem; TOB, tobramycin; CIP, ciprofloxacin; SXT, trimethoprim-sulfamethoxazole; TCN, tetracycline; VAN, vancomycin; S, susceptible; I, intermediate; R, resistant
Susceptibility of bacterial species to specific antibiotics was determined using breakpoints provided by the Clinical and Laboratory Standards Institute.
Figure 3. Detection of antibiotic resistance genes in blood culture isolates.
Each column represents a blood culture isolate from a single patient. The rows correspond to specific antibiotic resistance genes and are organized based on the antibiotic class to which the gene confers resistance. The color of each cell corresponds to the presence or absence of an antibiotic resistance gene within the blood culture isolate. FQ, fluoroquinolones; TCN, tetracyclines; SXT, trimethoprim-sulfamethoxazole; VAN, vancomycin
Rising abundances of BSI strains were frequently associated with antibiotic exposures
In the majority of BSI episodes, receipt of specific antibiotics was associated with changes in the relative abundance of the BSI strain in the gut microbiota (Figure 2). Specifically, there was a temporal association between a rise in the abundance of the BSI strain in fecal samples and exposure to an antibiotic to which this strain was resistant. In Patients 2 and 3, an increase in the relative abundance of the VRE strain in the gut microbiota was observed during exposure to vancomycin. In Patient 5, the relative abundance of E. coli in fecal samples increased from 12% (−12d) to 94% (−1d) while the patient was receiving cefepime for empirical treatment of febrile neutropenia. In Patient 6, the increase in abundance of E. coli from 0.2% (−10d) to 11% (−2d) occurred during treatment with trimethoprim-sulfamethoxazole. Finally, in Patient 7, exposure to piperacillin-tazobactam and subsequently moxifloxacin correlated with an increase in the abundance of E. coli from 18% (−25d) to 75% (−1d).
Gut colonization did not precede non-mucosal barrier BSI
To evaluate if the gut might serve as a reservoir for pathogens that are not classified as mucosal barrier injury organisms, we obtained the blood culture isolate for 3 patients with BSI episodes caused by Pseudomonas aeruginosa, Staphylococcus aureus, and Stenotrophomonas maltophilia, respectively. We performed whole-genome sequencing of the blood culture isolates and the fecal sample most closely preceding the BSI (P. aeruginosa, −1d; S. aureus, −2d; S. maltophilia, −4d). In each of these cases, we identified no sequencing reads in the fecal sample corresponding to the BSI species. Sequencing depth for these fecal samples (mean of 5,021,156 sequences per sample) was comparable to the sequencing depth for the fecal samples included in comparative genomics analyses of mucosal barrier injury BSI.
CONCLUSIONS
Using intensive prospective fecal sampling and high-resolution comparative genomics, this study demonstrates conclusively that the gut microbiota is a reservoir for bacterial strains that cause BSI after HSCT. This study further provides detailed information about pathogen dynamics within the gut that will inform future studies seeking to use applied microbiomics to predict and prevent BSI in HSCT recipients.
Prior studies in HSCT recipients focused primarily upon a relative abundance threshold in the gut microbiota that predisposes to BSI [6, 9, 10]. Most notably, Taur et al. established the concept of domination to represent when a single bacterial taxon comprises 30% or more of the gut microbiota and demonstrated that domination by Proteobacteria or Enterococcus was associated with BSI caused by these bacteria in adult HSCT recipients [6]. In support of these findings, more than half of the BSI episodes caused by E. coli and VRE in our cohort were preceded by gut domination by the identical strain for a median of 17 days. This was particularly striking for VRE, which comprised >99% of the gut microbiota in fecal samples preceding BSI episodes caused by this species.
In contrast, several patients developed E. coli BSI following a rapid rise in the relative abundance of E. coli in the gut microbiota in the absence of prolonged gut domination. Most strikingly, a patient developed an E. coli BSI following an increase in the relative abundance of E. coli in the gut microbiota from 0% to 29% over only 3 days. Similarly, another patient developed an E. coli BSI after the relative abundance of E. coli rose from 0.2% to 11% over an 8-day time period. This suggests that a rapid rise in the relative abundance of a strain in the gut microbiota may be an important harbinger of BSI, even if this strain is not highly abundant. Moreover, the rapidity of the shifts in gut microbiota composition after HSCT illustrates that frequent fecal sampling will be a necessary component of strategies seeking to use the gut metagenome for the prediction of BSI and other clinical outcomes in HSCT recipients.
Our results also indicate that serial monitoring of the gut microbiota is unlikely to be effective in predicting BSI episodes caused by organisms that do not typically have an ecological niche within the lower gastrointestinal tract. For instance, a patient in our cohort developed a BSI caused by R. mucilaginosa despite the absence of this species in several fecal samples closely preceding the BSI episode. Rothia species are typically abundant in the oral microbiota but are infrequently detected in samples from the lower gastrointestinal tract [11, 12, 13]. Interestingly, this patient had severe oral mucositis coinciding with the onset of the BSI episode. Moreover, we did not detect the BSI species in fecal samples preceding infections caused by several non-mucosal barrier injury pathogens (P. aeruginosa, S. aureus, and S. maltophilia). While each of these species was previously shown to colonize the gut of HSCT recipients, the predominant ecological niches for these organisms are the respiratory tract or skin [14, 15]. Tamburini et al. recently reported that strains of P. aeruginosa and Staphylococcus epidermidis detected in fecal samples were highly genetically similar to the blood culture isolates from BSI episodes occurring in adult HSCT recipients [7]. Notably, these strains were of low relative abundance (<3%) in the gut microbiota of these patients, and these species were not detected in fecal samples from several other patients who developed BSI episodes caused by these species. Thus, while it seems plausible that the gut could serve as a reservoir for typically non-enteric pathogens in HSCT recipients, it is likely that sampling of other body sites will be needed to predict many BSI episodes caused by these organisms.
The detection of antibiotic resistance genes predicted phenotypic resistance to several classes of antibiotics. As expected, we identified the vanA gene complex in both VRE isolates and several classes of extended-spectrum beta-lactamase (ESBL) genes in Enterobacteriaceae isolates with resistance to third-generation cephalosporins. These results support the utility of rapid molecular testing of positive blood cultures for the vanA and ESBL genes, as was previously described and is currently used in many clinical settings [16, 17, 18]. Reliably predicting phenotypic resistance to other classes of antibiotics with rapid molecular assays has proven to be more challenging. In our cohort, the detection of aminoglycoside, fluoroquinolone, tetracycline, and trimethoprim-sulfamethoxazole resistance genes correlated poorly with phenotypic resistance to these antibiotics, which may reflect the complex mechanisms that often account for resistance to these antibiotic classes in Enterobacteriaceae [19]. For instance, although the most common mechanism for high-level resistance to fluoroquinolones is though mutations in the genes that encode type II topoisomerases (gyrA, gyrB, parC), fluoroquinolone susceptibility can vary considerably even among strains with identical mutations in these genes, indicating the importance of other resistance mechanisms [20, 21].
This study has several limitations. First, it included a relatively small number of patients, and the majority of BSI episodes were caused by E. faecium or E. coli. While these species are among the most frequent pathogens in HSCT recipients, it is unclear if the dynamics of gut colonization that preceded these BSI episodes would apply to other bacterial species or fungi. In addition, the patterns of pathogen colonization that we describe prior to BSI onset are based on relative abundances that are necessarily correlated with the relative abundances of other gut microbes, and do not reflect absolute abundances. Moreover, although there were frequently temporal associations between antibiotic exposures and rising abundances of the BSI strains in the gut microbiota, other factors may have contributed in this medically complex patient population. Analyses were limited to children who developed a BSI episode, and it is likely that the observed patterns of pathogen colonization also occur in children who do not go on to develop BSI. We compared E. faecium and E. coli strains in fecal samples and blood at an unprecedented number of core genomic loci and applied a stringent definition of clonality, but there are no established cutoffs to determine definitively if two strains of a bacterial species are genetically-related. Finally, we were unable to compare the genetic similarity of the fecal strain to the blood culture isolate for two BSI episodes (patients 1 and 4) because too few reads were obtained for the BSI species in the gut microbiota.
In summary, we provide the most conclusive genomic data to date supporting a gastrointestinal source for BSI pathogens in HSCT recipients. High or rising relative abundances of bacterial strains in the gut microbiota preceded most BSI episodes in this small cohort. These data provide a basis for future studies evaluating the utility of serial whole-genome sequencing of the gut microbiota and other potential pathogen reservoirs to identify HSCT recipients who are high risk for BSI.
Supplementary Material
HIGHLIGHTS.
Gut colonization precedes most mucosal barrier injury BSIs in HSCT recipients
Antibiotics are associated with rising abundances of resistant strains in the gut
Detection of vanA and ESBL genes predicts phenotypic resistance in BSI strains
Acknowledgments
We appreciate F. Tamburini and A. Bhatt for their feedback on the study design and data analyses. We offer sincere gratitude to the children and families who participated in this study.
Funding
This work was supported by research grants from the Derfner Foundation and the Children’s Miracle Network. In addition, this research was supported in part by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (UM1AI104681). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. MSK was supported by a National Institutes of Health Career Development Award (K23-AI135090). VB was supported by the National Institute of Allergy and Infectious Diseases (R15-AI112985–01A1) and the National Science Foundation (1458347).
LIST OF ABBREVIATIONS
- BSI
bloodstream infection
- CARD
Comprehensive Antibiotic Resistance Database
- ESBL
extended-spectrum beta-lactamase
- HSCT
hematopoietic stem cell transplantation
- VRE
vancomycin-resistant Enterococcus faecium
Footnotes
Competing Interests
The authors have no competing interests to declare.
DECLARATIONS
Availability of data and material
The dataset supporting the conclusions of this study is available in the Sequence Read Archive (https://www.ncbi.nlm.nih.gov/bioproject/524322). The statistical code used for data analyses have also been made publicly available (https://gitlab.com/buccilab_public/pediatric_bsi).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES
- 1.Castagnola E, Bagnasco F, Farad M, et al. Incidence of bacteremias and invasive mycoses in children undergoing allogeneic hematopoietic stem cell transplantation: a single center experience. Bone Marrow Transplant. 2008;41:339–47. [DOI] [PubMed] [Google Scholar]
- 2.Almyroudis N, Fuller A, Jakubowski A, et al. Pre- and post- engraftment bloodstream infection rates and associated mortality in allogeneic hematopoietic stem cell transplant recipients. Transpl Infect Dis. 2005;7:11–7. [DOI] [PubMed] [Google Scholar]
- 3.Mikulska M, Del Bono V, Raiola AM, et al. Blood stream infections in allogeneic hematopoietic stem cell transplant recipients: reemergence of Gram-negative rods and increasing antibiotic resistance. Biol Blood Marrow Transplant. 2009;15:47–53. [DOI] [PubMed] [Google Scholar]
- 4.Krawczyk B, Śledzińska A, Szemiako K, et al. Characterisation of Escherichia coli isolates from the blood of haematological adult patients with bacteraemia: translocation from gut to blood requires the cooperation of multiple virulence factors. Eur J Clin Microbiol Infect Dis. 2015;34:1135–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Samet A, Śledzińska A, Krawczyk B, et al. Leukemia and risk of recurrent Escherichia coli bacteremia: genotyping implicates E. coli translocation from the colon to the bloodstream. Eur J Clin Microbiol Infect Dis. 2013;32:1393–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Taur Y, Xavier JB, Lipuma L, et al. Intestinal domination and the risk of bacteremia in patients undergoing allogeneic hematopoietic stem cell transplantation. Clin Infect Dis. 2012;55:905–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Tamburini FB, Andermann TM, Tkachenko E, et al. Precision identification of diverse bloodstream pathogens from the gut microbiome. Nat Med. 2018;24:1809–1814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Chen K, Pachter L. Bioinformatics for whole-genome shotgun sequencing of microbial communities. PLoS Comput Biol. 2005;1:106–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ubeda C, Taur Y, Jenq RR, et al. Vancomycin-resistant Enterococcus domination of intestinal microbiota is enabled by antibiotic treatment in mice and precedes bloodstream invasion in humans. J Clin Invest 2010; 120:4332–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Kamboj M, Blair R, Bell N, et al. What is the source of bloodstream infection due to vancomycin-resistant enterococci in persons with mucosal barrier injury? Infect Control Hosp Epidemiol. 2014;35:99–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Tsuzukibashi O, Uchibori S, Kobayashi T, et al. Isolation and identification methods of Rothia species in oral cavities. J Microbiol Methods. 2017;134:21–6. [DOI] [PubMed] [Google Scholar]
- 12.Zamakhchari M, Wei G, Dewhirst F, et al. Identification of Rothia bacteria as gluten-degrading natural colonizers of the upper gastro-intestinal tract. PLoS One. 2011;6:e24455. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Gevers D, Kugathasan S, Denson LA, et al. The treatment-naive microbiome in new-onset Crohn’s disease. Cell Host Microbe. 2014;15:382–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Nesher L, Rolston K, Shah D, et al. Fecal colonization and infection with Pseudomonas aeruginosa in recipients of allogeneic hematopoietic stem cell transplantation. Transpl Infect Dis. 2015;17:33–8. [DOI] [PubMed] [Google Scholar]
- 15.Simojoki ST, Kirjavainen V, Rahiala J, Kanerva J. Surveillance cultures in pediatric allogeneic hematopoietic stem cell transplantation. Pediatr Transplant. 2014;18:87–93. [DOI] [PubMed] [Google Scholar]
- 16.Banerjee R, Teng CB, Cunningham SA, et al. Randomized trial of rapid multiplex polymerase chain reaction-based blood culture identification and susceptibility testing. Clin Infect Dis. 2015;61:1071–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Birkett CI, Ludlam HA, Woodford N, et al. Real-time TaqMan PCR for rapid detection and typing of genes encoding CTX-M extended-spectrum P-lactamases. J Med Microbiol. 2007;56:52–5. [DOI] [PubMed] [Google Scholar]
- 18.Mancini N, Infurnari L, Ghidoli N, et al. Potential impact of a microarray-based nucleic acid assay for the rapid detection of gram-negative bacteria and resistance markers in positive blood cultures. J Clin Microbiol. 2014;52:1242–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Blair JM, Webber MA, Baylay AJ, Ogbolu DO, Piddock LJ. Molecular mechanisms of antibiotic resistance. Nat Rev Microbiol. 2015;13:42–51. [DOI] [PubMed] [Google Scholar]
- 20.Redgrave LS, Sutton SB, Webber MA, Piddock LJ. Fluoroquinolone resistance: mechanisms, impact on bacteria, and role in evolutionary success. Trends Microbiol. 2014;22:438–45. [DOI] [PubMed] [Google Scholar]
- 21.Weigel LM, Steward CD, Tenover FC. gyrA mutations associated with fluoroquinolone resistance in eight species of Enterobacteriaceae. Antimicrob Agents Chemother. 1998;42:2661–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Centers for Disease Control and Prevention. Bloodstream infection event (central line-associated bloodstream infection and non-central line-associated bloodstream infection). Atlanta, GA: Centers for Disease Control and Prevention; 2016;4:1–32. [Google Scholar]
- 23.Patel JB. Performance standards for antimicrobial susceptibility testing: Clinical and Laboratory Standards Institute; 2017. [Google Scholar]
- 24.Caporaso JG, Lauber CL, Walters WA, et al. Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc Natl Acad Sci USA. 2011;108 Suppl 1:4516–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.LaTuga MS, Ellis JC, Cotton CM, et al. Beyond bacteria: a study of the enteric microbial consortium in extremely low birth weight infants. PLoS One. 2011;6:e27858. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30:2114–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods. 2012;9:357–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Truong DT, Franzosa EA, Tickle TL, et al. MetaPhlAn2 for enhanced metagenomic taxonomic profiling. Nat Methods. 2015;12:902–3. [DOI] [PubMed] [Google Scholar]
- 29.Abubucker S, Segata N, Goll J, et al. Metabolic reconstruction for metagenomic data and its application to the human microbiome. PLoS Comput Biol. 2012;8:e1002358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.McArthur AG, Waglechner N, Nizam F, et al. The comprehensive antibiotic resistance database. Antimicrob Agents Chemother. 2013;57:3348–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kaminski J, Gibson MK, Franzosa EA, et al. High-specificity targeted functional profiling in microbial communities with ShortBRED. PLoS Comput Biol. 2015;11:e1004557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Bankevich A, Nurk S, Antipov D, et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J Comput Biol. 2012;19:455–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Li H, Handsaker B, Wysoker A, et al. The sequence alignment/map format and SAMtools. Bioinformatics. 2009;25:2078–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Walker BJ, Abeel T, Shea T, et al. Pilon: an integrated tool for comprehensive microbial variant detection and genome assembly improvement. PLoS One. 2014;9:e112963. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Page AJ, Taylor B, Delaney AJ, et al. SNP-sites: rapid efficient extraction of SNPs from multi-FASTA alignments. Microb Genom. 2016;2:e000056. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Stamatakis A RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics. 2014;30:1312–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Rambaut A FigTree: tree figure drawing tool. 2009.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.