Molecular epidemiology and antimicrobial resistance phenotype of paediatric bloodstream infections caused by Gram-negative bacteria.

Background: Gram-negative organisms are common causes of bloodstream infection (BSI) during the neonatal period and early childhood. Whilst several large studies have characterised these isolates in adults, equivalent data (particularly incorporating whole genome sequencing) is lacking in the paediatric population.
Methods: We perform an epidemiological and sequencing based analysis of Gram-negative bloodstream infections (327 isolates (296 successfully sequenced) from 287 patients) in children <18 years old between 2008 and 2018 in Oxfordshire, UK.
Results: Here we show that the burden of infection lies predominantly in neonates and that most infections are caused by Escherichia coli, Klebsiella spp. and Enterobacter hormaechei. There is no evidence in our setting that the proportion of antimicrobial resistant isolates is increasing in the paediatric population although we identify some evidence of sub-breakpoint increases in gentamicin resistance. The population structure of E. coli BSI isolates in neonates and children mirrors that in adults with a predominance of STs 131/95/73/69 and the same proportions of O-antigen serotypes. In most cases in our setting there is no evidence of transmission/point-source acquisition and we demonstrate the utility of whole genome sequencing to refute a previously suspected outbreak. Conclusions: Our findings support continued use of current empirical treatment guidelines and suggest that O-antigen targeted vaccines may have a role in reducing the incidence of neonatal sepsis.

Introduction

Gram-negative bloodstream infections (GNBSI) are a common cause of substantial morbidity and mortality globally in neonates and young children[1-4]. Their incidence has increased in both the UK and the US over the past decade, particularly in very low birth-weight neonates (VLBW, defined as <1500 g)[5,6]. Their association with antimicrobial resistance (AMR) has been highlighted by a recent study in the United States of 721 E. coli isolates (including 393 bloodstream infection [BSI]-associated isolates), which found high rates of non-susceptibility to commonly used empirical antibiotics, including ampicillin (66.8%) and gentamicin (16.8%), as well as an extended beta-lactamase (ESBL) phenotype in 1 in 20 cases[7]. A recent study of 2483 neonates with culture-confirmed sepsis in low and middle-income countries showed that Klebsiella spp. was the predominant pathogen causing multidrug-resistant neonatal sepsis[8]. In Greece, a retrospective observational study in 16 neonatal intensive care units (NICUs) revealed almost half (45%; 36/80) of Klebsiella spp. were resistant to either gentamicin or amikacin[9]. The ability of many Gram-negative bacilli (GNB) to readily acquire and exchange genetic material (particularly AMR genes [ARGs]) via mobile genetic elements means that the proliferation of drug-resistant strains remains a constant threat.

The molecular epidemiology of E. coli and Klebsiella spp. isolates causing invasive infection in adults has been characterised in large sequencing studies.[10,11] These have demonstrated the emergence of particular AMR-associated sequence types (e.g., E. coli ST131)[12], the genetic homogeneity of isolates causing community and nosocomial onset infections suggesting a common reservoir[13], the potential for vaccines to play a role in reducing the incidence of these infections[14,15], and the emerging threat of the convergence of multidrug resistance and hypervirulence in Klebsiella spp[16]. To our knowledge, no study to date has systematically evaluated the molecular epidemiology of E. coli/Klebsiella spp. and other common causes of GNBSI in a paediatric population; published studies focus predominantly on evaluations of outbreaks caused by AMR-associated strains and/or on neonates (see above)[8,17,18]. In this study, we, therefore, aimed to investigate sequencing data from a relatively large collection of sequentially acquired, unselected bloodstream isolates from neonates and children presenting to hospitals in Oxfordshire, UK, over the past decade. We find that GNBSIs in this population are primarily caused by E. coli/Klebsiella spp./Enterobacter hormaechei, with the greatest burden of disease occurring in neonates; overall, there were no substantial changes in antimicrobial susceptibility, supporting the continued use of current empirical regimens. We demonstrate that the population structure of E. coli in the paediatric population mirrors that seen in adults and demonstrate the utility of WGS to monitor determinants of AMR and the genomic relatedness of isolates, leading us to refute a previously suspected outbreak.

Methods

Isolate selection

Oxford University Hospitals NHS Foundation Trust is a large healthcare provider in the South East of England, serving a paediatric population of ~142,000 across four hospitals (of which two have emergency and acute general paediatric medicine, and one provides all neonatal/paediatric critical care and specialist paediatric services for the region). The microbiology laboratory additionally provides a service to all regional community healthcare providers. All E. coli and Klebsiella spp. isolates (deduplicated to one morphotype per 90-day period) from Oct-2008 to Nov-2018 collected from blood cultures of patients <18 years old on the day of collection were included in the study. The same selection criteria were applied to other GNB from August 2011 to September 2018, which were excluded from the initial period due to resource limitations. Prior to 2013, antimicrobial susceptibility testing was performed using disk diffusion; after this, the Phoenix BD system was used with European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints. Amikacin susceptibility phenotyping was not routinely performed prior to 2013.

Sequencing procedures

Frozen stocks were sub-cultured onto Columbia blood agar and incubated overnight at 37 °C. DNA extractions were performed using the QuickGene DNA extraction kit (Autogen, MA, USA) as per the manufacturer’s instructions (with an additional mechanical lysis step—FastPrep, MP Biomedicals, CA, USA; 6 m/s for 40 s, done twice). Sequencing was performed using Illumina HiSeq 2500/3000/4000/MiSeq instruments as described previously[19]. Briefly, library preparation was performed using the NEBNext Ultra DNA Sample Prep Master Mix Kit (New England Biosciences). Illumina Multiplex Adaptors were used to create ligated libraries with subsequent size selection using Agencourt Ampure magnetic beads (Beckman Coulter), followed by PCR-based enrichment and adaptor extension. We purified the products using Agencourt Ampure XP beads (Beckman Coulter) using a Biomek NXp per the manufacturers' instructions. We used a Tapestation system to assess size distributions of library preparations and quantified their concentration using a Qubit system. All sequencing data have been deposited under NCBI accession number PRJNA604975.

Bioinformatics

De novo assembly was performed using Shovill (v1.0.4)[20]. Reads were mapped to sequence type (ST[21,22]) specific references using Snippy (v4.6.0)[23] (Table S1). For the four E. coli STs with the largest number of isolates (i.e., E. coli ST131/95/73/69), we created core genome alignments using Snippy-core with the –mask auto setting, padded with the reference base at invariant positions; these whole genome alignments were used as input to Gubbins (v2.3.4)[24]. Such recombination-corrected phylogenies were also created for E.hormaechei (the most common non-E. coli/Klebsiella species detected in our study) and S. marcescens (because there was thought to have been an outbreak in our neonatal intensive care unit in 2016). We also used genomic distances calculated by Mash[25] (using -k21 -s 100,000 for within-study comparison of isolates and -k21 -s 1000 for comparison of isolates in this study to our collection of sequences of adult bloodstream infection isolates[26] from the same region over the same time period due to computational feasibility). Annotation against reference databases (VFDB/ResFinder) was performed using ABRicate (v2.3.4)[27] with genes called as being present if there was ≥80% coverage and DNA identity compared to the reference. Sequence types were predicted using the MLST tool (v2.19)[28]. For Klebsiella spp. isolates speciation and virulence gene detection (Supplementary Methods) was performed using Kleborate (v2.0.4)[29]. Detailed QC metrics and raw Abricate/Kleborate output have been uploaded to Figshare (https://figshare.com/projects/Paediatric_GNBSIs_in_Oxfordshire/135254).

Definitions

We defined isolates as being likely of neonatal origin if they originated from infants (i) in their first 30 days of life, or were (ii) in the neonatal intensive care unit, or (iii) under the care of a neonatologist on the day the blood culture was taken. For analytical purposes, we classified other children as <12 months, 1–4 years, 5–9 years and 10–17 years of age. Early-onset infection was defined as a disease within the first 72 h of life[30]. We further categorised BSIs according to healthcare exposure prior to onset as follows: nosocomial (>48 h after admission to hospital), ‘quasi-nosocomial’ (within 30 days of last discharge), ‘quasi-community’ (31–365 days since last discharge) and community (>365 days since last discharge)[31]. Genetic relatedness in the form of single nucleotide polymorphism (SNP) thresholds definitively associated with the transmission is variably defined for the species evaluated; based on recent studies, we considered a threshold of >20 SNVs between isolates as highly unlikely to be representative of a transmission event[32].

Epidemiology/statistics

Routinely collected healthcare data were acquired via pseudonymised linkage in the Infections in Oxfordshire Research Database (IORD). IORD has generic Research Ethics Committee, Health Research Authority and Confidentiality Advisory Group approvals (19/SC/0403, 19/CAG/0144) as a de-identified electronic research database. Data on suspected infectious focus (only available for E. coli/Klebsiella spp.) were acquired via linked local infection control records, which had been submitted to Public Health England as part of the mandatory surveillance programme. Likely source was identified by infectious disease/microbiology physicians using best clinical judgement or designated as an unknown where there was uncertainty. For each species, we modelled the number of bloodstream infections (BSIs) per year using negative binomial regression, with the total number of paediatric admissions in each year used as an offset to account for changes in the population over time. Only complete years (i.e., excluding 2008 and 2018) were included in this part of the analysis. All statistical analysis was performed in R v4.0.3[33,34].

Ethics statement

The Infections in Oxfordshire Research Database (IORD; https://oxfordbrc.nihr.ac.uk/research-themes-overview/antimicrobial-resistance-and-modernising-microbiology/infections-in-oxfordshire-research-database-iord/) has generic Research Ethics Committee, Health Research Authority and Confidentiality Advisory Group approvals (19/SC/0403, 19/CAG/0144) which facilitate the pseudo-anonymised linkage of routinely collected NHS electronic healthcare record data from the Oxford University Hospitals NHS Foundation Trust Clinical Systems Data Warehouse and research data (e.g., sequencing data) from the Antimicrobial Resistance and Modernising Microbiology Theme of the Oxford NIHR Biomedical Research Centre, Oxford. IORD links records by a specific, random number ensuring that no patient-identifiable information is shared with researchers using this resource. Individual informed consent is not required under these permissions, which allow the lawful collection, storage and use of this data as a ‘Public Task’ under GDPR; individuals can opt-out of having their data included should they wish. Further details are available at https://oxfordbrc.nihr.ac.uk/wp-content/uploads/2020/01/IORD-privacy-note-2019-10-29.pdf. We sequenced bacterial isolates from bloodstream infections that are routinely stored by the John Radcliffe Hospital Microbiology laboratory. In the UK, bacterial isolates (such as those sequenced in this study) routinely cultured from human clinical samples do not require ethical approval for analysis under the provisions of the Human Tissue Act as they do not contain any material considered to be human tissue.