Predictors of the extended-spectrum-beta lactamases producing Enterobacteriaceae neonatal sepsis at a tertiary hospital, Tanzania

Highlights • ESBL-PE sepsis was predicted by admission at ICU and ESBL-PE colonization.• Neonates infected with ESBL-PE had significantly high mortality.• ESBL-producing Klebsiella pneumoniae (ST45) carrying blaCTX-M-15 were predominant.• Whole genome SNP analysis revealed clonal origin in 50% of ESBL-PE paired cases with similar sequence type.


Introduction
Neonatal sepsis is one of the top three causes of morbidity and mortality during the neonatal period (Bhutta et al., 2010). Diagnosis and management of sepsis pose a great challenge for neonatologists in NICUs especially in low-income countries (Sharma et al., 2016). The increase of multidrug-resistant organisms in neonatal units limits the treatment options and delays effective treatment, resulting in increased morbidity and mortality (Patel and Saiman, 2010). In many low-income countries including Tanzania, extended-spectrum beta-lactamase-producing Enterobacteriaceae (ESBL-PE) are the commonest cause of neonatal sepsis (Christopher et al., 2013;D'Andrea et al., 2013;Ghotaslou et al., 2007;Mshana et al., 2009;Thaver et al., 2009). Infections due to ESBL-PE are associated with increased morbidity and mortality (Blomberg et al., 2005;Kayange et al., 2010;Mhada et al., 2012). The antibiotics of choice for the treatment of neonatal sepsis due to ESBL-PE pathogens are not available in most settings and even when stocked, they are too expensive to be afforded (Vandijck et al., 2008). Varieties of ESBL-PE genotypes have been found in humans, animals and the environment in the city of Mwanza, Tanzania . Despite ESBL-PE being prevalent in the city of Mwanza, the transmission pathways and risk factors of these pathogens in relation to neonatal sepsis is not well studied. This study was undertaken to investigate factors associated with ESBL-PE neonatal sepsis and mortality among neonates admitted at the Bugando Medical Centre (BMC), Mwanza, Tanzania. In addition, the study aimed to characterize selected isolates to show their virulence potential and transmission dynamics.

Study duration, area and inclusion criteria
The study was conducted between July and December 2016 at the premature unit and neonatal Intensive care unit (NICU) of BMC. All neonates (0-28 days) with signs and symptoms of sepsis were enrolled. The sample size was estimated using the Kish Leslie formula for the cross sectional studies (Kish, 1965).

Data collection
Using signs and symptoms published by "The WHO Young infants Study group" (Margolis et al., 1999), a data collection tool was designed and used to obtain socio-demographic and clinical data together with other relevant factors related to risks for neonatal sepsis. At the time of enrollment, clinical data and blood samples were collected. Antenatal history such as use of antibiotics, presence of chronic diseases, febrile illness during pregnancy, labor history, duration of labor, history of premature rupture of membrane, mode of delivery were recorded. Neonates were subjected to a full clinical examination to assess temperature, respiration rate, presence or absence of cyanosis, jaundice, umbilical redness, convulsion, reduced movement and ability to feed.

Microbiological procedures
Umbilical-rectal swabs were collected using sterile cotton swabs in Amies transport medium (Biolab, HUNGARY) from all neonates to determine colonization status. Analysis of stool samples obtained from mothers/guardians provided data on colonization status. Swab and stool samples were inoculated on MacConkey agar (Oxoid, UK) supplemented with 2 μg/ml cefotaxime (Medochemie Ltd, Cyprus EU) and processed as previously described (Nelson et al., 2014). Using aseptic techniques approximately 1.5-2 ml of blood was obtained and inoculated directly into Brain Heart Infusion broth (BHI) (Oxoid Ltd, UK) in a ratio of blood to BHI of 1:10 and transported to the Catholic University of Health and allied Sciences (CUHAS) Microbiology laboratory for incubation and subsequent processing as previously described (Kayange et al., 2010). All blood culture isolates were tested for their antibiotic susceptibility pattern using disc diffusion methods and interpreted following the Clinical Laboratory Standard Institute guidelines (CLSI, 2015). Confirmation of the ESBL phenotype was performed using the disk approximation method (Mshana et al., 2009). All neonates enrolled in the study were assessed and managed adhering to the BMC neonatal unit protocols. The first line antibiotic treatment option included ampicillin and gentamicin, the second line included cefotaxime followed by the third line that incorporated the use of meropenem. All neonates were evaluated on a daily basis prior to discharge, after which they were monitored for a further 28 days.

Whole genome sequencing and in silico analyses
Fifty-three (16 from blood and 37 from colonized neonates) ESBL-PE isolates were subjected for whole genome sequencing (WGS). These isolates were serially selected as they were isolated, no other criteria were used, and aim was to sequence the first 50 isolates. DNA was isolated using PureLink® Pro 96 Genomic DNA Purification Kit (Thermo-Fisher Scientific, Dreieich, Germany) according to the manufacturer's instruction. WGS was performed on an Illumina NextSeq 500 instrument (Illumina, San Diego, CA, USA) using an Illumina Nextera XT library with 2 × 150bp paired-end reads. The data was assembled using SPAdes (version 3.6.2) (Bankevich et al., 2012).
The SNP analysis of isolates depicting identical STs was performed using SAMtools (Li, 2011). Regions of high recombination (> 50 SNPs per 1 kb) were excluded manually from the analysis.
The raw data of all sequenced ESBL-PE isolates are available at the European Nucleotide Archive (ENA) under the project number PRJEB20875.

Ethics approval
This study obtained clearance from the Joint CUHAS-BMC Ethics and Scientific Review committee with certificate no: CREC/162/2016. Written informed consent was obtained from parents/guardians of the study participants.

Statistical analysis
Data was entered using Microsoft Excel 2007 and processed using STATA version 13.0 (Stata Corp, College Station, TX, USA). Categorical variables such as history of antibiotic use, history of admission, colonization status etc., were presented proportionally and compared using chi squared test or Fisher's exact test. Continuous variables such age, gestation age, length of hospital stay, weight etc., were described as medians (inter quartile range/ IQR)). To determine predictors of ESBL-PE neonatal sepsis, ESBL-PE neonatal colonization and the outcome, univariate followed by multivariate logistic regression analysis was performed. All factors with p value of less than 0.2 were further subjected to multivariate logistic analysis and controlled for both age and sex. To estimate survival rates and hazard ratios, a Cox regression model was used. All factors with p value less than 0.05 at 95% confidence interval (CI) were considered statistically significant.

Demographic and clinical characteristics
A total of 304 neonates were enrolled in the study. Male neonates formed the majority 192 (63.2%) of study participants. The median age of the enrolled neonates was 6 days (IQR: 3-9). A total of 94 (30.1%) neonates had birth weights of below 2.5 kg. The median gestation age was 39 weeks (IQR: 37-40) and the median body temperature measured was 37.9°C (IQR-37-38.2°C). A total of 57 (14.1%) of all neonates were delivered by cesarean section (C/S) and 16 (5.3%) were delivered at home (Table 1).
An Enterobacter cloacae from a swab was typed as ST93, while another isolate obtained from blood was classified as ST116. Both Enterobacter cloacae isolates harboured IncHI2 plasmids were assigned to the plasmid (p) MLST ST1. Resistance genes detected in Enterobacter cloacae included bla ACT-7 , bla TEM-1B , bla OXA-1 and bla CTX-M-15 . The ST93 strain did not harbor a qnrB1 gene but carried sul2, dfrA1 and aadA2 genes that were not present in ST116. The three Acinetobacter baumannii isolates were members of ST1470 and ST405 (n = 2) and carried carbapenemase resistance genes bla NDM-1, bla OXA-58 , bla OXA-69 , bla CARB-8 and bla OXA-69 (Table 4). IncF plasmids of pMLST K5 and FI were predominant in Klebsiella pneumoniae strains and Escherichia coli respectively, while for Enterobacter spp. the predominant plasmid was IncHI2.

Virulence genes of Klebsiella pneumoniae
Klebsiella pneumoniae isolates harboured up to four different virulence-associated determinants (Fig. 1). The yersiniabactin operon (ybtAEPQSTUX, irp1, irp2, fyuA) a metallobactin involved in iron acquisition was detected in 32/38 (84.2%) of the examined Klebsiella pneumoniae isolates. Other isolates harbouring yet another operon involved in iron acquisition (kfuABC), which was detected only in 4/38 isolates. Genes associated with pilus formation (mrkABCDFHIJ) were present in all but one of the (42i) Klebsiella pneumoniae examined. Four isolates harboured the kvgAS two-component system involved in sensing iron-limiting conditions and free radical stress.

Transmission of Klebsiella pneumoniae. Isolates to the blood
Twelve pairs of ESBL-PE between neonates blood and neonate swab were detected. In eight neonates, blood culture and neonatal swab isolates depicted identical sequence types. Single nucleotide polymorphisms were identified between these pairs to detect close relationship that indicates a transmission to the blood. Four pairs (50%) depicted very low SNP counts (< 20 SNPs) indicating the isolates are of clonal origin. In the other pairs, higher SNP counts were identified, indicating some relationship (Table 5).

Predictors of ESBL-PE neonatal sepsis
We investigated different factors to establish whether they could predict ESBL-PE neonatal sepsis. The median age of neonates with ESBL-PE neonatal sepsis was 6 days (IQR: 4-8) while that of ESBL-   (Table 6).

Predictors of ESBL-PE neonatal colonization
A total of 166 (54.6%; 95%CI: 49-60.1) neonates were colonized by ESBL-PE. Following univariate analysis, out of the 82 neonates with history of antibiotic use prior being admitted at PREM/NICU, 54(65.9%) were found to be colonized by ESBL-PE as compared to 112 (50.5%) of those without history of antibiotic use (P = 0.017). It was further noted that neonates born from mothers who were colonized by ESBL-PE were more likely to be colonized than those born from mothers with no ESBL-PE colonization (68.6% vs. 49.1%, OR: 2.27, 95%CI 1.34-3.84, P = 0.002). Using multivariable logistic regression analysis, predictors of neonatal ESBL colonization were history of antibiotic use (P = 0.048) and positive ESBL-PE colonization of the mother (P = 0.005) ( Table 7).

Discussion
The current study has demonstrated that ESBL-PE contributed to half of the blood culture isolates obtained from neonates in a tertiary hospital in Mwanza, Tanzania. The overall prevalence of the laboratory confirmed sepsis (20%) in the present study is lower than the 50% reported previously from the same centre (Kayange et al., 2010). The difference is probably the result of prophylactic use of broad-spectrum antibiotics in the current cohort. In addition, the proven occurrence of sepsis is also significantly lower than that observed in Mulago National hospital, Kampala Uganda (Mugalu et al., 2006). However, the level of sepsis in the current study is comparable to that observed in Muhimbili national hospital neonatal unit (Mhada et al., 2012).
In low-income countries the major burden of sepsis is borne by children, who are 18 times more likely to die before the age of five than those children in higher-income countries, with infections being a significant source of this disparity (Chan and Lake, 2012). We found that admission to NICU, colonization of neonates, and their mothers/guardians with ESBL-PE were independent predictors of ESBL-PE neonatal sepsis. Neonates infected with ESBL-PE had significantly higher mortality and the majority had severe illness requiring NICU admission. The practice in the neonatal unit/NICU considers carbapenem treatment as the last resort antibiotic. The sad fact of substantial mortality in neonates suffering from blood stream infection due to bacteria with ESBL enzymes that has been demonstrated in the study may very well be a consequence of the unavailability of antimicrobial agents active against those bacteria or due to delay initiation of the appropriate therapy. The slow turnaround times of current blood culture tests lead to a considerable delay in initiating of the appropriate antibiotic treatment and thereby promotes mortality among neonates infected with ESBL-PE. In the current study, the preliminary blood culture results were only obtained after 48 h, with final results for the positive culture given only after 72 h, with incubation for 7 days to confirm negative culture in the majority of cases. The introduction of the semiautomated assays will improve turnaround time of the blood culture and significantly save lives of neonates infected with ESBL-PE. Clinicians should initiate appropriate treatment to high-risk neonates especially those admitted in NICU.
Despite being a cross-sectional study, the study has highlighted important factors that might predict ESBL-PE neonatal sepsis. Previous studies (Kritsotakis et al., 2011;Le et al., 2008;Mshana et al., 2016;Sandiumenge et al., 2006) have demonstrated that exposure to antibiotics can lead to infections/colonization with ESBL-PE, this was also the case in the present study. As observed in previous studies (Chacha et al., 2014;Kayange et al., 2010), Klebsiella pneumoniae was the predominant species isolated. However, unlike a previous study where ST14 was predominant , in the current study ST45 has been found to be the most common clone. We conclude that the majority of ESBL-PE sepsis are endogenous since isolates of Klebsiella pneumoniae ST45 was both a predominant colonizer of the neonate microflora as well as causing infections at the same time. These data show that most ESBL-PE neonatal sepsis within a given time period might be caused by clonally related isolates as observed in the current study as evidenced by the fact that 50% of paired isolates (blood: neonatal swab) with similar sequence types were clonal. It should be noted that, these isolates could also originate from guardians/mothers, environment or contaminated hands of healthcare workers, underscoring a follow-up exploration study for other sources of neonates colonization. It should be emphasized that hand hygiene can reduce healthcare-associated infections, including those caused by ESBL-PE (Allegranzi and Pittet, 2009). Interestingly, ST45 isolates either colonizing the neonates or associated with neonate infections, harboured different plasmid replicon types and indicate that these strains might have either evolved separately or promiscuously acquired plasmids from different sources. Thus, blood isolates harboured IncFIB and IncHI1 plasmids of hitherto unclassified pMLST types, while ST45 strains colonizing the neonates had IncFII, IncFIB and IncR plasmids.
The majority of Klebsiella pneumoniae isolates harboured genes for siderophores involved in sequestering iron from the environment. The presence of siderophores have been linked to community infections (Holt et al., 2015) emphasizing the broad occurrence of ESBL-PE in our setting. Surprisingly only two Escherichia coli strains, both of which were not ESBL, were isolated from the blood despite the fact that about 70% and 36% of guardians/mothers and neonates were colonized by ESBL-producing Escherichia coli. The intrinsic ability of Klebsiella pneumoniae to survive in the environment and its ability to survive in the serum, through siderophores-dependent iron acquisition, could partially account for its transmission to neonates and development of blood stream infections (Podschun and Ullmann, 1998). Indeed in the majority of the isolates sequenced, virulence factors predisposing to infections were detected.
As in previous studies (Bhutta et al., 2010;Mshana et al., 2011Mshana et al., , 2016Seni et al., 2016) in the hospital and community among humans and animals, 90% of isolates were found to carry bla CTX-M-15 . Acinetobacter baumannii carrying a bla NDM-1 gene was detected as a colonizing isolate. This is worrisome because the inappropriate use of meropenem could promote selective growth and hence increase challenges in the management of the neonatal sepsis in low-income countries. Furthermore the bla NDM-1 might translocate to conjugative plasmids and be transmitted to other Enterobacteriaceae through horizontal genetic transfer (Struelens et al., 2010).
The limitations of this study include the following: Because of the low number of culture-positive blood samples we decided to compare those with ESBL-PE neonatal sepsis and other neonates assuming that all neonates were, from the onset, equally at risk of getting ESBL-PE neonatal sepsis. Being a single center, the results may not be generalizable. A prospective multicenter cohort study would have been more suitable. Another limitation was the lack of colonization and mortality data from neonates without signs and symptoms of neonatal sepsis for comparative analysis. Finally, not all isolates underwent molecular characterization including those colonizing the mother/guardians. Future studies should consider whole genome sequencing of larger numbers of isolates from neonates, guardian/mothers and health care workers to reveal dynamics of transmission and other source of ESBL-PE.

Conclusion and recommendation
The majority of septic neonates has been shown to be colonized by ESBL producers and among those colonized with strains secreting ESBL enzymes; the majority was shown to be colonized by ESBL-producing strains of K. pneumoniae. ESBL-producing Klebsiella pneumoniae was the most frequent isolated from blood of neonates at BMC neonatal unit. The virulent isolates of the Klebsiella pneumoniae ST45 carrying bla CTX-M-15 were predominantly infecting and colonizing neonates. Maternal and neonatal ESBL-PE colonization are important factors predicting ESBL neonatal sepsis. Our study demonstrates that neonatal mortality is significantly high in neonates infected with ESBL-PE. There is a need to improve and emphasize on infection control interventions and antimicrobial stewardship because these are cheap inexpensive strategies to control the increase of multi-resistant gram-negative bacteria. Furthermore, rapid diagnostic tests to detect ESBL-PE from neonates' blood are highly needed in the low-income countries in order to reduce the mortality with ESBL-PE neonatal sepsis. In addition, there is a need to improve turnaround time of blood culture assays for the neonatal units in order to guide antibiotic treatment. NB: Blood culture and Bacteria/gram reaction were not subjected on multivariable analysis due to collinearity with ESBL-PE sepsis.