Whole-Genome Sequencing of Invasive Neonatal Escherichia coli From Uppsala County, Sweden

Abstract

Background

This study sought to investigate associations between virulence factors and phylogeny in all neonatal Escherichia coli bloodstream infections from patients admitted to the neonatal intensive care unit at Uppsala University Hospital between 2005 and 2020.

Methods

A total of 37 E. coli isolates from 32 neonates were whole-genome sequenced and analyzed for virulence factors related to extraintestinal E. coli; patient-related data were collected retrospectively from the medical records.

Results

E. coli isolates that belong to phylogroup B2 were associated with mortality (odds ratio [OR], 26; P < .001), extreme prematurity with delivery before gestational week 28 (OR, 9; P < .05), and shock (OR, 9; P < .05) compared with isolates of non-B2 group. Female neonates were more often infected with isolates of phylogroup B2 E. coli compared with male neonates (OR, 7; P = .05). The identification of the genotoxin determinant clb coding for colibactin exhibited strong associations with mortality (OR, 67; P < .005), gestational age (OR, 18; P < .005), and shock (OR, 26; P < .005).

Discussion

The study highlighted the correlation between neonatal E. coli bacteremia caused by phylogroup B2 and the role of colibactin. Results emphasize difference between male and female neonates in E. coli populations in bloodstream infections.

Graphical Abstract

Open in new tabDownload slide

Escherichia coli bloodstream infections present a substantial challenge within neonatal care and are acknowledged to be an important global public health concern [1]. In particular, the incidence of early-onset infections—occurring within the first 72 hours after delivery—have risen among extremely premature infants born before 28 gestational weeks [2]. Identified risk factors for early-onset E. coli infection in neonates are intrapartum fever, prolonged rupture of membranes, and chorioamnionitis, and it is associated with prematurity [2, 3].

In contrast, late-onset infections, which manifest after 72 hours postnatally and are caused by invasive E. coli, are more often linked to compromised barrier defenses that lead to severe conditions such as necrotizing enterocolitis or birth defects that contribute to infections in the urinary tract [4]. Additionally, an extended stay in the intensive care unit heightens the risk for nosocomial infections due to gram-negative rods such as E. coli [5]. The widely adopted prevention strategies for neonatal infections involve risk stratification and antibiotic treatment policies for both mother and child. However, these approaches pose challenges in balancing the risks associated with the spread of antimicrobial resistance and the prevention of infections in a vulnerable patient group [3].

E. coli is renowned for its highly adaptable pathogenic nature, and its pathovars are well-adapted to diverse ecological niches. Extraintestinal pathogenic E. coli (ExPEC) are responsible for infections in typically sterile body sites such as the bloodstream and urinary tract, or neonatal meningitis. These strains are disproportionately found in phylogenetic groups B2 and D [6]. A multitude of virulence factors contribute to the pathogenicity of ExPEC isolates. Interestingly, when present in phylogroups A and B1, commonly associated with commensalism, these virulence factors can lead to lethal infection in mice [7].

This study sought to investigate associations between a broad spectrum of virulence factors and phylogeny in all neonatal E. coli bloodstream infections from patients admitted to the neonatal intensive care unit at Uppsala University Hospital between 2005 and 2020, and to examine relationships with neonatal characteristics.

METHODS

Patients

At Uppsala University Hospital, all blood stream and cerebrospinal fluid isolates were stored in −80°C freezers in glycerol stock at the Department of Clinical Microbiology. For the purpose of the study, a search for E. coli isolates was conducted for all patients treated at the Neonatal Intensive Care Unit, Uppsala University Hospital, during the calendar years 2005 to 2020, and a total of 32 neonates were identified. For all patients, a review of the clinical records regarding the perinatal and neonatal care periods was performed. Informed consent from the included subjects was not necessary; the study was approved by the Swedish Ethical Review Board (document number 2020-03297).

The review of the neonates’ records included collection of information about mortality, gestational week, weight, and sex. Infections were divided into early onset (debut 0–3 days postnatally) and late onset (after 3 days postnatally), which was confirmed by a positive blood culture. Furthermore, information about perinatal infections in the mother and/or the neonate was registered, and included necrotizing enterocolitis, chorioamnionitis, and urinary tract infection. Other laboratory parameters for the investigated E. coli bacteremia were included as follows: C-reactive protein (mg/L), bilirubin (μmol/L), hemoglobin (g/L), white blood cell count (×10⁹/L), and platelet count (×10⁹/L) when blood cultures were taken and peak values for C-reactive protein and bilirubin after blood culture samples. Furthermore, the symptoms when the blood culture was taken were evaluated and categorized as (pre)shock (fluid substitution and/or catecholamines were given), impaired (affected general condition regarding vital signs, respiration, circulations), and no symptoms (good general condition).

Culture-Based Methods

Blood or cerebrospinal fluid specimens were inoculated in BacT/Alert pediatric bottles including antimicrobial neutralization media and promptly placed in a BacT/Alert (Biomérieux) incubator. Cultures with growth were plated on blood agar, hematin agar, cysteine-lactose electrolyte-deficient, and fastidious anaerobic agar plates, and incubated for 2 days at 35°C in 5% CO₂, air, and anaerobic conditions, respectively. Species identification was performed using standard laboratory procedures and VITEK (before 2013; Biomérieux) and MALDI-TOF (since 2013; Bruker). All isolates were stored in glycerol stock at −80°C.

Whole-Genome Analysis

All E. coli isolates were analyzed by whole-genome sequencing. Extraction and library preparation was performed as described elsewhere [8]. Sequencing was performed on an Illumina HiSeq4000 platform ( Oxford Genome Center, Oxford, United Kingdom) or Illumina NovaSeq 6000 SP platform (National Genomics Infrastructure, Stockholm, Sweden), generating 150-bp paired-end reads.

Paired-end reads were assembled using SPAdes assembler (version 3.11.1, https://github.com/ablab/spades) using the −careful flag and kmer length 21, 33, 55, 77, 99, and 127. Species confirmation was performed using the ribosomal multilocus sequence typing (rMLST) species tool on http://pubmlst.org/rmlst. All sequence data are available through the European Nucleotide Archive (www.ebi.ac.uk/ena) under project accession number PRJEB62832.

Composition of an ExPEC Virulence Factor Database and Application With the Study Isolates

The ExPEC database was constructed through an intense literature search for previously described virulence factors that have been associated with ExPEC. ExPEC-associated genes and gene alleles were included in the database when a nucleotide entry in the NCBI GenBank nucleotide collection (www.ncbi.nlm.nih.gov/nucleotide) and/or UniProt (www.uniprot.org) was linked to experimental data showing evidence for significant association with extraintestinal E. coli infections. The database included 674 entries corresponding to 133 virulence factors that are assigned to the following functional groups: fimbriae and adhesins (n = 67), iron metabolism (n = 13), exotoxins (n = 19), immunomodulation (n = 15), bacteriocins (n = 29), and capsule types (n = 2) (Table 1, Supplementary Table 1, and database ExPEC.fasta). Gene prediction on draft genomes of the 37 study isolates was performed using BLASTn algorithm with standard settings and virulence factors with a nucleotide coverage ≥ 99% and nucleotide identity ≥ 98% [9].

Calculations and Statistics

Genetic determinants for virulence factors were treated as presence/absence data; mortality, gestational age categorized by delivery before and after 28 gestational weeks, symptoms, necrotizing enterocolitis, chorioamnionitis, urinary tract infection, and maternal infection perinatally as dichotomic variables; and C-reactive protein, hemoglobin, bilirubin, white blood cell count, and platelet count as continuous variables. All data were tabulated and prepared in Microsoft Excel for further statistical computations, which were performed using the statistical software R (version 4.3.0, 2023-04-21; R Foundation for Statistical Computing).

Univariate description of data involved median and range or absolute and relative frequencies where appropriate. Analysis of associations between dichotomous variables was performed using Pearson correlation (ϕ coefficient for binary variables) as implemented in basic R (“cor” function, “corrplot,” corrplot package) for all variables with more than 4 observations. For variable pairs with correlation or ϕ coefficients of > 0.4 or < −0.4, the correlation was examined for meaningfulness. Subsequently, for meaningful associations, hypotheses regarding clinical parameters were formulated and odds ratios (OR) and statistical significance were calculated by Fisher exact test for small sample sizes.

The phylogenetic relationship of all isolates was analyzed using rMLST allele types in a cluster analysis calculated with MS Tree V2 algorithm as implemented in Enterobase (https://enterobase.warwick.ac.uk). The isolates’ phylotypes were determined by Clermont typing, core genome MLST (cgMLST), and sequence types (ST) with concordant ST-clonal complexes using the 7-gene Achtman scheme were determined by Enterobase. Associations were evaluated graphically among mortality, gestational age, sex, symptoms, necrotizing enterocolitis, chorioamnionitis, urinary tract infection, and maternal infection perinatally to phylogenetic lineages according to Clermont, and where appropriate ORs and statistical significance were calculated using Fisher exact test for small sample sizes. Graphical illustrations of metadata were accomplished using GraphTree [10].

RESULTS

Patients

During the 15-year period (2005–2020), in 14 very premature neonates (gestational age 21.8–26.3 weeks), 10 premature neonates (28.3–35.8 weeks), and 8 full-term neonates (36.7–42 weeks) E. coli blood stream infections were identified. For all 32 neonates, information on gestational age, weight, and sex were accessible. The neonates were born at a median gestational age of 29.6 weeks (range, 21.8 to 42 weeks) with a medium weight of 1140 g (range, 417 to 4080 g) corresponding to 7 of 32 (22%) infants being small for gestational age (below −1.5 × standard deviation), 22 of 32 (69%) being appropriate for their gestational age, and 2 of 32 (6%) being large for their gestational age (above 1.5 × standard deviation). More neonates were male (22/32, 69%) than female (10/32, 31%).

For 1 patient, no record was accessible, thus information on mortality, symptoms, and other clinical conditions was available for 20 of 31 neonates (65%) that survived and 11 of 31 (35%) deceased neonates. A summary of the patient data and virulence factors is given in Table 2; no association between gestational age or mortality and chorioamnionitis or urinary tract infection was seen (Table 2).

General Comments on Bacterial Analysis

Bacteria

A total of 37 isolates were cultured from 32 neonates; 29 of 32 (91%) patients had 1 isolate, and in 3 of 32 (9%) neonates more than 1 isolate was detected. One patient had 3 bacteremic episodes and 1 simultaneous infection of cerebrospinal fluid during the second episode. The first bacteremic episode was at day 2 postnatally, the second 10 days later, and the last 6 weeks postnatally. One further patient had 2 bacteremic episodes with 12 days in between and 1 neonate had simultaneous bloodstream and cerebrospinal fluid infections. The first cultured isolate per patient and corresponding parameters were included in the statistical analysis. High-quality reads were obtained for 37 isolates and species verification on sequence data confirmed the species identification as E. coli and the purity of the whole-genome extracts.

Phylogenetics

All phylogenetic lineages according to Clermont were represented, and the majority was found to belong to phylogroup B2 (16/32, 50%). The distribution among other lineages was D (7/32, 22%), A (4/32, 13%), F (3/32, 9%), C (1/32, 3%), and E (1/32, 3%). According to 7-gene MLST, 26 isolates could be assigned to ST-clonal complexes: CC ST69 (5/32, 16%), CC ST10 (4/32, 13%), CC ST95 (4/32, 13%), CC ST12 (2/32, 6%), CC ST59 (2/32, 6%), CC ST568 (2/32, 6%), with single isolates for CC ST14, CC ST23, CC ST28, CC ST31, CC ST38, CC ST73, and CC ST131. The remaining 6 isolates corresponded to sequence types ST62, ST141, ST127, ST567, ST526 and ST1583. For patients from whom multiple isolates were cultured, both cgMLST and rMLST analysis consistently revealed identical sequence types. Specifically, 1 patient's isolate found during 3 bacteremic episodes and 1 isolate from cerebrospinal fluid, were all designated as ST69. Likewise, the other isolates originating from a patient who had 2 blood stream infections, could be assigned to ST62. The patient with E. coli isolates from both the bloodstream and cerebrospinal fluid exhibited isolates assigned to ST59. Additionally, 2 patients shared identical sequence types according to cgMLST and rMLST analysis. However, in the 7-gene MLST, only 1 isolate was assigned to ST69 and the other to ST31. Notably, these individuals underwent intensive care simultaneously in the neonatal intensive care unit; therefore the isolates were assumed to be epidemiologically related (Figure 1).

Virulence Factors

Overall, virulence factors from all virulence factor groups were found: 29 of 67 (43%) fimbriae and adhesins, 13 of 13 (100%) iron metabolism, exotoxins 9 of 19 (47%), immunomodulation 15 of 15 (100%), bacteriocins 14 of 29 (48%), and capsule types 2 of 2 (100%). Identical isolates according to cgMLST shared the same virulence factor pattern.

Associations Between Bacterial Analysis and Patient Data

Phylogenetics, Patient Data, and Virulence Factors

Neonates infected with isolates belonging to phylogenetic lineage B2 according to Clermont (15/31, 48%) were more likely to die compared with neonates infected with isolates from other lineages (B2 10/15, 67% vs non-B2 1/16, 16%; OR, 26; P < .001). Extremely premature neonates delivered before 28 gestational weeks were more often infected with E. coli belonging to lineage B2 compared with neonates delivered after 28 gestational weeks (11/14, 78% vs 5/18, 28%; OR, 9; P = .01). Similarly, neonates with isolates belonging to lineage B2 were more likely to present with shock compared with those with impaired or good general condition (B2 6/15, 40% vs non-B2 1/16, 6%; OR, 9; P = .03). Also, E. coli that belong to phylogroup B2 were more frequently found in female patients compared with male patients (female 8/10, 80% vs male 8/22, 36%; OR, 7; P < .05). The mortality of infection with phylogroup B2 isolates was comparable between female and male patients (female 5/7, 71%; male 5/8, 63%). No associations between phylogenetic lineage and necrotizing enterocolitis, chorioamnionitis, urinary tract infection, or maternal infection perinatally were seen. Presence of determinants encoding fimbriae Sfm, Ycb, and Yra were statistically significantly associated with isolates belonging to non-B2 lineages. In contrast, determinants involved in iron metabolism (salmochelin, Sit, Hma, Fec, ChuA), exotoxins (colibactin, CNF, Usp), immunmodulation (TcpC, AslA, and OmpT), and bacteriocins (microcin H47 and M) were statistically significantly associated with phylogroup B2 (Table 3, Figure 1, and Figure 2).

Virulence Factors and Patient Data

Most statistically significant associations between virulence factors and patient data were found for mortality (14/133, 11%) and gestational age (8/133, 6%). Very premature neonates delivered before gestational week 28 had a high mortality; however, prematurity was also associated with 7 virulence factors (cnf, yde, clb, hma, sfm, aslA, and iro). The overall function of the 14 virulence factors that were associated with mortality could be assigned to exotoxins (4/14, 29%), immunomodulation (4/14, 29%), bacterial iron metabolism (2/14, 14%), fimbriae (2/14, 14%), and bacteriocins (2/14, 14%). While most virulence factors were positively associated with stillbirth and extremely premature delivery, the γ₁-fimbriae encoded by sfm and the self-recognizing antigen 43, encoded by ag43 were associated with survival and delivery after gestational age 28, respectively. Of the 8 virulence factors that were associated with gestational age, most common were fimbriae (4/8, 50%), exotoxins (2/8, 25%), immunomodulation (1/8, 12.5%), and bacterial iron metabolism (1/8, 12.5%). Above all, colibactin showed strong associations with mortality, gestational age, sex, and necrotizing enterocolitis. Patients with E. coli isolates carrying genes encoding the exotoxins colibactin, clb, were more likely to develop shock symptoms (shock 6/7, 86% vs nonshock 4/24, 17%) and to have necrotizing enterocolitis (6/10, 60% vs 4/21, 19%; OR, 6; P = .04). There were no strong associations of virulence factors with necrotizing enterocolitis, chorioamnionitis, urinary tract infection, or maternal infection perinatally (Table 4).

DISCUSSION

The present study examined 37 neonatal E. coli isolates, obtained from infants admitted to the neonatal intensive care unit, Uppsala University Hospital across a 15-year period, for ExPEC virulence factors. Notably, E. coli isolates categorized under phylogroup B2 were associated with mortality, extreme prematurity, shock, and female sex. The identification of exotoxin determinants, particularly the colibactin determinant clb, exhibited associations with mortality, gestational age, and the occurrence of shock. Additionally, the study introduced a comprehensive database of ExPEC virulence factor determinants, capable of efficiently analyzing whole-genome sequenced E. coli isolates.

The E. coli population is categorized into at least 8 phylogroups (A, B1, B2, C, D, E, F, and G), with phylogroup B2 notably linked to extraintestinal infections [11]. This association is believed to originate from colonization advantages of the human intestinal tract, enabling the acquisition of virulence traits and antimicrobial resistance determinants [12]. E. coli blood stream isolates associated with fetal and neonatal infections are often found to belong to phylogroup B2, which correlate with increased morbidity and mortality rates compared with other phylogroups [13, 14]. Half of the isolates belonged to phylogroup B2, and were associated with higher mortality compared with non-B2 and presented more often with shock symptoms. Phylogroup B2 was also found more often amongst extremely premature neonates, which have a higher mortality irrespective of infection status. However, a noteworthy observation is that while only 36% of the female neonates were extremely premature, 80% of all female infants had bacteremia linked to phylogroup B2, which resulted in higher mortality rates compared with male infants. This is in line with previous observations of heightened virulence potential of the B2 phylogroup and indicates sex-related differences in susceptibility to infections.

Sex-related differences in morbidity and mortality in early life have been documented since the 1930s, but evidence concerning sex-dependent distribution of E. coli populations is still limited [15, 16]. After perinatal and respiratory or nervous system conditions, infections are the fourth leading cause of infant mortality and exhibit a notable skew towards higher mortality in male patients [17]. While cultural factors can contribute to mitigating the male disadvantages in infant morbidity and mortality, an increasing body of evidence highlights biological sex differences in immune response to infections [18]. The previously described male vulnerability is reflected in the findings of this study, because bacteremia caused by E. coli was more common in male neonates born before or after 28 gestational weeks. Female infants had both a significantly higher likelihood of being infected with E. coli isolates associated with phylogroup B2, and had an increased risk of infection with isolates expressing capsule types K1 and K5 and genes encoding colibactin production, an association that might also be explained by phylogenetic linkages [13]. Possible biological explanations have been found in a genome-wide study involving 233 cases with late-onset neonatal sepsis from 6 countries that linked single-nucleotide polymorphisms in interleukin 10 (IL-10), tumor necrosis factor-α (TNF-α), and the NOTCH signaling pathway through sex stratified analysis [19]. Moreover, placentas from male fetuses exhibited more chronic inflammation, potentially offering female fetuses greater survival chances due to cardiovascular stability and lower levels of cytokines [20, 21]. Hormonal influences on the immune system have been documented, with testosterone demonstrating an essentially immunosuppressive effect and estrogens exhibiting an immune amplifying effect [22].

Blyton et al investigated the distribution of fecal E. coli phylogeny in relation to sex and found no significant influence of sex on the phylogenetic lineage of colonizing E. coli [23]. However, the same study did note that female patients, who predominantly harbored E. coli B2, carried fewer strains of E. coli A and B1, suggesting a potential advantage for E. coli B2 colonization in female patients. Consideration of sex-related differences in immune response to infections prompts speculation on whether the higher prevalence of E. coli B2 in female neonates found in the present study reflects sex differences in establishing bacteremia or instead indicates differences in symptom manifestations of bacteremia. The latter might introduce a selection bias, potentially resulting in a reduced frequency of blood culturing in female patients.

Unlike non-B2 E. coli, group B2 E. coli demonstrate significant adaptation to the human intestine. Many virulence factors prevalent in this phylogroup confer a colonization advantage, contributing to prolonged persistence [24]. Colibactin, linked genetically to phylogroup B2 through the pathogenic island, induces DNA damage by alkylating DNA, resulting in interstrand cross-links that promote mutagenesis and cell death [25, 26]. Infants acquire colibactin-carrying E. coli isolates during delivery from their mother, aiding in persistent colonization in the neonatal gut [24, 27, 28]. Similarly, factors associated with bacterial iron acquisition enhance both bacterial virulence and colonization properties, predominantly linked to group B2 [29].

In the context of evolution where bacterial populations strive for persistence, the properties enhancing colonization fitness and transmission to new hosts seemingly conflict with virulence traits that impede transmission by killing the host. Le Gall et al have suggested, based on phylogenetic evidence, that virulence in the B2 group might be a fortuitous side effect of successful commensalism [30]. Extraintestinal infections primarily involve an array of virulence factors; however, no single factor deems E. coli an obligate pathogen in extraintestinal infections. Notably, several virulence factors correlated with mortality in our study. Intriguingly, our findings indicate a strong association between colibactin-positive E. coli bacteremia and severe outcomes, including shock and mortality in both female and male patients. Conversely, the absence of colibactin was associated with survival. This prompts speculation about whether E. coli B2 harboring colibactin in the bloodstream might represent an obligate pathogen.

The sample size of this study is a clear limitation; however, we believe that the detailed examination of clinical features yielded some new insights, particularly concerning sex differences, mortality, and symptoms. Several variables in the study are interconnected; for instance, very premature age is linked to mortality from conditions other than infection, warranting a multifactorial analysis to unravel these associations. While the retrospective retrieval of patient data was incomplete for some individuals, prospective studies on infrequent infections demand an extended time horizon and poses logistical challenges. Sequence analysis is limited to showing the presence of genes and provides no information about the expression and effects of virulence factors. Nonetheless, the study underscores the diverse nature of neonatal E. coli septicemia: comprehensive analysis that integrates whole-genome sequencing of bacteria with corresponding extended patient profiles is imperative to unveil associations. Additionally, the study offers a comprehensive repository of extraintestinal virulence factors, potentially valuable for further investigations.

This study highlights the correlation between neonatal E. coli bacteremia caused by phylogroup B2, as well as isolates harboring certain virulence factors associated with adverse outcomes such as mortality, shock, and extreme prematurity. Notably, female patients were predominantly affected by group B2 infections, whereas male patients had bacteremia caused by both group B2 and non-B2 E. coli. Furthermore, the severity of E. coli bacteremia was notably higher in female patients, correlating with increased mortality rates. It is intriguing to note that despite recent recognition of significant sex-based differences in immune response and infectious diseases, these factors are not yet considered in clinical practice.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Author contributions. A. H. and S. S. contributed conceptualization and methodology. A. H., Å. M., E. N., E. T., and S. S. performed investigation, analysis, and project administration. A. H., Å. M., E. N., K. G., E. T., and S. S. wrote the article.

Data availability. All sequence data are publicly available through the European Nucleotide Archive (www.ebi.ac.uk/ena) under accession number PRJEB62832.

Financial support. This work was supported by the Centre for Research and Development, Region Gävleborg, Sweden (grant numbers CFUG-698631 to A. H. and CFUG-978549 to K. G.); the Region Uppsala, Sweden (grant number 810901 to S. S.); and the Swedish Research Council (grant number 2019-05909 to S. S.). Generation of sequence data was supported by Afa Insurance, Sweden (grant number 150411) and the Alf de Ruvo Memorial Foundation, Sweden. The National Genomics Infrastructure in Stockholm (funded by Science for Life Laboratory, the Knut and Alice Wallenberg Foundation, and the Swedish Research Council), and Swedish National Infrastructure for Computing/Uppsala Multidisciplinary Center for Advanced Computational Science supported assistance with massively parallel sequencing and access to the UPPMAX computational infrastructure.

References

Author notes