Liu et al. [1] reported that worldwide the estimated mortality in children younger than 5 years in 2010 was 7,600,000. Neonates, in particular those with very low birth-weight, are of particular concern due to their weak immune system [2]. The risk of infection in neonates increases with low birth weight and additionally prolonged hospitalization [3, 4]. Mortality among neonates is attributed to infectious causes, preterm birth complications, intrapartum-related complications, sepsis and meningitis.

The most common cause of Gram-negative bacterial neonatal meningitis is E. coli K1. It has a mortality rate of 10–15 %, and neurological sequelae in 30–50 % of cases [5–8]. Other bacteria responsible for neonatal and infant morbidity and mortality include Group B streptococci (GBS), Enterobacter spp., Citrobacter koseri, Neisseria meningitidis, Serratia spp., and Cronobacter spp. [9, 10].

E. coli serotypes associated with neonatal meningitis are primarily O18:K1:H7, O1:K1, O7:K1, O83:K1 and the more recently reported O45:K1:H7 [11–13]. These are in the E. coli extraintestinal pathogenic (ExPEC) subgroup B2, and are sequence type (ST) 95 (Achtman scheme) [14]. Bacterial invasion across the blood–brain barrier is multifactorial, requiring several genes for binding, invasion and intracellular survival. Proposed virulence genetic determinants include ibeA, sfaS, cnf1, gimA, and ompA [15, 16]. However these are not always supported in experimental assays and are not demonstrable in all E. coli strains isolated from cerebral spinal fluid [12, 13, 17].

Although there have been some concerns regarding bacterial biofilm formation inside neonatal nasogastric feeding tubes by opportunistic pathogens, few systematic studies have been undertaken [18–20]. Mehall et al. [18] reported both feeding intolerance and a link to necrotizing enterocolitis in neonates following the bacterial colonised of their feeding tubes. Previous studies by Hurrell et al. [19] have revealed that in situ the inside of such tubes can be colonised by a variety of fungi and various opportunistic bacterial pathogens producing a biofilm of mixed microbial composition; Candida spp., E. coli, Enterobacter hormaechei, Klebsiella pneumoniae, Cronobacter sakazakii, Yersinia enterocolitica, and Pseudomonas fluorescens. The aim of this study was to investigate the diversity of the E. coli strains previously isolated by Hurrell et al. [19] from the residual liquid in the lumen and biofilm from 30 neonatal nasogastric feeding tubes, which had been collected from among 129 neonates on two intensive care units.

Microbial colonisation of the neonate starts at birth, or even sooner through the meconium [39]. This initial flora is largely commensal but may include E. coli pathovars [40]. Hurrell et al. [19] has already given a general overview of the Enterobacteriacae isolates from bacterial biofilms inside neonatal nasogastric feeding tubes collected from these two hospitals. Those studies included electron micrographs of multi-organism (bacteria and fungi) biofilms inside the used feeding tubes. E. coli was isolated from 29 % (n = 129) of these tubes. However, in that study, the E. coli strains had not been genotyped to determine if there was a common source.

Our follow-up study shows, as expected, that the same pulsotype strains were isolated from both the residual liquid of the lumen and inner surface biofilm from 66 % of tubes; Table 1. For example, PT1 were three strains from one neonate, which had been isolated on the same day. This neonate had been fed ready-to-feed formula. Given that this is a sterile product, the source of the E. coli strains inside the feeding tube is uncertain. This issue has already been considered by Hurrell et al. [19] who proposed that a possible secondary source of the enteral tube flora was the throat due to gastroesophageal reflux. In preterm neonates this occurs 3–5 times per hour when the lower oesophageal sphincter relaxes. This would increase the exposure of the feeding tube to the throat flora.

However, Fig. 1 also shows that the 30 E. coli isolates only formed five pulsotypes, and therefore multiple indistinguishable strains had been isolated from different neonates over a one year period. PT4 strains had been isolated from three neonates over a 4 week period. These neonates had all received breast milk, and two had also received reconstituted infant formula. This demonstrates the possible dissemination of strains in the neonatal intensive care unit. Of particular significance was pulsotype 3 which was composed of 19 indistinguishable E. coli strains from 11 neonates on different feeding regimes, over a two week period; Table 1. This reinforces the probability that strains were acquired due to dispersion within the intensive care unit by carers and the environment and not a specific feed source such as contaminated infant formula.

MLST revealed that each pulsotype corresponded with a unique sequence type. It is noted that although ST394 and ST2076 only differ in one nucleotide in the parA allele, that the strains differed in their antibiotic susceptibilities and virulence; Table 2. The E. coli ST127 strain 1008 was the only strain which was motile, and also showed β-haemolysis on sheep blood agar; Table 2.

Since the clinical representation of the neonates in the study was not available, the potential pathogenicity of the strains was assessed using both genetic analysis for virulence traits (PCR-probes, and genome sequence analysis) as well as in vitro tissue culture. E. coli K1 translocates from the neonatal intestines to the bloodstream, where they multiply and cross the blood–brain barrier by invading the brain microvascular endothelial cells. These steps were investigated using attachment and invasion studies of human colonic carcinoma epithelial cells (Caco-2), rat blood brain barrier cells (rBECE4) and human brain microvascular endothelial cells (HBMEC) tissue culture cells; Fig. 3. Macrophage survival was studied using the (U937) cell line of human monocyte cells. These assays revealed there was considerable variation in the presence of virulence traits and in vitro pathogenicity according to the E. coli sequence type.

The three E. coli K1 ST95 strains were notable for their ability to attach and invade intestinal and both human and rat brain cells at levels comparable to Salmonella enterica and C. koseri, respectively; Fig. 3a, c, d. Macrophage uptake and persistence was comparable to C. koseri; Fig. 3b. The three other sequence types (ST394, ST73, ST2076) also attached and invaded human intestinal and brain cells, and ST394 was also able to invade rat brain cells.

These five sequence types are in the ExPEC biogroup B2, and combining the results of the motility assay with serotyping showed that the ST95 strains were E. coli O1:K1:NM. This group is of high significance due to their strong association with neonatal meningitis, and on this occasion 19 indistinguishable strains had been isolated from the tubes of 11 neonates. E. coli phylogroup B2 ExPEC strains of serotypes O1, O2, O18, and O45 are most frequently in ST95, and have been a focus of considerable research in recent years [12–14, 41,42]. In order to assess the virulence potential of the strains, a total of 30 virulence genes were screened for. These included adhesins, invasins, capsule, toxins, siderophores and others commonly associated with neonatal meningitic E. coli (NMEC), avian pathogenic E. coli (APEC) and uropathogenic E. coli (UPEC) [28].

The presence of the virulence related genes differed depending on the sequence type; Table 5. For example, E. coli O1:K1:NM ST95 strains encoded adhesin genes fimH, papACEFGI, papG allele II, siderophores fyuA (yersiniabactin receptor), and UPEC pathogenicity associated island (PAI) marker (malX) as well as the serum resistance associated gene traT. However despite the attachment and invasion of human and rat brain cells (Fig. 3), the ST95 strain however did not encode for ibeA or sfaS. Similarly Johnson et al. [17] reported their occurrence in only 33 % and 59 % of NMEC strains (n = 70), respectively. In contrast, the β–haemolytic ST127 E. coli K5 strain 1008 which did not attach or invade any cell line encoded for the sfaS adhesin as well as the haemolysin encoded by hlyA and the cytotoxic necrotizing factor (cnf). MalX, a marker for a UPEC PAI from the archetypal ExPEC strain CFT073 (serotype O6:K2:H1) [43] was present in sequence types 73, 95, and 127. A fuller description of the E. coli genomes derived from this study will be given in a separate publication.

Mora et al. [38] reviewed the source and virulence profiles of 59 ExPEC O1:K1:H7/NM ST95 strains of animal and human origin, recovered from different dates and geographic sources. They reported that some APEC isolates may act as potential pathogens for humans from poultry, suggesting no host specificity for this type of isolate. In contrast, the strains in this study had been isolated from preterm neonates in isolation units who had been fed breast milk and infant formula. Microbiological analysis of the feeds and microbial carriage by staff was not assessed at the time by Hurrell et al. [19]. Therefore the source of these E. coli K1 strains is currently uncertain.

E. coli K1 are the second most common cause of severe neonatal infections after Group B streptococcal (GBS) meningitis [44]. Although 85 % of infected neonates recover, this is often not as full as would occur with older infants and children. Sources and dissemination of such pathogenic organisms needs further investigation, especially since E. coli K1 causes 80 % of neonatal meningitis cases. Neonates acquire their initial flora at birth from the mother, environment and other carers. Maternal to child transmission of E. coli has been reported, and has been linked to late-onset neonatal infection [45–47]. There is also the possible transmission of E. coli K1 by nurses’ hands [40]. In addition, recent microbiome studies have indicated the possible dispersion of bacteria in the neonatal intensive care units [48]. It should also be noted that E. coli ST73, ST394 and ST2076 strains demonstrated the ability to invade human cells lines and therefore their occurrence in nasogastric feeding tubes may additionally pose a pathogenic risk towards neonates.

Due to confidentiality reasons, the clinical condition of the specific neonates from whom the E. coli strains in this study were isolated is not available. However during this collection period there were four cases of E. coli infection. It should be noted that the genomic analysis showed the E. coli K1 strains (ST95) had two streptomycin resistance genes belonging to the aminoglycoside class antibiotics (Table 4). This could be of clinical significance since aminoglycoside antibiotics such as gentamicin are regularly used as 1^st and 2^nd line combinations on NICUs. Since the patients’ details and isolates were not available for analysis, no direct causal infection route from the nasogastric tube can be made. Attribution of the source of the E. coli K1 ST95 is not feasible as there no environmental sampling or screening of carriage by staff or mothers. Nevertheless given the indistinguishable strains were obtained from neonates on different feeding regimes it seems probable that strains were disseminated in the NICU by carers and the environment, and not directly from a single feeding source.