The strains in this study were from the largest reported outbreak of E. sakazakii in a NICU, with the most deaths, that has been reported to-date . In addition, it is the first report comparing in vitro virulence assays of clinical isolates with associated patient details and clinical symptoms.
E. sakazakii is associated with NEC, bacteraemia, sepsis and meningitis. Therefore, attachment to intestinal cells and translocation to systemic tissue are important in the pathogenesis of the organism. Variations in symptoms for each pulsotype could have varied according to dose, neonate age and antibiotic administration. However, pulsotype 2 appeared to be more virulent than pulsotypes 1 and 3 based on the number of NEC, septicaemia, meningitis and fatal cases. In addition, strains 695 and 767 in pulsotype 2 acquired extended spectrum β-lactamase activity . It is accepted that other unidentified factors could have contributed to the number of neonatal cases and severity of infection, such as duration of exposure.
The attachment of E. sakazakii strains to Caco-2 cells after a 3-h exposure period was greater than the negative control (p ≤ 0.001) E. coli K12 (Fig. 1). Strain 696 (pulsotype 1) attached to a greater extent than the positive control S. Enteritidis (2.5 and 1.5% respectively). Strain 693 (pulsotype 3) was from an asymptomatic case and showed lower attachment values than the E. sakazakii strains which had caused NEC, and meningitis. The attachment of strain 716, from powdered infant formula, was significantly (p = 0.026) higher than 693 and 709 (p = 0.017) and similar to 695, 701 and 767. E. cloacae strain 766 showed the greatest attachment value (3.6%). There was no direct correlation between attachment and invasion rates of Caco-2 cells. For example, strains 696 and 766 had the highest attachment rates, yet 701 and 767 had the highest invasion rates (Fig. 1 &2). All E. sakazakii strains and E. cloacae showed similar potential for invasion. Strains 701 and 767, associated with fatal cases of NEC and meningitis, showed the highest invasion rates (0.18 and 0.23%). This suggests that the ability to invade intestinal cells may be a strong indication of potential virulence and poor outcome in the neonatal host.
Most E. sakazakii strains persisted in macrophages for 48 h. Strains 701 (pulsotype 2, fatal NEC case) and 716 (pulsotype 4 from powdered infant formula) multiplied significantly in the first 24 h. Persistence and replication within macrophages suggests that E. sakazakii possess virulence mechanisms to withstand the bactericidal activities within macrophages and evade the host immune response. Controls for macrophage persistence (E. coli K1), killing (E. coli K12), and replication (C. koseri SMT319) comparatively showed that E. sakazakii strains could persist within macrophages for at least 48 h and some strains demonstrated moderate (767 and 696) to high (701 and 716) levels of replication. In contrast to the E. sakazakii strains, there was either no uptake of E. cloacae (strain 766) by U937 macrophage cells, or it was rapidly killed following macrophage uptake.
E. sakazakii is associated with rare cases of bacterial meningitis. Therefore, the ability to invade cells comprising the blood-brain barrier was investigated using rat brain capillary endothelial cell line rBCEC4. E. coli K1 and C. freundii have been tested in both Human Brain Microvascular Endothelial Cells (HBMEC) and rBCEC4 cells and show similar levels of invasion, supporting the use of rBCEC4 cells as a model of BBB cell invasion . The E. sakazakii strains and E. cloacae 766 were able to invade rat brain cells at similar levels as E. coli K1 (Fig. 4). E. coli K1 is known to cause acute bacterial meningitis in neonates and further demonstrates the virulence potential of E. sakazakii for this host site. E. sakazakii is invasive for capillary endothelial cells and may gain access to the brain via crossing the normally impenetrable blood-brain barrier with specific virulence mechanisms .
The four pulsotypes have been previously described with respect to source, patient symptoms, and phenotypic traits . Pulsotype 4 strains were isolated from an unopened tin of powdered infant formula and were not associated with any neonatal infections. It is notable that pulsotype 4 strains showed no protease activity on skimmed milk agar plates or capsule production. Whether these phenotypes are related to virulence determinants is unknown, but are under current investigation.
The outcome of infection due to E. sakazakii will be dependent upon a number of factors such as birth weight, age, immune status and antibiotic therapy . In our study, it was not possible to directly correlate clinical symptoms and outcomes with in vitro studies. Nevertheless, we have shown the invasive potential of E. sakazakii with intestinal and blood-brain barrier cells. In addition, E. sakazakii was able to persist and even replicate for a period within macrophage cells. These traits appear to facilitate host immune evasion and dissemination. In addition, E. sakazakii can acquire antibiotic resistance during infection, as shown by the acquisition of extended spectrum β-lactamase activities by strains 695 and 767. These two strains were isolated from two neonates, one month apart (Table 1). There was variation within pulsotype 2 for attachment and invasion of intestinal and blood-brain barrier cells. Whether this reflected the variation in patient symptoms is uncertain. But it is notable that strain 767 had the highest level of endothelial cell invasion, greater than E. coli K1, and was isolated from a fatal meningitis case, the outcome of which could have been influenced detrimentally by the extended β-lactamase spectrum activity the strain had acquired due to its resistance to standard antibiotic therapy.
Strain 766 was initially identified as E. sakazakii and was isolated from a fatal case of septic shock during the period of the NICU E. sakazakii outbreak (Table 1). However, recent 16S rRNA gene sequencing revealed it was E. cloacae . This strain showed attachment, and invasion of Caco-2 endothelial cells, no invasion or no survival in macrophages, and low invasion of rat brain capillary endothelial cells. Therefore, it had the ability to attach and invade intestinal cells, but no propensity for macrophage survival or dissemination, nor an ability to invade the brain. It had however been able to enter to bloodstream and cause fatal septic shock.
The potential attachment and invasion of host tissue, dissemination and invasion of the blood-brain barrier have been investigated and demonstrated the potential of E. 11 sakazakii to infect exposed neonates. The source of the three pulsotypes could not definitely be attributed to contaminated reconstituted infant formula. Nevertheless the feeding practices of preparation for 24-h periods, and prolonged (>3 h) administration could have enabled bacterial growth in the reconstituted formula and increased the risk of infection.