Cronobacter spp. – opportunistic food-borne pathogens. A review of their virulence and environmental-adaptive traits

Infantile or neonatal infections

These organisms are regarded as opportunistic pathogens linked to life-threatening infections predominantly in neonates (infants <4 weeks old) (Bar-Oz et al., 2001; Mullane et al., 2008). Clinical presentation of Cronobacter infections in infants include NEC, bacteraemia and meningitis, with case fatality rates ranging between 40 and 80 % being reported (Bowen & Braden 2006; Friedemann, 2007). Infections in older infants have also been noted (Bowen & Braden, 2006).

The Food and Agriculture Organization of the United Nations/World Health Organization statistics for 2006 estimated that the annual incidence rate in the USA among low-birth-weight infants who weighed <2500 g and were <1 year old was 8.7 per 100 000 (FAO/WHO, 2006). Globally, there is no active surveillance system for tracking this pathogen; however, in their 2008 report, the WHO Expert Panel tracked cases from 1961 to 2008, and found 120 recorded cases of Cronobacter among infants and children <3 years old (FAO/WHO, 2008). Although only ~120 cases have been reported worldwide, the actual number of cases is considered far higher (CDC, 2009; Friedemann, 2009; Teramoto et al., 2010). Stoll et al. (2004) reported that only one case of sepsis caused by Cronobacter was diagnosed among 10 660 neonates, indicating that outside the epidemic outbreaks, Cronobacter are very rare in low-birth-weight infants. Even though septicaemia was the diagnosis in this single case, the infant was breast fed, further emphasizing that other routes of nosocomial transmission may be present. Nonetheless, it is not the number of infections caused by Cronobacter that is the issue; it is rather the 80 % fatality rate associated with illness. In addition, it was reported that all patients recovering from central nervous system (CNS) infections develop chronic neurological and developmental disorders (Lai, 2001).

An early hypothesized source of Cronobacter infection was the birth canal. However, this source now seems to be unlikely, as none of the vaginally delivered infants developed signs of infection until several days after birth and it is also unlikely that Cronobacter constitute a part of the vaginal microflora (Yan et al., 2012; Hunter & Bean, 2013). In addition, Biering et al. (1989) observed that infants who showed Cronobacter meningitis had all received reconstituted powdered-milk formula before falling ill. Furthermore, after a review of feeding procedures, a significant association was found between the consumption of certain brands of PIF, the development of NEC in neonates and the isolation of Cronobacter (van Acker et al., 2001). Several Cronobacter outbreaks have occurred in neonatal intensive care units that were linked epidemiologically to contaminated PIF (Himelright et al., 2002; Teramoto et al., 2010). In a recent survey, the majority of the infant formula packages were found completely devoid of any micro-organisms (Jaradat et al., 2009). However, extrinsic contamination of opened PIF cans or formula preparation utensils has also been reported (Noriega et al., 1990; Friedemann, 2009; Jaradat et al., 2009). Table 1 shows a list of the various foods and environments where Cronobacter have been isolated. Accordingly, the risk of infection may rely on different factors, including the number of bacterial cells contaminating the product, the source of Cronobacter, handling after preparation and the health status of the infants (e.g. low birth weight, prematurity, or immunosuppression) (Himelright et al., 2002).

NEC

It is generally thought that Cronobacter gain entrance to the human body through the gastrointestinal tract (GIT) where they may cause NEC (Liu et al., 2012b). Grishin et al. (2013) point out that the development of NEC requires a susceptible host, typically a premature infant with physiological impairment (i.e. hypoxia, hypothermia, intestinal ischaemia), administration of enteral formula feeds (which lack the beneficial protective components normally found in breast milk) and uncontrolled bacterial colonization. These conditions lead to an increased degree of mucosal inflammation, which subsequently results in the production of high levels of host inflammatory factors, including cytokines, nitric oxide, platelet-activating factor and prostanoids, which further damage the apical GIT epithelium. These authors and others also suggest that the virulence of Cronobacter spp. is dose-dependent, but not due to a property of a particular bacterial species as a whole, rather a characteristic(s) of certain strains, which may be innocuous in a full-term, yet pathogenic in a pre-term infant (Hamby et al., 2011; Cetinkaya et al., 2013). In vitro models of infection also suggest that the organism may also gain entrance into the systemic circulation through transcytosis of the GIT epithelium (Giri et al., 2012).

CNS infections

Once the organism has entered the systemic circulation, it has a tropism toward the CNS, thus increasing the propensity to cause meningitis among low-birth-weight neonates and infants, whilst causing bacteraemia or sepsis among slightly higher-birth-weight infants (Yan et al., 2012) or adults. Once the pathogen crosses the BBB and enters the brain, it might cause ventriculitis and form cysts or brain abscesses, which later may develop into hydrocephalus – a condition where excessive accumulation of cerebrospinal fluid (CSF) in the brain occurs (Bowen & Braden, 2006; Chenu & Cox, 2009). The disproportionate accumulation of CSF results in an abnormal enlargement of the spaces in the brain called ventricles. The enlarged CSF-filled ventricles create a situation where the balance between CSF production and absorption is disturbed, which potentially leads to an increase of cranial pressure on the tissues of the brain.

Infections associated with adults

Compared with infant cases, Cronobacter are more commonly detected in adults, with most case reports describing infections related to elderly patients who were previously ill or immunocompromised. Up to 50 % of adults with Cronobacter infection had an underlying malignancy (Lai, 2001; See et al., 2007). Interestingly, Cronobacter have been linked to or should be considered as a pathogen that may cause sepsis, conjunctivitis, aspiration pneumonia in stroke patients, osteomyelitis, diarrhoea, acute cholecystitis, wound infections, abscesses associated with indwelling catheters and urinary tract infections (Gosney et al., 2006; Friedemann, 2009; Flores et al., 2011; Yan et al., 2012; Tsai et al., 2013).

Nosocomial infections by Cronobacter, such as aspiration pneumonia, urinary tract infections and conjunctivitis, are unlikely to be related to contaminated PIF, so medical equipment surfaces, water outlets and intrapersonal contacts may serve as possible sources of these infections (Friedemann, 2009; Flores et al., 2011). However, other environmental sources of contamination are also possible. Jaradat et al. (2009) isolated Cronobacter from household vacuum dust in addition to isolating it from spices, indicating the strong association of the pathogen with a variety of dried food products and environmental factors. See et al. (2007) described a case of Cronobacter bacteraemia with multiple splenic abscesses in a 75-year-old institutionalized woman. It was the first reported case of Cronobacter with splenic abscesses and was the first case in a non-immunocompromised adult.

Multilocus sequence typing (MLST) sequence types (STs)

The reclassification scheme described by Iversen et al. (2007, 2008a) was supported subsequently by both optical mapping and genome sequencing data, which confirmed the revision of the taxonomic positions of the seven species (Yan et al., 2012; Kotewicz & Tall, 2009; Kucerova et al., 2010; Stephan et al., 2011; Joseph et al., 2012; Grim et al., 2013). Although C. sakazakii and C. malonaticus were found to be closely related and difficult to distinguish by 16S rDNA gene sequence analysis, a MLST scheme reported by Baldwin et al. (2009) was shown to support the separation of these two species (Yan et al., 2012). Other supporting evidence which suggested that C. malonaticus and C. sakazakii are separate species were the single nucleotide polymorphisms identified for two genes, rpoB and cgcA, as reported by Stoop et al. (2009) and Carter et al. (2013). Their work led to the development of two species-specific end-point PCR assay schemes that are now widely used for speciation of isolates. The sequencing of another 15 Cronobacter strains by Grim et al. (2013) fully corroborates both the taxonomic scheme described by Iversen et al. (2008a) and the MLST scheme described by Baldwin et al. (2009). Grim et al. (2013) concluded that, in particular, C. sakazakii and C. malonaticus must have evolved or acquired accessory genes that have enhanced their virulence capacity and host species adaptation, and thus augmented their overall pathogenicity. Furthermore, the MLST scheme has defined over 136 STs and C. sakazakii ST4 was the predominant ST found associated with CSF isolates from cases of neonatal meningitis (Joseph et al., 2012a). In fact, the MLST typing method has alone defined 55 STs for C. sakazakii (Sonbol et al., 2013). As suggested by Holý & Forsythe (2013), the clonality and prevalence of the ST4 strains are clearly associated with meningitis, but not NEC. Fortunately, this information gives direction for further meningitis research with the bacterium.

Nevertheless, C. sakazakii has been implicated in most of the infant and neonatal meningitis cases. Similarly, in another study by Hariri et al. (2013), the authors found that most serious meningitis cases were caused by C. sakazakii ST4. Furthermore, they also found that all the five CSF isolates were ST4. In another study, Joseph & Forsythe (2011) showed that half of the studied Cronobacter strains from all clinical sources over 50 years from seven countries were C. sakazakii ST4. These findings indicated the predominance of this particular clonal or clonal complex in causing neonatal meningitis and provided evidence that ST4 may be a stable virulent clone. Furthermore, it was suggested that C. sakazakii ST4 may be composed of multiple STs and comprises a clonal complex, and ST4 was predominantly involved in infant meningitis and was frequently isolated from milk powder processing facilities. C. malonaticus contained other STs such as ST8 that were associated with clinical cases, but not necessarily associated with neonates, whilst ST7 was associated with adult infections (Sonbol et al., 2013). Sonbol et al. (2013) described the occurrence of different Cronobacter MLST STs in global surveys and their findings suggest that 24 % of strains from PIF manufacturing sites are the ST4 clone. In their study, Müller et al. (2013) reported that 132 of 141 Cronobacter isolates obtained from an infant formula production factory were C. sakazakii, whilst only seven were C. malonaticus. Furthermore, when the isolates were analysed by MLST, C. sakazakii ST4 was the predominant type among them.

Molecular O-antigen typing

The O-antigen is a component of the LPS structure located on the outer surface of Gram-negative bacteria and is responsible for serological diversity. As a Gram-negative pathogen, the cell surface of Cronobacter is covered with a dense area of LPS (endotoxin) with a lipid A core, the toxic moiety forming ~8 % of the total LPS weight (MacLean et al., 2009). It was found that different strains of Cronobacter have different LPS structures, and show differences in pathogenicity based on differences in structure (linear versus branched) and composition of LPS (MacLean et al., 2009). It was found that the LPS of C. sakazakii and C. malonaticus was branched, whilst LPS from other Cronobacter strains such as C. muytjensii was unbranched (Joseph et al., 2012c). Indeed, C. muytjensii LPS failed to provoke the production of mAbs due to its unbranched structure (Jaradat et al., 2011).

As a useful method of characterizing Gram-negative bacteria, O-antigen typing and studies describing the nature of the O-antigen associated with Cronobacter spp. have been reported (Mullane et al., 2008; Jarvis et al., 2011, 2013). Mullane et al. (2008) initially developed a molecular serotyping method, based on amplification of the rfb encoding locus (in Gram-negative enteric bacteria this is located between galF and gnd) followed by MboII digestion. Using this approach, a PCR-RFLP profile was generated that could be compared across several isolates. Based on this approach, the first two O-antigen serotypes were characterized, and denoted as O1 and O2. More recently, Sun et al. (2011) added five additional C. sakazakii O-antigen serotypes that correlated with the PCR-RFLP profiles. Jarvis et al. (2011) extended the original Mullane molecular characterization scheme to include other Cronobacter spp. and defined eight new molecular O-serotype gene clusters. Two of these O-serotype gene clusters were shared among C. sakazakii, C. muytjensii, C. malonaticus and C. turicensis strains (Jarvis et al., 2011, 2013). The structural composition of several O-serotypes has now been described (MacLean et al., 2009; Arbatsky et al., 2010, 2011; Shashkov et al., 2011). Presently, there are 17 recognized serotypes among the seven species. Over a 1 year period (2011–2012), Müller et al. (2013) found that C. sakazakii serotype O2 was the predominant serotype observed among 141 confirmed Cronobacter isolates obtained during a surveillance study of a Swiss PIF manufacturing facility. Correlations between ST4 and C. sakazakii serotypes O2/O3 and between ST83 and C. sakazakii serotype O7 were observed, suggesting that a ST may contain multiple serotypes.

Toxins and proteolytic enzymes

Historically, enterotoxin production has been a major virulence factor that marks the pathogenicity of microbes. Pagotto et al. (2003) were the first to describe putative enterotoxin activity in Cronobacter. Using the suckling mice assay, they were able to show an enterotoxin effect by four of the 18 tested Cronobacter isolates. When tested on cultured mammalian cells, filtrates from one strain were toxic to Vero and Y-1 cells, causing rounding and cell lysis. Boiling the filtrates for 20 min did not affect the toxin activity in suckling mice, but it reduced the cytopathic effect of the toxin on Vero cells. Yang et al. (2009) showed that filtrates from two out of eight strains studied were toxic to Vero cells. Raghav & Aggarwal (2007) were the first to purify and characterize a putative enterotoxin from Cronobacter isolates. The molecular mass of the toxin was determined to be 66 kDa, which is similar to the 62 kDa toxin of Shigella dysenteriae 1 and to the 65 kDa toxin of Pseudomonas aeruginosa with optimum activity observed at pH 6. Additionally, the activity of the toxin was unaffected by holding it at 50 or 70 °C and it was stable at 90 °C for 30 min, implying that it is a moderately heat-stable toxin. This indicates that the enterotoxin could be resistant to commercial milk pasteurization treatments and thus stays active in PIF.

However, it seems that the discovery of these toxins is still in its infancy, as the specific genes need to be identified to shed light on the importance of these enterotoxins (Chenu & Cox, 2009). Molecular studies using genome sequencing will help with the further characterization of these toxins.

In addition to enterotoxins, Cronobacter also secrete proteolytic enzymes that lyse cells and create tissue damage at the site of infection in mice (Pagotto et al., 2003). Kothary et al. (2007) reported the presence of a zinc metalloprotease that deforms CHO cells, eventually causing their rounding and leading to cell damage. The genes for a haemolysin (hly) have been identified, but no activity has been found (Kucerova et al., 2010; Cruz et al., 2011; Grim et al., 2013).

Townsend et al. (2007a) provided evidence that the presence of endotoxin along with C. sakazakii in infant formula enhanced the translocation of C. sakazakii from rat guts through the BBB. One possible mechanism is thought to be the disruption of the tight junctions by LPS, thus increasing the permeability of host barriers to the pathogen (Kim & Loessner, 2008). In addition, endotoxin was found to impair enterocyte migration and epithelial restitution, and therefore inhibited tissue repair following bacterial infection, which leads to the translocation of the pathogen from the gut into the blood and internal tissue (Cetin et al., 2004). As the endotoxin is heat stable at 100 °C, its presence in formulas might also play an important role in enhancing the pathogenesis of Cronobacter in human infants (Townsend et al., 2007a).

Attachment and invasion

Role of OmpA in Cronobacter invasion of mammalian cells

OmpA appears to play a major role in Cronobacter invasion. In vitro, it was found that OmpA plays a critical role in the invasion of human intestinal epithelial cells (INT-407), Caco-2 cells and HBMECs (Nair & Venkitanarayanan, 2007; Kim & Loessner, 2008; Singamsetty et al., 2008; Kim et al., 2010). An 87 % reduction of Cronobacter invasion of INT-407 cells was observed in the absence of OmpA expression. Similarly, Nair et al. (2009) reported that OmpA is a determinant that influences the invasion of C. sakazakii into HBMECs, by providing evidence that the absence of OmpA expression resulted in 83 % reduction in C. sakazakii invasion. However, adherence of an OmpA mutant did not differ significantly from the WT strain, possibly because there are other surface determinants that contribute to C. sakazakii attachment to host cells that may not play a role in the invasion.

Mittal et al. (2009a) showed that OmpA expression is important for the onset of meningitis caused by Cronobacter in a newborn rat model as OmpA⁺ Cronobacter were present in high numbers in the brains of infected animals. They also showed that OmpA expression is essential for Cronobacter resistance to blood and serum killing (Mittal et al., 2009a). Although both OmpA⁺ and OmpA⁻ Cronobacter were efficiently taken up by dendritic cells, only OmpA⁺ Cronobacter resisted killing and multiplied in these cells, thus highlighting the critical role of OmpA in protecting the intracellular survival of this pathogen (Mittal et al., 2009b).

Similarly, Kim et al. (2010) reported that OmpA and OmpX of C. sakazakii are important for both apical and basolateral adhesion to and invasion of host cells, as well as for movement into deeper organs, such as the spleen and liver. Invasion studies proved that OmpA and OmpX are crucial for C. sakazakii invasion of INT-407 and Caco-2 cells. However, adherence studies demonstrated that OmpA was essential for binding to Caco-2 cells, but not INT-407 cells. OmpX did not show any significant role in binding to both types of cells examined. The authors concluded that C. sakazakii adherence to Caco-2 cells is mediated by OmpA, but adhesion to INT-407 cells is not, and that the initial interaction of C. sakazakii with Caco-2 and INT-407 cells is independent of OmpX. Nonetheless, not all Cronobacter spp. were positive for OmpA using the primers described by Nair & Venkitanarayanan (2007), as reported by Jaradat et al. (2009) and Giri et al. (2012).

Role of cytoskeleton rearrangement in Cronobacter invasion

To enable smooth entrance into the cytoplasm of a host cell, most pathogenic bacteria are able to manipulate the cytoskeleton of the host cell either by the secretion of effector molecules or by interfering with the phosphorylation cascades in the intracellular signal transduction pathways (Singamsetty et al., 2008; Kim & Loessner, 2008; Mittal et al., 2009b).

Results of several studies clearly demonstrated that Cronobacter exploit both microfilaments and microtubules to invade host cells. Actin microfilaments are the most well-known structures in this process. This was supported by the fact that treating INT-407 cells, Caco-2 cells and HBMECs with the actin-depolymerizing agent cytochalasin D completely blocked the invasion by C. sakazakii of these cell lines (Nair & Venkitanarayanan, 2007; Kim & Loessner, 2008; Li et al., 2010). C. sakazakii invasion of HBMECs involves participation of distinct signal transduction processes in which the PI3K/Akt signalling pathway also plays an essential role to initiate the rearrangement of actin filaments and subsequent bacterial internalization (Singamsetty et al., 2008; Li et al., 2010). Liu et al. (2012a) found that host cytosolic phospholipase A₂α is a downstream effector of PI3K/Akt signalling, which is needed for the actin reorganization in HBMECs elicited by C. sakazakii infection. In contrast, it was shown that Cronobacter severely impair the phosphorylation of p38, extracellular signal-regulated kinase 1/2 and c-Jun N-terminal kinase, which are major components of the MAPK pathway, and prevent the activation and maturation of dendritic cells by disarming the MAPK pathway – a process that may aid the pathogen in evading the immune system (Mittal et al., 2009b).

Tight junction disruption

The tight junction constitutes the principal barrier in intestinal epithelial cells. Kim & Loessner (2008) showed that when the tight junctions of Caco-2 cells were disrupted, the efficiency of C. sakazakii adhesion and invasion increased substantially. Apparently, after binding of Cronobacter to the intestinal enterocyte cell surfaces, Cronobacter caused the generation of increased amounts of nitric oxide, which is thought to have led to the disruption of the tight junctions between individual enterocytes. Disruption of the tight junctions subsequently led to apoptosis (Liu et al., 2012b). This, in turn, led to the leakage of the pathogen through these localized disruptions, eventually reaching the systemic circulation (Liu et al., 2012b).

Interestingly, the inhibition of the PKC-α intracellular signalling pathway abolished the disruption of the tight junctions. These results indicate that PKC-α influences the permeability of the intestinal epithelium to C. sakazakii (Liu et al., 2012b). Singamsetty et al. (2008) showed that PKC-α activation is also critical for Cronobacter invasion of HBMECs. The disruption of tight junctions might be a result of C. sakazakii interaction with dendritic cells, which could modulate the expression of the ‘tightness’ of the tight junctions (Emami et al., 2011).

Virulence genes and plasmids

Several virulence genes and plasmids were identified, and found to be specific to Cronobacter spp. Kothary et al. (2007) identified the gene locus of zinc-containing metalloprotease (zpx) – a protein that caused rounding of CHO cells in tissue culture. In addition, the presence of putative sodA genes might provide resistance for Cronobacter against intracellular macrophage oxidase and acidic conditions, and may contribute to its intracellular persistence (Townsend et al., 2007b). However, these genes and OmpA and OmpX were found in all Cronobacter spp., and thus cannot be responsible for the variation in pathogenicity among the different Cronobacter strains (Joseph et al., 2012c).

The complete genome sequence of C. sakazakii ATCC BAA-894 enabled Cruz et al. (2011) to compare other Cronobacter genome sequences with the sequence of C. sakazakii in their attempt to identify virulence genes. Three putative virulence genes – type III haemolysin (hly), siderophore-interacting protein (sip) and plasminogen activator (cpa) – were identified.

Among the annotated genes inspected, Kucerova et al. (2010) also showed that the complete cation efflux system (cusA, cusB, cusC and cusF) and its regulatory gene cusR were present in strains associated with neonatal infections, but missing in the other strains, which may explain the fact that not all Cronobacter spp. are equally pathogenic and cause meningitis.

In addition, Franco et al. (2011a) showed that all Cronobacter RepFIB plasmids encode two iron acquisition systems – a siderophore-mediated iron acquisition system (iucABCD/iutA operon) and ATP-binding cassette transport-mediated iron uptake and siderophore system (eitCBAD operon) – suggesting that these plasmids are common important virulence plasmids and may also contribute to systematic survival of Cronobacter. The ability to acquire iron is crucial for most pathogens in establishing infection after entering a host cell. Franco et al. (2011b) showed that the iucABCD/iutA siderophore (cronobactin) is the only functional siderophore possessed by Cronobacter. However, Grim et al. (2013) demonstrated that cronobactin – a hydroxamate-type, aerobactin-like siderophore – was not the only iron acquisition system possessed among Cronobacter. The cronobactin locus consists of five genes homologous to biosynthetic genes iucABCD and the receptor gene iutA. Indeed, iron acquisition is considered as growth promotion as it is related to utilization of iron in milk, but as iron utilization genes were found in all Cronobacter spp. its role is still not understood (Joseph et al., 2012c). Grim et al. (2013) reported targeted in silico sequence analysis of nine Cronobacter genomes, and showed that Cronobacter shared iron acquisition systems among the seven species and also possessed iron acquisition genes encoding ferric and ferrous transporters and haem-iron extractors, as well as putative TonB-dependent iron receptors and ferric reductases. For acquisition of ferrous iron, all Cronobacter have both the Feo and Efe systems, and for transport of ferric iron, all plasmid-harbouring strains (97 %) have the aerobactin-like siderophore cronobactin. All Cronobacter have the genes encoding the enterobactin-like siderophore, but this siderophore is thought not to be functional. Grim et al. (2013) also reported that in addition to receptors for cronobactin and enterobactin, all Cronobacter have five common receptors (FhuA, YncD, FoxA, FhuE and PfeA) for siderophores that are also produced by other organisms. The ferric dicitrate transport system was found specifically in a small subset of C. sakazakii and C. malonaticus strains, most of which were isolated from clinical samples, suggesting that this iron acquisition system plays a role in the virulence of Cronobacter. These studies also provided evidence that C. dublinensis and C. muytjensii have two other receptors, Fct and FcuA, for heterologous siderophores produced by plant pathogens, indicating that these receptors may give an advantage to these Cronobacter spp. to compete more successfully for iron in a plant niche. This also supports the hypothesis that Cronobacter emerged from a common ancestor that possessed a plant-associated lifestyle prior to its species-level bi-directional divergence.

Secretion systems, referred to as type IV secretion systems, that can transfer proteins and nucleoprotein complexes are considered to be important virulence factors as they were identified in C. sakazakii and C. turicensis as a plasmid-borne (pESA2/pCTU2) gene cluster (Franco et al., 2011b). In contrast, several newly identified type VI secretion systems (T6SSs) were found in the chromosomes of most Cronobacter spp. as well as on pESA3 carried by C. sakazakii, including pathogenic C. sakazakii ST4 strains (Joseph et al., 2012c). Interestingly, the T6SS was also found to be important for E. coli K1 invasion of the BBB (Joseph et al., 2012c).

Other virulence characters

Whole-genome studies of Cronobacter revealed that C. sakazakii is the only Cronobacter species that has the nanAKT gene cluster encoding utilization of exogenous sialic acid as a carbon source (Joseph et al., 2013b). Other genome-sequencing studies revealed the presence of two genomic regions (GR127 and GR129) that are involved in the utilization of sialic acid in the genome of C. sakazakii BAA-894 (Grim et al., 2013). Furthermore, Grim et al. (2013) found that 55 out of 57 strains of C. sakazakii were able to utilize sialic acid or its derivative N-acetyl-neuraminic acid, whilst no other Cronobacter strains were able to do this. Sialic acid is found in breast milk, infant formula, the mucin lining of the intestinal tract and is a component of the brain ganglioside complex (Siqueira Santos et al., 2013). Hence, it is possible that the ability of C. sakazakii to utilize sialic acid will enhance its pathogenicity for neonates and young infants. Cronobacter might use sialic acid to produce its exopolysaccharide (capsule) which helps it to avoid the immune system, and this behaviour is probably exacerbated in the presence of different milk sources and their products, which also contain sialic acid or its precursors. In addition, C. sakazakii sialic acid may also help in colonizing the intestinal tract via interactions with host mucins (Joseph et al., 2012c).

Inositol fermentation has been proposed recently as a marker of pathogenicity for Cronobacter based on the presence of the inositol monophosphatase gene (suhB) in some pathogenic strains. However, the inositol utilization operon GR29 was found in Cronobacter strains isolated from the environment, whilst also being absent in genomes of pathogenic strains (Grim et al., 2013). Therefore, its role in virulence is unclear at present.

Biofilm formation

Cronobacter have been reported to attach to and form biofilms on stainless steel, glass, latex, silicon, polyvinyl chloride and polycarbonate (Iversen et al., 2004; Lehner et al., 2005). Capsular polysaccharides on the bacterial cell surface play a pivotal role in biofilm formation. Colanic acid production is encoded by ESA_0115-01175, the wzABCKM gene in all Cronobacter spp. (Joseph et al., 2012c). Bacterial cells appear to attach more rapidly to hydrophobic non-polar materials such as Teflon and plastics than to hydrophilic surfaces such as glass or metals (Lehner et al., 2005). Furthermore, Cruz-Córdova et al. (2012) reported that flagella from C. sakazakii are involved in biofilm formation. In addition to flagella, two hypothetical proteins (ESA_00281 and ESA_00282) were identified as possible adhesin proteins that may contribute in biofilm formation (Cruz-Córdova et al., 2012). Flagella seem to aid Cronobacter cells to adhere to mammalian cells such as Caco-2 and other epithelial cells.

The ability of Cronobacter spp. to form biofilms provides protection from environmental stresses that impart resistance to cleaning and sanitizing agents (Ravishankar & Juneja, 2003; FAO/WHO, 2006). Kim et al. (2007) showed that resistance of Cronobacter spp. to disinfectants was increased when they formed biofilms compared with planktonic cells that were inoculated and dried onto the surface of stainless steel or even planktonic cells in culture. Additionally, a higher UV dose was required to kill Cronobacter growing in a biofilm than in the planktonic state (Jo et al., 2010).

Biofilm formation is of special importance in the food industry, because biofilms can act as a source of microbial contamination that might lead to spoilage of foods or contamination of food products undergoing processing (Lehner et al., 2005; Hartmann et al., 2010). Biofilms of Cronobacter formed on feeding areas and on equipment surfaces used in formula preparation are thought to have contributed to several neonatal outbreaks (Kim et al., 2006). Indeed, these observations support the increased risk of Cronobacter for formula-fed infants.

Antibiotic susceptibility and resistance

Prior to 1985, patients with Cronobacter infections were frequently treated with ampicillin, gentamicin and/or chloramphenicol. Willis & Robinson (1988) recommended the combination of gentamicin and ampicillin for the treatment of meningitis caused by Cronobacter. In contrast, Lai (2001) found that all Cronobacter isolates were uniformly resistant to ampicillin, cefazolin and extended-spectrum penicillins, whereas they were uniformly susceptible to trimethoprim/sulfamethoxazole and aminoglycosides. Recently, Al-Nabulsi et al. (2011) showed that streptomycin, gentamicin, kanamycin and ciprofloxacin are effective against both stressed and unstressed C. sakazakii cells, and thus these antibiotics might be appropriate choices for patient treatment regimes. Nevertheless, Nazarowec-White & Farber (1999) found that two Cronobacter strains out of eight were resistant to tetracycline and chloramphenicol. Kim et al. (2008) described Cronobacter strains recovered from foods that were resistant to cephalothin or ampicillin and susceptible to tetracycline. Hochel et al. (2012) found that all Cronobacter isolates varied in their susceptibility to tetracycline.

Muytjens & van der Ros-van de Repe (1986) examined Cronobacter susceptibility to 29 antimicrobial agents and found that Cronobacter were the most susceptible among other Enterobacteriaceae studied. This was supported by the finding of Stock & Wiedemann (2002) who showed that Cronobacter were repeatedly susceptible to β-lactams with no evidence for the expression of β-lactamase. In contrast, Pitout et al. (1997) demonstrated that some strains of Cronobacter produce β-lactamase at low levels. Caubilla-Barron et al. (2007) identified two Cronobacter isolates with extended-spectrum β-lactamase activity. Furthermore, Zhou et al. (2011) showed that one Cronobacter isolate from PIF produced extended-spectrum β-lactamase. According to the same study, susceptibility results showed that each isolate had different levels of resistance to β-lactam antibiotics (Zhou et al., 2011).

Although multiple antibiotic resistance (mar) operons were found in Cronobacter (Burgos & Varela, 2002), the overall level of antibiotic resistance was low compared with other food-borne pathogens and the increase of antibiotic-resistant Cronobacter could be a result of the overuse of antibiotics (Lee et al., 2012).

Heat adaptation

Some reports indicated that Cronobacter are more thermotolerant than most other bacteria belonging to the Enterobacteriaceae, whilst other studies reported that some Cronobacter isolates have no such tolerance (Nazarowec-White & Farber, 1997; Breeuwer et al., 2003; Osaili et al., 2009). The contradiction between these studies might be attributed to the growth conditions, isolate source (i.e. food or clinical) and growth phase or due to variation among the different Cronobacter strains (Arroyo et al., 2009; Walsh et al., 2011). In addition, different properties of food or formula, including low water activity and high fat content, affect thermotolerance of Cronobacter (Osaili et al., 2009). Nonetheless, adequate pasteurization conditions are sufficient to inactivate Cronobacter spp. during processing and reconstitution of PIF (Al-Nabulsi et al., 2011). Recently, an 18 kbp region containing 22 ORFs that are upregulated under heat adaptation conditions has been reported (Gajdosova et al., 2011). The major feature of the region is a cluster of conserved genes – most of them having significant homologies with known bacterial proteins involved in some type of stress response, including heat, oxidation and acid stress.

Desiccation adaptation

Cronobacter are considered more resistant to dry and osmotic stresses than other members of the Enterobacteriaceae (Feeney & Sleator, 2011). It is thought that this is probably due to the accumulation of trehalose inside the cells, which works as a protectant (Breeuwer et al., 2003). Furthermore, Álvarez-Ordóñez et al. (2014) concluded that de novo protein synthesis, DNA repair proteins and the maintenance of the structural integrity of cells are vital for the survival of Cronobacter in hyperosmotic media. Another explanation was offered by Barron & Forsythe (2007), who showed that Cronobacter were recoverable from PIF even after 2.5 years, assuming that their extracellular polysaccharides could protect the cells against the desiccation effects. Furthermore, Feeney & Sleator (2011) showed bioinformatically that C. sakazakii contained multiple copies of the osmotolerance homologues ProP and OpuC, and they concluded that the osmotic stress response of C. sakazakii appears to be regulated at the transcriptional, translational and post-translational levels, with RpoS most likely functioning as the global transcriptional regulator of the osmotolerance response.

Gurtler & Beuchat (2007) reported that the reduction in the viability of Cronobacter in PIF was less evident in formulas with water activity (a_w) values 0.25–0.30 compared with formulas with a_w = 0.43–0.50. Additionally, Cronobacter spp. can survive in infant cereals at low populations (2 c.f.u. g⁻¹) over a wide range of a_w (0.30–0.83) for up to 12 months with enhanced viability at lower a_w values (Lin & Beuchat, 2007). Furthermore, Dancer et al. (2009) showed that Cronobacter grown and dried in PIF survived better throughout drying than those cells grown and dried in tryptic soy broth, highlighting the role of some PIF components in the process of desiccation adaptation. Genes involved in desiccation resistance and osmotic stresses have been identified. For instance, genes encoding osmoprotectant glycine betaine and trehalose were found in all Cronobacter (Joseph et al., 2012c). Furthermore, all Cronobacter contain genes for β-carotene production, which is believed to protect bacteria against harmful oxygen radicals (Joseph et al., 2012c). Products of these genes are believed to help the bacteria to survive in dry food ingredients utilized for making infant formula preparations. Furthermore, several other genes were identified whose products could help Cronobacter to tolerate low water activity environments. The identified genes include cellulose biosynthesis operons, colonic acid exopolysaccharide, the capsular biosynthetic operon, the environmental persistence capsule and the curli GR55. The synergetic expression of these genes might provide resistance to desiccation stress (Grim et al., 2013). Using transposon mutagenesis, Álvarez-Ordóñez et al. (2014) identified genes responsible for the desiccation tolerance of Cronobacter. They found that the Cpx system, known as an envelope stress response regulator, and the sigma factors RpoN and RpoS seem to be the main signals regulating the bacterial response to hyperosmotic conditions. Furthermore, among the other identified genes, only dnaK and dnaJ, encoding two molecular chaperones, were important for C. sakazakii survival under desiccation.

Acid adaptation

Dancer et al. (2009) reported that 72 Cronobacter strains were able to grow at pH 4.5, whilst 98.6, 95.8 and 79.2 % of these tested strains were able to grow at pH 4.3, 4.1 and 3.9, respectively, indicating that Cronobacter cells are more acid tolerant than most related enteric pathogens. In addition, Edelson-Mammel et al. (2006) reported that 10 out of 12 strains showed <1 log decline at pH 3.5 over a 5 h incubation at 37 °C, whilst the most acid-sensitive strain showed ~3.5 log decline. Johler et al. (2010) showed that Cronobacter strains not only survived low pH, but were also able to grow to ~10⁹ c.f.u. ml⁻¹ at pH 4.5 within 24 h. Overall, Cronobacter have a good potency to survive in acidic environments that may enable them to cope with stomach acidity. This in turn may explain the increased susceptibility of infants to infection with Cronobacter spp., where stomach acidity is not fully developed.