Complete genome sequence and phenotype microarray analysis of Cronobacter sakazakii SP291: a persistent isolate cultured from a powdered infant formula production facility

Outbreaks of human infection linked to the powdered infant formula (PIF) food chain and associated with the bacterium Cronobacter, are of concern to public health. These bacteria are regarded as opportunistic pathogens linked to life-threatening infections predominantly in neonates, with an under developed immune system. Monitoring the microbiological ecology of PIF production sites is an important step in attempting to limit the risk of contamination in the finished food product. Cronobacter species, like other microorganisms can adapt to the production environment. These organisms are known for their desiccation tolerance, a phenotype that can aid their survival in the production site and PIF itself. In evaluating the genome data currently available for Cronobacter species, no sequence information has been published describing a Cronobacter sakazakii isolate found to persist in a PIF production facility. Here we report on the complete genome sequence of one such isolate, Cronobacter sakazakii SP291 along with its phenotypic characteristics. The genome of C. sakazakii SP291 consists of a 4.3-Mb chromosome (56.9% GC) and three plasmids, denoted as pSP291-1, [118.1-kb (57.2% GC)], pSP291-2, [52.1-kb (49.2% GC)], and pSP291-3, [4.4-kb (54.0% GC)]. When C. sakazakii SP291 was compared to the reference C. sakazakii ATCC BAA-894, which is also of PIF origin, the annotated genome data identified two interesting functional categories, comprising of genes related to the bacterial stress response and resistance to antimicrobial and toxic compounds. Using a phenotypic microarray (PM), we provided a full metabolic profile comparing C. sakazakii SP291 and the previously sequenced C. sakazakii ATCC BAA-894. These data extend our understanding of the genome of this important neonatal pathogen and provides further insights into the genotypes associated with features that can contribute to its persistence in the PIF environment.

Cronobacter can be isolated from a wide range of foods and environments (Baumgartner et al., 2009;Chap et al., 2009;El-Sharoud et al., 2009;Jaradat et al., 2009;Schmid et al., 2009). Specifically, contaminated powdered infant formula (PIF) has been epidemiologically linked with many of the neonatal and infant infections (Himelright et al., 2002;Bowen and Braden, 2006;Mange et al., 2006). Previous studies reported the isolation of Cronobacter from PIF, and the PIF production environment (Drudy et al., 2006;Mullane et al., 2008a,b), suggesting that this bacterium has the capacity to adapt to, survive and persist under desiccated environmental conditions. Comparison of environmental and clinical Cronobacter isolates, indicated that the desiccation tolerance exhibited might play a role in the persistence of Cronobacter in PIF and its associated low-moisture ingredients (Walsh et al., 2011;Beuchat et al., 2013). Stress response factors identified previously in Cronobacter, which include heatshock, cold-stresses, survival in dry conditions, water activity (a w ), and pH may contribute to this phenotype (Dancer et al., 2009a,b;Carranza et al., 2010;Chang et al., 2010;Arku et al., 2011). Genome sequencing efforts of Cronobacter species commenced in 2010. To date, 16 genomes are currently available, of which three, C. sakazakii ATCC BAA-894, C. sakazakii ES15 and C. turicensis z3032, are complete (Kucerova et al., 2010;Stephan et al., 2011;Shin et al., 2012;Grim et al., 2013).
Following on-going surveillance of a PIF production facility in our laboratory, an interesting isolate C. sakazakii SP291 was identified which exhibited a thermo-adapted phenotype when compared with other Cronobacter and Salmonella species tested under laboratory conditions (Cooney, 2012). In an effort to better understand C. sakazakii SP291, its genome was completely sequenced and compared to that of a PIF isolate C. sakazakii ATCC BAA-894, a whole grain isolate C. sakazakii ES15, a clinical isolate C. turicensis z3032 and other selected draft genomes. Additionally, we interrogated the phenome of C. sakazakii SP291, to determine the functionality of strain-specific genotypic traits that may contribute to its adaption capacity in a PIF production environment.

BACTERIAL ISOLATES STUDIED AND THEIR CULTURE CONDITIONS
Seventeen Cronobacter isolates used in this study are listed in Table 1. Cronobacter sakazakii SP291 was assigned according to the classic rpoB method described previously (Stoop et al., 2009;Lehner et al., 2012). The isolate was cultured routinely in an Isotherm® Forced Convection Laboratory Incubator (Esco GB Ltd., Downton, UK) at 37 • C on Trypticase Soy Agar (Oxoid Limited, Hampshire, UK) and stored at −80 • C on cryo-beads (Technical Service Consultants Ltd., Lancashire, UK).

DNA SEQUENCING, ANNOTATION, AND COMPARATIVE GENOMIC ANALYSIS
Total genomic DNA was purified using a DNeasy Blood and Tissue Kit (QIAGEN, Hilden, Germany) following the manufacturer's instructions. Concentrations were measured using a Nanodrop® (ND 1000) spectrophotometer (Labtech International Ltd., Luton, UK). Purified DNA was maintained at −20 • C until required. The whole genome sequencing and assembly methodology is described elsewhere . The complete chromosome and plasmid sequences were uploaded to the RAST (Rapid Annotation using Subsystem Technology) (Aziz et al., 2008) annotation server in a FASTA file format. The RAST server automatically identifies proteinencoding, tRNA and rRNA genes, assigns their functions, predicts the presence of subsystems in the genome, and reconstructs the metabolic network (Aziz et al., 2008). Genome to genome comparative analysis was performed in the SEED viewer as previously reported (Overbeek et al., 2005;Grim et al., 2013). Three complete genomes of C. sakazakii ATCC BAA-894 (Accession number CP000783-CP00785), C. sakazakii ES15 (Accession number CP003312) and C. turicensis z3032 (Accession number FN543093-FN543096) were uploaded and annotated in RAST, and used as reference sequences. Most probable insertion or deletion genome regions of C. sakazakii SP291 were identified as previously reported . In addition, nitrogen metabolism genes, stress-coding genes, as well as antibiotic and toxic compound resistant genes were determined based on significant identity alignments using BLAST. The genome sequence of C. sakazakii SP291 was deposited in GenBank under the accession numbers CP004091-CP004094. The accession numbers for other genome sequences studied were included in Table 1.

PHENOTYPE MICROARRAY ANALYSIS
Phenotype microarray (PM) analysis was performed on C. sakazakii ATCC BAA-894 and C. sakazakii SP291 using the OmniLog® automated incubator/reader (Biolog Inc., Hayward, USA) following manufacturer's instruction. All 20 plates (PM-1 through PM-20) inoculated with bacterial cell suspensions, were incubated at 37 • C and cell respiration was measured every 15 min for 48 h. The tetrazolium redox dye is reduced when bacteria respire, which provides both amplification and quantitation of the phenotype. Analysis was carried out using OmniLog® phenotype microarray software v1.2 to determine the phenotypic differences. Negative control wells, which contained the inoculated Omnilog™ growth medium, but without any substrate, were measured to normalize differences in inocula and redox dye oxidation between samples. The respiration profiles for both strains were compared using PM's integration function software and a significant divergent phenotype was identified when a difference in Omnilog™ units of 20, 000 ± 1800 or greater between the two strains was obtained.

Cronobacter sakazakii SP291 GENOME
The complete genome sequence of C. sakazakii SP291 is composed of a single, circular chromosome, 4.34 Mb in length with an average GC content of 56.9% along with three plasmids (denoted as 118.136 kb,57.2% GC;52.134 kb,4.422 kb, 54.0% GC) (Accession number CP004091-CP004094). The general features of the genome are presented in Table 2. A total of 4129 genes were identified on the chromosome, including 82 tRNA and 22 rRNA genes. The protein coding sequence (CDS) represents 86.3% of the genome and is organized into 4025 CDS, with an average length of 931 nucleic acids ( Figure A1). From the annotation of the three plasmids, it was determined that 116 genes cover 87.1% of pSP291-1, 74 genes cover 77.2% of pSP291-2 and 7 genes were located on pSP291-3 and accounts for 48.6% of this structure.

COMPARATIVE GENOMIC ANALYSIS OF C. sakazakii SP291 WITH THREE OTHER COMPLETED Cronobacter GENOMES
Cronobacter sakazakii SP291 and three other completed genomes: C. sakazakii ATCC BAA-894, C. sakazakii ES15 and C. turicensis z3032 were compared (Figure 1). For the purposes of this comparison, the C. sakazakii ATCC BAA-894 genome was used as the reference. Five genomic regions (denoted as GR-1 through −5, in Figure 1A) were identified and these were present in the other genomes but missing in C. sakazakii SP291 (Table S1). These GRs are discussed in detail below. Several unique prophages or phage-like elements of C. sakazakii ATCC BAA-894 were noted in GR-1 (genome positions 578,739. . .601,816), 245,990. . .2,272,660) and 962,630. . .3,022,134). Three specific genes were observed in GR-1, which included the DNA-methyltransferase subunit M and the S subunit of a type I restriction-modification system, along with a zinc binding domain/DNA primase, which is a phage P4-associated/replicative helicase denoted as RepA. A lambda phage portal protein, a large subunit of a terminase enzyme, along with some hypothetical proteins were noted in GR-3. In GR-4, a 1478 bp uncharacterized translocase gene required for O-antigen conversion and two-recombination genes, part of the bacteriophage ninR region, and denoted as ninB and ninG were identified. These annotations confirmed those previous reported (Kucerova et al., 2010). Interstingly, a putative bactoprenol glucosyl transferase was identified in C. sakazakii ATCC BAA-894, and shared with C. sakazakii ES15 and C. turicensis z3032, but not C. sakazakii SP291. Regulatory protein CII along with a phage Kil protein were annotated in C. sakazakii ATCC BAA-894 but not C. sakazakii SP291. Although protein CII was absent in C. turicensis z3032, the Kil protein was present (Stephan et al., 2011).
Genes unique to C. sakazakii SP291 were also noted and these were used as a reference to interrogate the genomes of the other strains. Six genomic regions ( Figure 1B, denoted as GR-a through -f, Table S2) were identified as being unique to C. sakazakii SP291. GR-a (genome regions 319,823. . .350,809), 643,874. . .165,774) and 471,984. . .2,499,614) contained a set of phage-and phage-related proteins along with some hypothetical proteins. A phage regulatory CII-like protein was identified in C. sakazakii SP291 and mapped within GR-a, which also matched a similar homolog observed in C. turicensis z3032. A holin protein, which controls the length of an infective cycle for bacteriophage (Wang et al., 2000), together with membrane proteins related to metalloendopeptidases were present in C. sakazakii SP291 alone, being located in GR-e. In 890. . .638,586), a YkfI toxin-encoding protein along a YfjZ-antitoxin encoding protein (the corresponding antitoxin to YpjF) were identified and unique to C. sakazakii SP291. This toxin-antitoxin protein pair was also reported in E. coli previously and was shown to regulate cell death through the disruption of essential cellular processes (Brown and Shaw, 2003). It has been proposed by Lewis (2000) that, under some circumstances, it may be evolutionarily advantageous for some cells in a population to undergo programmed cell death in order to provide nutrients for the remainder. Toxin-antitoxin pairs were noted in a previous study as most Cronobacter genomes contain a large number of them, which might be conserved, shared, or unique . GR-c (genome regions 891,557. . .912,700) contains seven interesting genes, which includes an uncharacterized protein YeeT, a NgrB protein, an ATP-dependent Clp protease, an ATP-binding subunit ClpA, a small HspC2 heat shock protein, a galactoside Oacetyltransferase-encoding gene and an anti-restriction protein KlcA which have been reported as a component part of a type I DNA restriction system (Serfiotis-Mitsa et al., 2010). A helicase protein, a glycerol dehydrogenase enzyme-encoding gene, and a DNA-cytosine methyltransferase were identified within 363,441. . .3,392,072). Of note, a type I restriction-modification system, specificity the S-subunit-like gene, was identified in C. sakazakii SP291, a feature which was noted earlier in C. sakazakii ATCC BAA-894 (Kucerova et al., 2010;.

COMPARATIVE GENOMIC ANALYSIS OF C. sakazakii SP291 AND SELECTED AVAILABLE GENOMES WITHIN THIS GENUS
Two earlier studies described the core genome of Cronobacter Grim et al., 2013). The availability of C. sakazakii SP291 genome has provided an opportunity to reevaluate the content of the Cronobacter core gemome, comparing it to other currently available genome sequences within the genus. Thus, a comparison between C. sakazakii SP291 and 16 other Cronobacer genomes ( Table 1) was performed in SEED viewer server.
Within the 11 Cronobacter sakazakii isolates compared, 57 annotated genes were present in C. sakazakii SP291, but absent in all other genomes, including 41 hypothetical proteins, 12 phage-and prophage-related genes/proteins and four other genes/proteins (Table S3). Among all seven Cronobacter species, there were 154 annotated genes/proteins absent in other species, including 122 hypothetical proteins, 4 phage-and prophagerelated genes/proteins and 28 unique genes/proteins (Table S4).
Interestingly, a conserved domain protein was identified that was unique to C. sakazakii SP291, which is associated with retron-type reverse transcriptase. Fifteen genes were shared with other species by C. sakazakii SP291, but were absent in all the C. sakazakii genomes compared to date, and these consisted of a retron-type RNA-directed DNA polymerase, a holin protein which controls the timing of bacteriophage infections as mentioned earlier, a topoisomerase IA-encoding protein, and 12 phage-and prophage-related proteins. There were 31 proteins, which are only shared with some of the C. sakazakii genomes by C. sakazakii SP291 and which were absent among the other six species. These included a sodium-dependent phosphate transporter protein, a RelE antibacterial toxin protein, a RelB protein (antitoxin to RelE), a probable poly (beta-D-mannuronate) O-acetylase protein, two putative periplasmic proteins, a possible secretory protein, a GTPase protein, denoted as NgrB, a mobile element protein, a galactoside O-acetyltransferase protein, a mannose-6-phosphate isomerase, class I protein, a different locus type I restriction-modification system, specificity the S subunit-like protein, a predicted transcriptional regulator COGs COG2378, permeases of the major facilitator superfamily, a superfamily II DNA/RNA helicases, SNF2 family, a DNA modification methylase, an IS1 transposase OrfA protein, a probable tonB-dependent receptor yncD precursor, a putative ORF-4 protein, a putative ORF (located using Glimmer/Genemark), seven beta-fimbriae probable major subunits, and four phage related proteins.
Several common genes were identified in plasmid group 1, These consisted of a complete ABC transporter (which could function to transport iron; vitamin B12; siderophores and hemin), including the ATP-binding component, the periplasmic substrate-binding module and the permease element. These genes were identified in all three plasmids. An aerobactin siderophore receptor (the IutA/TonB-dependent siderophore receptor) was shared between the three plasmids, while a Cronobacter plasminogen activator (cpa) homolog has only been mapped to pESA3 and pSP291-1, but not pCTU1, which is in agreement with the results reported by Franco et al. (2011) and Grim et al. (2012) (Figure A2). Three arsenical resistance genes were identified on all three plasmids along with pCTU3. Genes corresponding to commonly shared proteins on all three plasmids included a C-terminal helicase protein, a HipA protein previously reported to be required for growth arrest and multi-drug resistance in Escherichia coli (Correia et al., 2006), a hypothetical-encoding gene ycgF reported to be a direct anti-repressor which acts in the blue-light response of E. coli (Tschowri et al., 2009), a starvation sensing protein RspA, a magnesium transporting P-type 1 ATPase protein, transcriptional regulators, including members of ArsR family, GntR family (Kucerova et al., 2010;, LysR family and TetR-family, a MFS superfamily transporter, a Trk system encoding the potassium uptake protein TrkG, and an uncharacterized protein ImpD. A two-component response regulator protein, a two-component system sensor protein, three uncharacterized proteins ImpB, ImpC, and ImpJ/VasE, a glutathione S-transferase protein, a membrane protein, suppressor for copper-sensitivity ScsB, a hypothetical ABC transport system, periplasmic component, a RND efflux transporter, a suppression of copper sensitivity: putative copper binding protein ScsA were shared between pSP291-1 and pESA3, but not pCTU1, which confirmed the findings from previous studies (Kucerova et al., 2010;. In plasmid group 2, 15 heavy metal (copper, cobalt, zinc, cadmium, lead, and mercury) resistance genes were shared by both plasmids ( Figure A3). An osmosensitive K + channel histidine kinase protein (KdpD), and a virulenceassociated protein vagC were also present in both plasmids. PCR analysis confirmed the presence of a pCTU3 IncH1-like origin of replication gene, repA in C. sakazakii SP291 (data not shown).
Interestingly, pSP291-1 contained two unique proteins, a histone acetyltransferase HPA2 and related acetyltransferases protein, along with an uncharacterized protein ImpH/VasB. Six specific proteins were found in pSP291-2, which included a putative glutathione S-transferase protein, a LysR family transcriptional regulator, a putative phage-associated acyl carrier protein, a Sadenosylmethionine: tRNA ribosyltransferase-isomerase protein, permeases of the major facilitator superfamily and an abortive infection protein. Various pSP291-3 proteins including mobilization proteins MobB, MobC, MobD, and DNA relaxase MbeA, which were not shared with any of the other plasmids, were also identified.

COMPARATIVE PHENOTYPIC PROFILING OF C. sakazakii ATCC BAA-894 AND SP291
The phenotypic microarray (PM) platform was used previously to support the re-classification of this bacterial genus (Iversen et al., 2008). By comparing the phenotypes of C. sakazakii ATCC BAA-894 and C. sakazakii SP291 expressed across the complete array, interesting differences were observed and these were subsequently assessed in light of comparisons made at the genome level despite of the same PIF orgin. General differences, noted in the phenotypes between the two strains were described in the form of a heat map shown in Figure 2 (the corresponding numerical data is shown in Table S6). Phenotypic differences based on the bacteria's ability to utilize carbon, nitrogen, phosphorous and sulfur sources, as well as other nutrient supplements were noted. Furthermore, growth responses to osmolytes and different pH growth environments, as measured by the array were also observed for both strains. Antibiotic resistance patterns and the ability to respire in the presence of toxic compounds differed.

Carbon, nitrogen, phosphorous, and sulfur, other nutrient supplement metabolite utilization
Bacteria require a sufficient supply of various biomolecules to support their metabolic activity. In natural environments, where these bacteria are often found, it is to be expected that only limited amounts of these nutrients may be available. To support efficient adaptation and to enable growth in these conditions, bacteria will evolve redundant metabolic systems to support the utilization of a broad range of different substrates, with varying efficiencies. The PM array data gives an insight into how these features differ, between C. sakazakii SP291 and C. sakazakii ATCC BAA-894.
A number of phenotypic differences based on their ability to utilize a range of carbon sources were noted (Figure 2A and Table S6). When compared with C. sakazakii ATCC BAA-894, C. sakazakii SP291 could grow faster in m-inositol and slower in succinic acid, dulcitol, D,L-α-glycerol phosphate, D,L-malic acid, Tween 20, α-ketoglutaric acid, uridine, bromosuccinic acid, glycolic acid, inosine, and dextrin. In contrast there were little or no differences in growth rates when other carbon sources such as methyl pyruvate, mannose, and β-methyl-D-glucuronic acid were compared.
Differences in phenotypes based on the metabolism of carbon sources were compared at the genome level within the carbohydrate subsystem (Table S7). Interestingly, nine inositol catabolism genes were annotated in the C. sakazakii SP291 genome (Table S8), which supported the PM data. Furthermore, a pentose phosphate pathway gene, a lactose utilization gene, and a sucrose utilization gene were also annotated in the C. sakazakii SP291 genome specifically although no evidence to support their activity was found following PM analysis. Similarly, a maltose and maltodextrin utilization gene and a lactate utilization gene were annotated in C. sakazakii ATCC BAA-894 alone, with supporting evidence for the activity lacking from the PM array data. In all, 428 annotated genes related to carbon metabolism were shared between C. sakazakii ATCC BAA-894 and C. sakazakii SP291, which included 10 chitin and N-acetylglucosamine utilization genes, five fructoselysine (amadori product) utilization pathway genes, five dehydrogenase complexes genes, a dihydroxyacetone kinases gene, 14 Entner-Doudoroff pathway genes, and others. Dancer et al. (2009a,b) reported that for Cronobacter species the availability and utilization of a nitrogen source was an important determinant for biofilm formation when growing in skim milk, and that strong biofilm formers were responsible for coagulation of skim milk (Dancer et al., 2009a). Data from the phenotypic microarray, showed no differences in nitrogen metabolism when C. sakazakii ATCC BAA-894 and C. sakazakii SP291 were compared ( Figure 2B and Table S6). Interestingly, when regions of these two genomes known to encode genes associated with nitrogen metabolism were compared, a 16-kb locus, consisting of eight genes was found to be absent in C. sakazakii SP291 compared to C. sakazakii ATCC BAA-894 (Table 3). BLAST analysis of the region facilitated the identification of the corresponding genes located at this position. This locus, contained two nitrate transport proteins NrtB and NrtC, two nitrite reductase proteins NasB and NasA, a respiratory nitrate reductase NarL, a nitrate/nitrite-sensing protein NarX, a nitrite extrusion protein 1 NarK, and a nitrate reductase 1, alpha subunit NarG. This region was also present in C. sakazakii ES15 and C. turicensis z3032. Furthermore, 24 nitrate genes were broadly shared between both of the genomes, which was supported by data from the PM analysis.
C. sakazakii SP291 was found to grow significantly slower in minimal media supplemented with phosphorous containing compounds (Figure 2A and Table S6), particularly in O-phospho-D-tyrosine, phospho-L-arginine, D,L-α-glycerol phosphate, β-glycerol phosphate, phosphoryl choline, phosphoenol pyruvate, D-glucose-6-phosphate, adenosine 3 -monophosphate, guanosine 2 -monophosphate, guanosine 3 -monophosphate, guanosine 5 -monophosphate, guanosine 2 ,3 -cyclic monophosphate, cytidine 2 -monophosphate, cytidine 3 -monophosphate, thymidine 5 -monophosphate, and uridine 5 -monophosphate. Genome annotation provided a conflicting view as determined by the genes identified (Table S9). Twenty-nine phosphorus metabolism genes were broadly shared between C. sakazakii SP291 and C. sakazakii ATCC BAA-894, including eight high affinity phosphate transporters and control of pho-related regulon genes, 18 phosphate metabolism genes, and three polyphosphate genes. Cronobacter species cultured from a PIF production site were compared for their ability to grow in different food matrices (Cooney, 2012). Some demonstrated a slower growth rate compared to others, a feature that might contribute to their enhanced survival in this environment.
No differences in the metabolism of sulfur containing compounds were observed following a comparison of these strains after PM analysis (Figure 2A and Table S6). Forty-nine sulfur metabolism genes were shared by C. sakazakii ATCC BAA-894 and C. sakazakii SP291 (Table S10). These consisted of 17 inorganic sulfur assimilation genes, eight alkanesulfonate assimilation genes, five alkanesulfonates utilization genes, six L-cystine uptake and metabolism genes, four taurine utilization genes, three galactosylceramide and sulfatide metabolism genes, and six thioredoxin-disulfide reductase genes.
Iron is an essential nutrient for bacterial growth and the process of iron acquisition is generally thought to be a prerequisite for a pathogen to establish an infection when entering a host, a feature previously reported in Cronobacter species (Crosa and Walsh, 2002;Franco et al., 2011;Grim et al., 2012). Highaffinity iron binding molecules, such as siderophores, and specific iron transport systems function to sequester iron from the environment when bacteria are subjected to iron-limiting growth conditions (Grim et al., 2012). Interestingly, analysis of the PM data showed no major differences between C. sakazakii SP291 and C. sakazakii ATCC BAA-894, in terms of their metabolism of iron or other nutrient supplements. Several transport systems were annotated in C. sakazakii SP291, and which are shared with C. sakazakii ATCC BAA-894 (Kucerova et al., 2010;, including a ferric hydroxamate ABC transporter denoted as FhuCDBA, 16 ferric enterobactin transporter proteins (including EntA,EntE,EntD,EntB,Fes,EntS,EntF,YbdZ,FepC,FepD,FepG,FepE,FepB,EntC,FepA2,and EntH), a ferrous iron transporter EfeUOB, along with a hemin transporter system, including a ferric reductase protein FhuF and a periplasmic binding protein TonB. A gene summary of iron acquisition and metabolism markers in C. sakazakii SP291 chromosome is shown in Table S11.
Additionally, iron acquisition and metabolism genes were also identified on pSP291-1 (Table S5), which were indistinguishable from that previously reported to be present on pESA3 of C. sakazakii ATCC BAA-894 (Kucerova et al., 2010; and pCTU1 of C. turicensis z3032 Grim et al., 2012). Target genes from previous reports, such as the RepFIB-like origin of replication gene repA, two plasmidborne iron acquisition systems (eitCBAD and iucABCD/iutA), as well as the Cronobacter plasminogen activator cpa gene were all present in pSP291-1, with no evidence of the 17-kb type VI secretion system (T6SS) locus identified previously in pESA3 along with a 27-kb region encoding a filamentous hemagglutinin gene (fhaB), its specifc transporter gene (fhaC), and associated putative adhesins (FHA locus) identified in pCTU1 (Kucerova et al., 2010;Franco et al., 2011;Grim et al., 2012Grim et al., , 2013. These features support the hypothesis that these plasmids have evolved from a single archetypal backbone that included an iron acquisition system. Our sequence analysis and those of other groups Grim et al., 2013) did not find evidence of plasmid mobilization genes associated with the several plasmid group 1 genomes.

Osmolyte tolerance and survival in different pH environments
When present in different environments, bacteria must develop strategies that promote their survival. Genetic adaptation is derived from modifications of gene expression, via mutations, the acquisition of new and beneficial gene traits, or when these new traits are brought under control of a regulator that was already present in the core genome of the organism's ancestor (Maurelli, 2007). The outcome is that the organism is now better equipped to survive within the new ecological niche. It is generally thought that genes that are no longer compatible with the new lifestyle are selectively inactivated either by point mutation, insertion, or deletion and the contribution of gene loss to an organism's evolution is only now beginning to be appreciated (Maurelli, 2007). Based on our understanding of Cronobacter species epidemiology, these organisms are considered as environmental bacteria. Therefore their ability to survive adverse conditions would be critical. Phenotypes associated with growth in a range of osmolytes and in different pH growth environments were measured by PM analysis (Figure 2C and Table S6) as an indirect reflection of challenging environmental niches. In response to the presence of osmolytes, C. sakazakii SP291 could tolerate 100 mM sodium nitrate compared with C. sakazakii ATCC BAA-894. In contrast, the former grew slower in solutions containing 5% NaCl, 4% potassium chloride, 4% urea, 4-11% sodium lactate, 200 mM sodium phosphate at pH 7 and 20 mM sodium benzoate at pH 5.2. These observations are consistent with what has been suggested previously, in that when a selected adaptation event occurs, and the bacterium enters a new environment such as the human host, phenotypes change (Tall, unpublished observations). Comparing the ability of the environmental isolate C. sakazakii SP291 to survive over a range of different pH growth conditions with that of the PIF isolate C. sakazakii ATCC BAA-894, the former grew faster in a growth condition of pH 9.5 with phenylethylamine, whilst its growth was slower in pH 4.5 with L-proline. This example demonstrates the gain of one phenotype consistent with the inability to survive in the human host (ability to survive in high pH growth conditions) compared to the loss of a sufficient acid resistance response. In this case, the pathoadaptative event that resulted in an increased persistence in the environment comes at the expense of decreased commensal fitness of the microbe (a patent acid response) to survive the acidity of the stomach. However, a greater number of genomes and strains should be evaluated to rule out strain to strain variation.
Annotation of the genome suggested that C. sakazakii SP291 contained a repertoire of genes that could function to aid survival under stressed conditions, such as osmolyte tolerance and different pH environments (Table 4). One hundred and fifty-two annotated genes were identified as being involved with various stress responses. Their presence in the genome may provide early insights into how C. sakazakii SP291 adapts to and survives under different stressful growth conditions.
In recent studies involving Salmonella species, a picture of the transcriptome in low-moisture growth conditions has begun to emerge (Frossard et al., 2012;Finn et al., 2013). Allied to this, 25 genes involved in osmotic stress, and covering 16.4% of the stress response genes were identified in C. sakazakii SP291. Interestingly, the osmoprotectant ABC transporter denoted as YehZYXW in the Cronobacter genome, together with the L-proline glycine betaine MFS transporter ProP, and the ABC transporter ProU systems (composed of ProV, ProW, ProX) designed in Escherichia coli (Checroun and Gutierrez, 2004) and Salmonella Typhimurium (Cairney et al., 1985) were identified in the C. sakazakii SP291 genome. Moreover, an osmoregulator transporter including genes opuCA, opuCB, opuCC, and a fourth gene, which was also an ABC transporter denoted as opuCD here, was 77% similar to that of the osmU operon (osmVWXY ) in Salmonella (Frossard et al., 2012) at the gene level. Other osmotically functioning genes identified included the betaine/carnitine/choline transporter (BCCT) family, which acts to transport betaine and choline. This operon     consists of a high-affinity choline uptake gene betT, a helixturn-helix (HTH)-type transcriptional regulator betI, which was previously identified in E. coli (Lamark et al., 1991), a betaine aldehyde dehydrogenase betB gene and a choline dehydrogenase gene betA. An in silico assessment of those loci involved in osomotolerance comparing Cronobacter sakazakii ATCC BAA-894 and E. coli K12 MG1655 identified these latter features also (Feeney and Sleator, 2011). Interestingly, several other genes linked to osmotic stress conditions were identified in the genome of C. sakazakii SP291, which included five osmoregulation genes (aqpZ, glpF, osmY, ompA, and yiaD), four osmotic stress cluster genes (yciM, osmB, pgpB, and yciT) and six osmoregulated periplasmic glucan genes (mdoH, mdoC, mdoD, mdoG, opgC and mdoB). None of these genes were identified previously by Feeney and Sleator (2011). Finally, ompA which encodes an outer membrane porin, was identified in C. sakazakii SP291 and is a recognized virulence marker (Kim et al., 2010). Experiments to investigate the nature of the C. sakazakii responses to cold-and heat-shock conditions have been reported (Shaker et al., 2008;Carranza et al., 2010;Chang et al., 2010;Al-Nabulsia et al., 2011;Gajdosova et al., 2011). Following exposure to extreme temperatures of cold-shock at −20 • C, or heat-shock at 47 • C, survival of Cronobacter sakazakii was significantly enhanced (Chang et al., 2010). Carranza et al. (2010) reported that when exposed to higher temperatures, several potential virulence factors were up-regulated. The fact that the pathogenic potential of Cronobacter species may be related to its ability to survive at higher temperatures, warrents further investigation. From the genome sequence of C. sakazakii SP291, the cspA family of coldshock genes (including cspA, cspC, cspD, cspE and cspG) along with 11 other genes annotated as heat-shock genes, part of the dnaK gene cluster, including dnaJ, dnaK, yggX, gshB, grpE, rdgB, rpoH, hemN2, rph, rsmE, prmA, hslR, yraL, and smpB genes were conserved.
Using a top-down proteomics approach, Williams et al. (2005) identified a candidate protein in C. sakazakii, known to be associated with thermotolerance in Methylobacillus flagelatum, and which was denoted as KT. In a recent study, the genomic region containing this presumptive marker of thermotolerance was compared to similar regions in other bacteria (Gajdosova et al., 2011). An in silico analysis showed that this thermotolerace KT-region was present in 4 of 14 isolates consisting of seven Cronobacter species studied by . Cronobacter sakazakii SP291 can survive desiccation for long periods of time at an average temperature of 56.7 • C, similar to that recorded when spray drying is in operation during PIF production (Cooney, 2012). Interestingly, C. sakazakii SP291 was negative for the KT marker, as determined by PCR (data not shown). Apart from the locus between orfA-orfE, when the corresponding region of the SP291 genome was compared to that of C. sakazakii ATCC 29,544, this region was devoid of the KT-encoding homolog (Figure 3). In light of the thermo-adapted phenotype possessed by C. sakazakii SP291, this finding suggests that there may be other thermotolerance survival mechanisms expressed by C. sakazakii SP291.
As mentioned above, Cronobacter species have the capacity to survive in desiccated environments for long periods, a phenotype that is linked to their epidemiology and routes of infection. As an example of genes linked to this phenotype, the yih-encoding operons, consisted of 10 annotated genes present in the genome of C. sakazakii SP291. Desiccation-related proteins YihU, YihT, YihR, YihS, YihQ and YihV have conserved domains which function in carbohydrate transport and metabolism. YihO is a glucuronide  transport protein whilst YihP is homologous to it. YshA is an outer membrane sugar transport protein and YihW is a deoRtype transcriptional regulator, reported to negatively regulate the expression of yihU-oyshA in Salmonella (Gibson et al., 2006). This operon was reported to be up-regulated following desscication stress in Salmonella. Interstingly, the yih operon are conserved in 17 annotated Cronobacter genomes (strain information of the genomes are listed in Table 1) and noted previously . The ability of a bacterium to eliminate toxic compounds from the cell is an important survival mechanism. Nine genes involved in detoxification were identified in the C. sakazakii SP291 genome. These included a tellurite resistance-encoding gene tehB, which matches a 593 bp hypothetical protein in C. sakazakii ATCC BAA-894. However, a tellurite resistance region (terACDYZ) was reported only in C. sakazakii ATCC BAA-894 (Table 4), and with the exception of the terC-encoding marker in C. turicensis z3032, was not identified in any of the other Cronobacter species genomes sequenced (Kucerova et al., 2010;Grim et al., 2013). Genes involved in the detoxification of organic pollutants, including a D-tyrosyl-tRNA (Tyr) deacylase-encoding dtd, two glutathione-dependent pathway formaldehyde detoxification genes (frmA and yieG), three genes involved in the uptake of selenate and selenite (dedA, cysA, and its homolog tsgA), and two uncharacterized genes (ydsK and ydcL), were also identified in C. sakazakii SP291. This feature supports an earlier report describing the ability of Cronobacter species to detoxify and survive in tannery wastewater effluents (Chandra et al., 2011).
Oxidative stress is an example of an important bacterial stress response, with 62 annotated genes covering 40.8% genome linked to this sub-system. These genes included the zinc uptake regulation zur, which was reported as involved in the oxidative stress response of Streptomyces coelicolor (Shin et al., 2007). Other stress-related genes included seven periplasmic stress related genes, a bacterial hemoglobin gene, seven genes involved in carbon starvation, four commensurate regulon activation genes, a flavohaemoglobin gene, five hfl operon genes, five phage shock protein (psp) operon genes, a sugar-phosphate stress regulation gene, and five universal stress protein family genes.

Resistance to antibiotics and toxic compounds
C. sakazakii was originally reported to be susceptible to a panel of 69 antimicrobial agents (Stock and Wiedemann, 2002). Subsequently, a tetracycline-resistant C. sakazakii cultured from  a Chilean freshwater salmon farm (Miranda et al., 2003) was isolated, followed by a report of a trimethoprim and neomycin resistance isolate cultured from fresh domiati cheese (El-Sharoud et al., 2009). More recently, isolates resistant to cephalothin were recovered from dried food (Chon et al., 2012). The emergence of strains that have become resistant to antimicrobial compounds is of great concern to public health (Dumen, 2010;Yan et al., 2012). Figure 2D shows a heat map comparing C. sakazakii SP291 and C. sakazakii ATCC BAA-894 and Table 5 provides a summary of the significant changes in PM redox measurements, after bacterial growth in microtitre wells containing a number of antimicrobial compounds as part of the phenotypic microarray. Compared to C. sakazakii ATCC BAA-894, C. sakazakii SP291 exhibited more activity in the presence of 5,7-dichloro-8-hydroxyquinoline, 5-nitro-2-furaldehyde semicarbazone, hexamminecobalt (III) chloride, poly-L-lysine, protamine sulfate, ornidazole, tobramycin, streptomycin, apramycin, iodonitro tetrazolium violet, amoxicilin, neomycin, and sisomicin; while exhibiting a reduced activity in the presence of phleomycin, ciprofloxacin, cinoxacin, dichlofluanid, tolylfluanid, guanidine hydrochloride, colistin, methyl viologen, sodium azide, guanazole, rifamycin SV, glycine hydroxamate, D,L-methionine hydroxamate, cefmetazole, and cloxacillin.
Careful analysis of the PM data showed an interesting phenotype, related to bioactive and toxic anions. C. sakazakii SP291 survived significantly better in sodium metaborate, potassium tellurite, sodium m-arsenite, sodium tungstate, sodium periodate, sodium arsenate, and antimony (III) chloride compared with C. sakazakii ATCC BAA-894. In contrast the latter bacterium, exhibited a distinct phenotype in sodium bromate. These observations suggested that C. sakazakii SP291 elaborates a greater ability to counter the effects of a broader range of heavy metals, a characteristic that could be facilitate adaptation in powered infant formula manufacturing environments where metallic compositions such as quaternary containing disinfectants are used for decontamination. This resistance phenotype may be globally regulated as well. Together, this information may explain how this organism adapted to the manufacturing environment.

CONCLUSIONS
Stress responses and resistance to antibiotic and toxic compounds are interesting phenotypes identified in C. sakazakii SP291, when compared to C. sakazakii ATCC BAA-894, based on the comparative phenotypic microarray analysis in parallel with the annotation of its genome. Given this the fact that the PIF production environment is a stressful ecological niche, the osmoprotectant ABC transporters including YehZYXW, ProP, ProU, and OpuCABCD can be expected to play a role to support bacterial survival, as reported in other microorganisms previously (Cairney et al., 1985;Checroun and Gutierrez, 2004;Frossard et al., 2012;Finn et al., 2013). Notably, C. sakazakii SP291 possesses a greater ability to survive in a broader range of heavy metals, as quaternary containing disinfectants which include metallic compositions are often used for PIF manufacturing environment disinfection. In conclusion, genome analysis of C. sakazakii SP291 along with its metabolome highlighted a number of potential features, which might be considered as candidates for future studies to extend our understanding of the persistence and virulence mechanisms deployed by this bacterium in the production environment. These data provide an early insight into how a factory isolate may survive in a desiccation condition and adapt to the PIF production environment and associated food matrices.

ACKNOWLEDGMENTS
We thank Drs. Marta Martins and Orla Condell for their assistance with the PM data analysis and to Ms. Sarah Finn for critical reading of the manuscript.