Evolution in Quantum Leaps: Multiple Combinatorial Transfers of HPI and Other Genetic Modules in Enterobacteriaceae

Horizontal gene transfer is a key step in the evolution of Enterobacteriaceae. By acquiring virulence determinants of foreign origin, commensals can evolve into pathogens. In Enterobacteriaceae, horizontal transfer of these virulence determinants is largely dependent on transfer by plasmids, phages, genomic islands (GIs) and genomic modules (GMs). The High Pathogenicity Island (HPI) is a GI encoding virulence genes that can be transferred between different Enterobacteriaceae. We investigated the HPI because it was present in an Enterobacter hormaechei outbreak strain (EHOS). Genome sequence analysis showed that the EHOS contained an integration site for mobile elements and harbored two GIs and three putative GMs, including a new variant of the HPI (HPI-ICEEh1). We demonstrate, for the first time, that combinatorial transfers of GIs and GMs between Enterobacter cloacae complex isolates must have occurred. Furthermore, the excision and circularization of several combinations of the GIs and GMs was demonstrated. Because of its flexibility, the multiple integration site of mobile DNA can be considered an integration hotspot (IHS) that increases the genomic plasticity of the bacterium. Multiple combinatorial transfers of diverse combinations of the HPI and other genomic elements among Enterobacteriaceae may accelerate the generation of new pathogenic strains.


Introduction
The emergence of new pathogenic strains among the Enterobacteriaceae depends largely on the horizontal transfer of virulence determinants via mobile genetic elements such as plasmids, phages, genomic islands, and genomic modules [1][2][3][4][5][6][7][8][9][10][11]. Moreover, a single horizontal transfer event can alter the phenotypic and virulence characteristics. Thereby a normally benign organism can be transformed by a single step into a pathogen. This has been dubbed ''evolution in quantum leaps'' [3]. Enterobacter hormaechei is the most commonly isolated nosocomial pathogenic species of the Enterobacter cloacae complex (ECC) [12,13]. A nationwide outbreak of a multidrug-resistant E. hormaechei outbreak strain (EHOS) occurred in The Netherlands [14][15][16]. This strain spread throughout hospitals, despite the adequate implementation of internationally accepted infection prevention guidelines, and caused invasive infections in more than 100 patients [16]. Epidemic strains, due to their prevalence, have a greater chance of acquiring new virulence and resistance genes [17,18]. In a previous study, we showed that the chromosome of the EHOS contained the High Pathogenicity Island (HPI), which most likely increased virulence of the EHOS [19,20]. This is supported by the fact that Yersinia spp., Escherichia coli, and possibly Klebsiella pneumoniae isolates containing the HPI are more virulent than isolates lacking this island [7,9,[21][22][23][24]. The HPI segment that encodes the iron uptake and regulation system comprises 11 genes and is nearly identical in all detected isolates. Yersiniabactin is synthesized by a complex assembly line in which YbtS, YbtE, HMWP1, HMWP2, and YbtU are essential proteins [25,26]. HMWP1 and HMWP2 (High Molecular Weight Protein 1 and 2) are 350 and 230 kDa proteins that are encoded by irp1 and irp2, respectively [27,28]. Other genes are involved in the regulation and transport of yersiniabactin, such as the fyuA gene, which encodes the yersiniabactin receptor.
Three basic variants of the HPI have been described. All HPI variants contain a common P4-like integrase, called intB, and a region that encodes for an iron uptake system followed by an ATrich sequence. First, HPIs from Yersinia species with one or more IS elements at the 39-end were described [21,29]. Next, a putative integrative and conjugative element (ICE) with an HPI was found in E. coli ECOR31 (HPI-ICEEc1) [6,10]. Then, an ICE with a high similarity to HPI-ICEEc1 was found in a K. pneumoniae isolate [7]. This ICE, named HPI-ICEKp1, contains the conserved part encoding the iron uptake system, a putative integrative and conjugative element and two additional genetic segments relative to HPI-ICEEc1. One of these segments produces a catecholatetype siderophore and a regulator of the mucoid phenotype and the other segment contains hypothetical genes with similarity to genes found in Nitrobacter hamburgensis. The other segment, which is located after the HPI segment but in front of the putative ICE segment, is similar to part of a large plasmid in K. pneumoniae, and it contains genes responsible for DNA conjugative transfer [7]. Genes located on the ICE segment of HPI-ICEKp1 that are necessary for excision and conjugation are 99% identical to those from HPI-ICEEc1. However, HPI-ICEKp1 lacks a putative helicase gene described for HPI-ICEEc1 [7,10]. The integrase plays a critical role for horizontal transfer of genomic islands. Like phage integrases, it acts as a site-specific recombinase that catalyses both excision and integration [6]. attO, a 17 bp directrepeat that flanks HPI-ICEEc1 and HPI-ICEKp1, is the supposed integration site for HPI-ICE. HPI-ICE integrates by recombination at attO, resulting in a duplication of these sequences [7,10]. The first described self-transferable genomic island was HPI-ICEKp1. It was transferred to another strain in vitro via conjugation and possibly also in vivo to other K. pneumoniae isolates [7].
The aim of this study was to determine the heterogeneity of the HPI in the EHOS and its genetic relationship to HPIs in other Enterobacteriaceae. Therefore, whole genome sequencing was performed on the EHOS. The EHOS was found to contain a new HPI-ICE variant we termed HPI-ICEEh1, with an ICE element highly similar to the ICE element of HPI-ICEEc1 from E. coli ECOR31 [10]. Furthermore, HPI-ICEEh1 was located in a region of the chromosome between another genomic island and three other genetic modules, which were all located next to each other. This region is an integration hotspot (IHS) of E. hormaechei, and the possibility of excision, circularization and transfer of the different genomic islands was investigated.

Bacterial Isolate and DNA Isolation
Based on the similarity between PFGE patterns, the clinical invasive isolate 05-545 was selected as a representative for the EHOS strain. The isolate was identified in 2005 in a routine clinical setting as Enterobacter cloacae using the Phoenix 100 automated microbiology system with version V3.22 software (Becton Dickinson Biosciences, Sparks, MD, USA). By rpoB and hps60 genotyping, the isolate was identified as E. hormaechei (accession numbers: EU643417, EU643052) [13]. DNA was extracted using a NucleoSpin Tissue Kit (Macherey-Nagel GmbH & Co. KG, Düren, Germany) according to the manufacturer's guidelines.

Genome Sequencing and Annotation
DNA was sequenced using 454 pyrophosphate sequencing technology with 24-fold coverage by Roche Applied Sciences (Roche Diagnostics, Penzberg, Germany). Subsequently, resulting reads were assembled in contigs, using the 454 Newbler assembler. A hypothetical order of the contigs was established using the genome of Enterobacter sp. 638 as a scaffold. This hypothetical order of contigs was used to concatenate them to each other with the recognition sequence (NNNNNCACACACTTAATTAATTAAG-TGTGTGNNNNN) between contigs. This sequence contains a stop codon in each reading frame to prevent the formation of fusion proteins at the contig borders during automated annotation. Subsequently, the concatenated genome DNA sequences were submitted to the JCVI Annotation Service, where it was run through JCVI's prokaryotic annotation pipeline. Included in the pipeline are gene finding with Glimmer, Blast-extend-repraze (BER) searches, HMM searches, TMHMM searches, SignalP predictions, and automatic annotations from AutoAnnotate. (http://www.tigr.org/tigr-scripts/AnnotationEngine/ann_engine.cgi) Subsequently, the contig of interest was manually checked by comparing (putative) open reading frames (ORFs) with sequences deposited in GenBank via BLAST searches.

Bacterial Isolates Used for HPI Prevalence and Genotyping
A total of 717 ECC isolates, one isolate per patient, were included. All isolates were genotyped by pulsed-field gel electro-phoresis (PFGE) [14][15][16]20]. The isolates were divided into UMCU outbreak related (UMCU-ECC) and non-UMCU outbreak related isolates (non-UMCU-ECC). The 305 UMCU-ECC isolates originated from the University Medical Center Utrecht, Utrecht, The Netherlands (UMCU). These isolates were representing the ECC population in the UMCU from 2001-2005 during the EHOS outbreak [16]. Of the isolates, 128 belonged to the EHOS clone (outbreak I) and 14 isolates represented three small outbreaks in which the patients had entirely overlapping times of hospitalization: outbreak IV (E. cloacae IV; N = six patients), outbreak VIII (Enterobacter asburiae; N = five patients), and outbreak IX (E. hormaechei; N = three patients).
The 412 non-UMCU-ECCs consist of 217 aminoglycosideresistant ECC isolates obtained from 15 other Dutch clinical microbiology laboratories [16], 95 blood culture isolates obtained from 1989 through 2000 in the UMCU and 100 isolates obtained as part of a European antibiotic resistance surveillance (ENARE) study [30].
To study possible transfer of the HPI, a third group was included. This group contained HPI-positive Enterobacteriaceae isolates that did not fulfill the criteria of the first two groups. This third group contained 20 isolates, (four ECC, eleven E. coli, three Enterobacter aerogenes, and two Citrobacter freundii). The four ECC isolates were isolates with a different PFGE-type isolated from patients from whom an ECC was already included in the UMCU-ECC group. Enterobacteriaceae isolated from patients with an ECC during the outbreak period (2001)(2002)(2003)(2004)(2005) were E. coli (11), Enterobacter aerogenes (2), and Citrobacter freundii (1). Furthermore, an E. aerogenes (10E013) and C. freundii (10A275) from the previously mentioned European antibiotic resistance surveillance study were included [30]. These last two isolates were previously misidentified as E. cloacae.

HPI Detection
The presence of the irp2, intB, and fyuA genes were detected by PCR in lysates using the primers for irp2: Irp2-F, Irp2-R, fyuA: FyuA-F, FyuA-R, and intB: IntB-F, IntB-R. Characteristics and expected product sizes are described in Table S1. Amplification products were detected on 1.5% agarose gels with 1 mg/mL ethidium bromide visualized under UV light.

Analysis of the High Pathogenicity Island
The contig encoding the HPI-ICEEh1 was screened for the presence of attO repeats. To investigate excision and circularization of the HPI-ICEEh1, primers were designed for sites flanking all attO repeats (for primers see Table S1; the schematic location of the primers is depicted in Figure 1). Subsequently, PCR and nested PCR were performed on the possible remaining junctions (reformed attO sites) in the chromosome as well as on the extra chromosomal elements formed by the HPI-ICEEh1. Obtained products were sequenced after Qiagen Quick purification (Qiagen, Westburg b.v., Leusden, The Netherlands) using the BigDye Terminator v1.3 Cycle Sequencing Ready Reaction Kit and a 3100 capillary DNA sequencer (Applied Biosystems, Nieuwerkerk a/d IJssel, The Netherlands).

Characterization of Genomic Islands Integrated at the Integration Hotspot
To determine whether it was likely that other isolates contained a similar (putative) combination of genomic islands and genomic modules as the EHOS, several PCRs specific for the different genomic islands were performed. Primers for ICE typing were based on 17 genes described in the HPI-ICEEc1 present in E. coli ECOR31 (GenBank Accession No. AY233333, Tables S1). An additional PCR was performed for the detection of tnpA with primers designed based on sites adjacent to the tnpA gene, as tnpA was absent in HPI-ICEKp1. Primers to type the putative genomic islands and genomic modules were based on the sequences of strain 05-545 (for primers see Table S1; the schematic location of the primers is depicted in Figure 1). To determine if the genomic islands were located adjacent to each other, linkage PCR reactions were set up using primers directed outwards from the different genomic islands and genomic modules in the IHS (Table S1, Figure 1).

Characterization of Transferability of the HPI
To determine the genetic relationships between the conserved part of the HPI from different isolates, we extracted from GenBank all sequences with the conserved part of the HPI (the region from the intB gene to fyuA in Figure 1). Subsequently, with the 22 sequences (nine E. coli, two K. pneumonia, one Citrobacter koseri, two Yersinia pseudotuberculosis, and eight Yersinia pestis) and the sequence of the HPI from EHOS 05-545, a phylogenetic tree was constructed using ClonalFrame. ClonalFrame determines the genetic relationships of bacteria based on point mutations and homologous recombination [31]. The burn-in length was 50,000 iterations and Markov chain Monte Carlo iterations were set at 50,000. After every 100 iterations, a posterior sample was recorded. Finally, a majority-ruled consensus tree was generated from all posterior samples. Subsequently, the output was used to construct trees with SplitsTree4 (http://www.splitstree.org) [32]. Subsequently, congruence with the intB results were analyzed.

Characterization of intB Genes
To obtain additional evidence on whether different isolates carried identical HPIs, the intB genes of 40 irp2-positive isolates were sequenced. The intB gene of isolate Q2447 was not sequenced because PCR amplification failed. Additionally, the intB genes of three EHOS isolates were not sequenced because 11 intB gene sequences were already determined from the EHOS isolates. The complete gene was amplified by two sets of specific primers. (See primers sets used for amplification and sequencing the intB gene of HPI (HPI-ICEEh1) in Table S1, and Figure 2). Sequencing of purified PCR products was performed as described above. The resulting data were combined with 22 complete intB sequences present in GenBank. To test whether the phylogenetic information was biased by selection pressure, neutrality tests for selection (Tajima's D and Fu's F test) were performed using DNAsp, version 4.20.2 [33][34][35]. Subsequently, a phylogenetic tree was constructed using a Neighbor-Joining algorithm and 10,000 bootstrap iterations with Mega4.0 [36].
Furthermore, to put our data in a larger context, we constructed a phylogenetic tree of a 792 bp fragment of intB, previously also used for phylogenetic analysis by Schubert et al., and we combined the intB sequences from that study with our results.

Yersiniabactin Green Fluorescent Protein (GFP) Reporter Assay
To test the expression of the iron uptake system of the HPI, yersiniabactin production was tested using a GFP-reporter assay [37][38][39]. The presence of yersiniabactin in the medium can be chromosomal DNA of E. hormaechei 05-545; red: asn tRNA; turquoise: direct repeat attO; green: genetic island EhGI1; yellow: the conserved region of HPI-ICEEh1encoding the integrase and yersiniabactin production, regulation and uptake; orange: the putative integrative and conjugative element of HPI-ICE-Eh1; blue: genetic module 3 (EhGM3); purple: genetic module 4 (EhGM4); brown: genetic module 5 (EhGM5). In Box 1-5 the red arrows indicate primer positions for the linkage PCR to determine whether genomic islands are located next to each other and in which direction (see Table 3 (and Table S2) for results and Table S1 for primer characteristics). Red numbered boxes indicate positions of the amplified products by PCR for IHS characterization with primers as described in Table S1. doi:10.1371/journal.pone.0008662.g001 determined by increased production of GFP, because 267 amino acids of FyuA are fused with GFP and FyuA is upregulated in the presence of yersiniabactin. The knockout strain Yersinia enterocolitica WA-CS irp1::Kan r containing the pCJG3.3N plasmid, kindly provided by Prof. J. Heesemann and Dr. S. Schubert, was used as the reporter strain to detect yersiniabactin in supernatants.
Tested isolates were cultured for seven days at 37uC under continuous shaking at 150 rpm in either 10 ml Nutrient Broth (NB), Difco nutrient broth (Becton Dickinson, Sparks, MD, USA) with 85.6 mmol NaCl or 10 ml NB with 200 mM a,a'-dipyridyl (NBD). Filter-sterilized supernatant from a culture (450 ml) was added to 50 ml of a culture of the reporter strain WA-CS irp1::Kan r containing the pCJG3.3N plasmid. Moreover, this culture was grown overnight at 28uC in 10 ml NB with 200 mM a,a'-dipyridyl (NBD). Subsequently, the culture was centrifuged and diluted to an optical density of 0.1 at 660 nm. When the added supernatant contains yersiniabactin it will upregulate GFP production because the fyuA part fused to the gfp-gene is activated by yersiniabactin. The cultures with the GFP-strain were incubated at 28uC overnight, washed and diluted in phosphate buffered saline (PBS). The bacterial associated fluorescence and scatter data for 50,000 gated bacteria were measured using a flow cytometer (FACSCalibur, Becton Dickinson, Franklin Lanes, NJ, USA), and the mean fluorescence was determined.

HMWP1 and HMWP 2 Expression
To test the functionality of the HPI iron uptake system, the expression of HMWP1 and HMWP2 were monitored under various conditions using SDS-PAGE. Bacteria were grown in M9 minimal medium containing 60 mM Na 2 HPO 4 , 22 mM KH 2 PO 4 , 8.6 mM NaCl, 2 mM MgSO 4 , 0.1 mM CaCl 2 , 2 g/L glucose, and 5 g/L casamino acids (pH 7.4) (M9) and in irondepleted M9. Iron was depleted with 1% Chelex-100 for 48 h at 37uC followed by the addition of 0.5 mM a,a'-dipyridyl (Sigma-Aldrich). Proteins were extracted using sonication followed by centrifugation at 50,0006g for 1 h at 4uC. The pellet was incubated for 1 h in 2 mL 1% SDS and then centrifuged at 50,0006g for 1 h at 4uC [40]. Subsequently, the supernatant was concentrated using a Centricon YM-100 centrifugal filter (Millipore, Billerica, MA, US). Proteins were separated using 7.5% SDS-PAGE at 40 mA for 1 h, as described by Laemmli et al. [41]. The gel was stained with Coomassie Blue R. After staining, HMWP1 and HMWP2 were identified by their exceptionally large size, 350 and 230 kDa, respectively. Next, to confirm the identity of HMWP2, the proteins in the gel were electro-blotted onto an Immobilon-p Transfer Membrane (Millipore). The membrane was stained with Coomassie Blue R, and the suspected HMWP2 band was cut out and sequenced using Edman degradation (the Sequence Center, Utrecht, The Netherlands).

Analysis of the HPI Region
HPI-ICEEh1 was located on one contig (accession number: FN297818). Six attO sites were found in this contig ( Table 1). The six direct repeats are an indication that this region of 127 kb putatively contains two genomic islands and three genomic modules ( Figure 1). The first genomic module is likely a genomic island because it contains all the structural features of a genomic island. (i) it is integrated at a tRNA gene (asn tRNA); (ii) it carries a gene for a phage type integrase; (iii) its GC-content (47.9%) is different from that of the genome (55.4%); (iv) it is flanked by 17bp perfect direct repeats; and (v) it contains genes that possibly increase the fitness of the bacteria. This first putative genomic island of 34.2 kb (Enterobacter hormaechei Genomic Island 1 or EhGI1) comprises 28 open reading frames (orfs) that showed similarity with a comparable region in C. koseri ATCC BAA-895 and partly with a DNA region in E. coli (ECOR31, accession no. EU681267). Only for one orf that encodes a putative truncated intB2 integrase is the function clear. This intB2 had a sequence similarity of 95% with the intB of HPI-ICEEh1. However, it contains sequence mutations that likely lead to a truncated protein making it unlikely that this protein is functional. The remaining 27 ORFs had low similarities with known sequences. However, based on sequence comparisons, at least a part of these ORFs may encode genes involved in iron transport because a putative Fe 3+ -ABC transporter periplasmic-binding protein, a putative Fe 2+ /Zn 2+ uptake regulation protein, a putative Fe 2+ transport protein and an ATP-binding component of an iron transport protein were identified.
The second genomic module is also a genomic island because it also contains most of the structural features of a genomic island: (i) it carries a gene for a phage type integrase; (ii) it has a GC-content (52.4%) that is different from that of the genome (55.4%); (iii) it is flanked by 17-bp perfect direct repeats; and (iv) it contains genes that increase the virulence of bacteria. Although this genomic island is not directly integrated at a tRNA gene (Figure 1), homologues of this genomic island described in other Enterobacteriaceae are. This second genomic island, or HPI-ICEEh1, is 66.2 kb and consists of two segments. The first segment, a conserved part including intB and the 11 genes of the yersiniabactin iron uptake system, is known as HPI. This conserved sequence is 99% identical to sequenced HPIs from Yersinia species, E. coli isolates and a K. pneumoniae isolate. The second segment contains the variable part of HPI-ICEEh1. This segment is possibly involved in transfer of HPI-ICEEh1 and is 99% identical with that of HPI-ICEEc1. Starting from the fyuA gene, the first difference with this ICE is a hypothetical protein, which is likely to be transcribed in E. hormaechei 05-545 but not in the ECOR31 because of a 1-bp deletion. A second 1-bp deletion in ECOR31 compared to 05-545 leads to the lack of an ORF in ECOR31 that is present in the EHOS (ORF60). Furthermore, the insertion element IS630 found in ICEEc1 encoding the transposase TnpA is lacking in the ICE of HPI-ICEEh1. The ICE element is necessary for the conjugation and integration of HPI [7]. In addition, ORF21a in ECOR31 is 33 amino acids smaller than its counterpart in the EHOS. Finally, ORF23 and ORF28 differ in their N-terminal regions compared to their counterparts in EHOS, ORF72 and ORF77, respectively.
The third part, EhGM3, is a genomic module and possibly a genomic island. The GC-content of EhGM3 (48.0%) is different from the genome. The module is flanked by 17-bp perfect direct repeats and the 39-end is flanked by an asn tRNA. Although this genomic module does not contain a functional or cryptic gene encoding an integrase, it does contain an ORF that encodes a putative holin protein related to that of prophage CP-933X (51% amino acid similarity) and another ORF that encodes Tsx, a nucleoside-specific channel-forming membrane protein that also functions as a receptor for phages. Two transcriptional regulators are located in this region: a putative LysR-type and a nitrogen assimilation transcriptional regulator. EhGM3 is 10.9 kb and contains in total eight ORFs.
The fourth genomic module, EhGM4, is 13.7 kb and contains nine orfs, and it has a GC-content (56.3%) that is slightly different from that of the genome. It is flanked by perfect direct repeats (one orientated in reverse orientation) and is integrated between two asn tRNAs, of which one in the reverse orientation. Furthermore, it has no integrases or genes involved in conjugation. Most of the genes are associated with transport, including genes that encode for putative AcrA-and AcrB-like proteins. The Resistance Nodulation cell Division (RND) familytype transporter AcrB and the periplasmic accessory protein AcrA, together with the outer membrane factor TolC, form the tripartite efflux pump.
The fifth genomic module (EhGM5) is 1.9 kb and contains a gene encoding a putative MATE family transport protein. The GC-content of this module is 54.2%; because of the variability of the GC-content in the genome and because of the small size of the module, this is not significantly different. However, this region is flanked by perfect direct repeats (one orientated in reverse orientation), and it is integrated between two asn tRNAs, of which one is in the reverse orientation. The genetic modules EhGM3, EhGM4, and EhGM5 may be remnants of genomic islands, but they are now co-transferred by the HPI-ICEEh1.

Analysis of Integration Hotspots
The EhGI1 in isolate 05-545 is integrated at an asn tRNA after the iutA2 gene in the chromosome. By sequencing this region in 12 EHOSs and the two ECCs we showed that the integration site was identical in all these isolates (Figure 3, Accession no. GQ891736-GQ891749). The IHS was located after the same asn tRNA, indicating that the IHS, at least in the EHOS, was always located after iutA2. Excision and Circularization of the Genomic Islands in the HPI Region Because six direct repeats and two integrases (of which intB2 is almost certainly truncated) were identified, the possibility of various excision events and extrachromosomal circularizations in isolate 05-545 was investigated. Four different circular extrachromosomal elements were detected by PCR. These consisted of EhGI1 -HPI-ICEEh1 ( Figure S1), EhGI1 -HPI-ICEEh1 -EhGM3 ( Figure S2), HPI-ICEEh1 ( Figure S3) and HPI-ICEEh1 -EhGM3 ( Figure S4). The attO sequence was present in all these extrachromosomal elements (Accession no: GU086403, FN556610-FN556612). Other combinations were not detected but could not be excluded. The junction in the chromosome formed after excision of EhGI1 and HPI-ICEEh1 could also be detected ( Figure S5A). After this excision, EhGM3, EhGM4 and EhGM5 were still in the chromosome. The junction contained the attO site (FN556613). In the chromosome of an HPI-negative EHOS isolate (03-638), a junction containing the attO site was also detected with EhGM3, EhGM4, and EhGM5 in the chromosome. Furthermore, there was also an indication that the junction containing the attO site and only EhGM4 and EhGM5 were present in the chromosome of EHOS 05-545 and 03-638 ( Figure S5B). However, this was not confirmed by sequencing because of the small amount of products obtained after the amplification reaction.

Presence of the HPI in Other Enterobacteriaceae
PCR amplification of the intB, irp2 and fyuA genes was performed for 137 ECC isolates to determine which gene was the most representative for the presence of the HPI. The isolates were composed of a subset of UMCU-ECCs and included 59 EHOSs and 78 other isolates. All three PCRs were positive for 57 (56 EHOS) isolates, and these were considered HPI-positive. The intB PCR was also positive for six additional isolates, making the intB PCR unreliable for HPI detection. Possibly, these six intBpositive samples contained EhGI1 with the highly similar IntB2 gene or P4 phages that also have an integrase gene similar to intB. From the irp2 and fyuA genes, irp2 was arbitrarily chosen as the marker used to test the remaining isolates for the presence of HPI [42].
Based on the irp2 PCR, the HPI was present in 92% (184/199) of the isolates belonging to the EHOS and 2.1% (11/518) of all other ECC tested isolates. During the outbreak, EHOS isolates with and without the complete HPI-ICEEh1 were retrieved from at least four patients (data not shown), and 8% of all examined EHOS isolates did not contain the HPI, which is an indication that HPI is often lost in EHOS isolates. Six of the 12 HPI-negative EHOS isolates of which the origin of isolation could be determined were isolated from urine samples. Since yersiniabactin is unstable under acidic conditions, excision of the HPI may confer a fitness advantage to the bacteria in acidic environments [43]. The strains involved in the small outbreaks IV, VIII and IX were HPI-negative. The 4.0% (7/176) HPI-positive isolates in the non-EHOS isolates of the UMCU-ECC group was significantly higher (p,0.05) than the 1.2% (4/342) prevalence in the non-UMCU-ECC group. This suggests that in the UMCU during the outbreak, HPI-positive isolates were selected or that horizontal transfer of HPI occurred in vivo in the UMCU during the outbreak period ( Table 2).

Characterization of the ICE Region of the HPI
The presence of genes possibly involved in the transfer of HPI-ICEEh1 were investigated because in previous studies the ICE segment of the HPI-ICE was shown to be necessary for the autonomic integration and circularization of the HPI [7,10]. The presence of the 17 different genes known to be located on HPI-ICEEc1 was assessed in 44 irp2-positive isolates. All 14 irp2-positive non-EHOS isolates were selected, as were 14 EHOS isolates, 11 E. coli, 2 C. freundii, and 3 E. aerogenes. (Table 3 and Table S2). Thirteen irp2-negative ECC isolates, including two HPI-negative EHOS isolates, served as negative controls. All 14 EHOS isolates had an ICE region that was identical to the ICE of E. coli   Table 3. Prevalence, localization and partial characterization of the genomic islands located in the integration hotspot of the EHOS.             , with the exception that the EHOS ICE region lacked the tnpA gene, similar to HPI-ICEKp1 (Table 3 and Table S2). The HPI-ICEs of five other ECC isolates were identical to that of the EHOS. Four of the five were related to the outbreak in the UMCU (03-273, 05-349, 03-635 and 03-739). One isolate (14A001) was obtained in Poland. Of the 11 HPI-negative control isolates, 10 lacked all HPI-ICE genes tested, while one isolate was positive for orf19. Since orf19 was also present in four other isolates (two ECC and two E. coli), we assume that the primers selected to detect orf19 lacked sufficient specificity or that a copy of this gene is present elsewhere in these isolates.

Stability of the Other Genomic Islands and Genomic Modules in the EHOS
To test for the presence of other genomic islands and/or genomic modules in the EHOS, PCRs 1 to 32 and the five linkage PCRs used for the characterization of the IHS (Table S1) were also performed on 16 EHOS isolates (14 HPI-positive and 2 HPInegative, Table 3, Table S2). All 14 irp2-positive EHOS isolates contained (besides the HPI) the genomic island EhGI1 and the three putative genomic modules in the same order. Of the two HPI-negative EHOS isolates tested, one (03-797) harbored the genomic modules EhGM3, EhGM4, and EhGM5. The other HPInegative EHOS (03-638) contains EhGI1 in addition to the genomic modules EhGM3, EhGM4, and EhGM5. EhGI1 was apparently located on another region of the genome because no product was obtained with the linkage PCR from EhGI1 to the expected chromosomal border of EhGI1. However, EhGM3, EhGM4, and EhGM5 were located in the same IHS because with a linkage PCR we could show that EhGM3 was located after the asn tRNA behind the iutA2 gene. This shows that EhGM3, EhGM4, and EhGM5 were integrated at the same IHS. Furthermore, in this isolate a junction containing the attO site with only EhGM4 and EhGM5 was also detected. The two combinations of genomic islands in this isolate indicate the flexibility of the IHS in isolate 03-638.
To determine if the same structure of the genomic islands and genomic modules present in the EHOS were also present in the other 41 isolates, PCRs were performed to characterize the IHS region of the HPI of these isolates ( Table 3, Table S2). Two of the five ECC isolates with the same HPI-ICEEc1 were PCR-positive for the genomic island/module-specific PCRs and linkage PCRs (isolates 03-273 and 05-349). This indicated that in these two isolates, EhGI1 through EhGM5 were present and located in the same order and orientation as in the EHOS. Two isolates with a similar HPI-ICE structure to the EHOS did not contain EhGI1, but contained EhGM3, EhGM4 and EhGM5 (isolates 03-739 and 03-635). One isolate contained EhGM5 next to the HPI-ICE (isolate 14A001). However, linkage PCRs showed that in this 14A001 isolate, EhGM5 and HPI-ICE were apparently not located next to each other. Furthermore, different combinations of the genomic islands/modules were found in HPI-positive and HPI-negative isolates, and these genomic islands/modules were not always located next to each other based on the results of the linkage PCRs. All HPI-positive E. coli and C. freundii tested contained none of the other genomic islands except the HPI. All three E. aerogenes and one ECC isolate contained EhGI1, but it was likely at a different chromosomal location than the HPI because linkage PCRs from EhGI1 to the HPI were all negative.

Phylogenetic Analysis of the Conserved Part of the HPI
To determine the genetic relationships between the HPI-ICE in EHOS 05-545 and 22 other HPI-ICEs in other Enterobacteriaceae, ClonalFrame was used. Therefore, the conserved regions Table 3. Cont. Genes, open reading frames (ORFs) and DNA sequences known to be located on genomic islands of the EHOS. ''+'' represents that a product was amplified and ''2'' represents a negative result for the different DNA fragments investigated (results are depicted in more detail in Table S2). of HPI-ICEs (homologies of sequences bp 42,589-72,874 of FN297818 bp) encoding the integrase and yersiniabactin production, regulation and uptake genes were used. The recombination/ mutation ratio was 0.740 (95% confidence interval 0.414-1.250), indicating that the mutation rate is influencing the evolution of the conserved region of the HPI more than recombination events. The phylogenetic tree based on the conserved region of the HPI-ICEs shows that the conserved region of the HPIs of Y. pseudotuberculosis and Y. pestis isolates are highly related to each other. This indicates that the conserved region of the HPI in Yersinia is highly stable and not subjected to recombination events. These HPIs most probably descend from a single ancestor (Figure 4) The branch with E. coli ED1a contains four isolates that belong to different species. As mentioned, the HPI-ICE of K. pneumoniae has been transferred to other species [7]. Furthermore, we showed that ECOR31 and two ECC isolates also contained a highly similar HPI-ICE. BLAST analysis of the ICE segment of HPI-ICE (the variable region of the HPI) of ECOR31 showed that it is most similar to the HPI-ICE of 05-545. However, E. coli ED1a, C. koseri BAA-895, K. pneumoniae NTUHK2044, and E. coli UMN026 also contain segments highly similar to the integrative and conjugative part of the HPI of ECOR31 and EHOS 05-545, while no other HPI-ICE-containing isolates annotated in GenBank contained genes that may be involved in conjugation and transfer. This result indicates that the branch with multiple species contains mostly HPI-ICEs that may be transferred. This implies that multiple species are able to transfer their HPI.
The intB gene is genetically the most diverse in the analyzed region of the HPI. However, this did not affect the data. When the intB gene was excluded from the HPI analysis, results were similar ( Figure S6A). If only the intB sequence is used for this set of 23 isolates, the results are almost congruent with the results from the conserved region of the HPI ( Figure S6B). Therefore, only the intB gene sequences were used for genetic comparison. Since our aim was to determine if the HPI of the EHOS was closely genetically related to other described HPIs, intB was a reasonable alternative to compare many isolates. The intB genes of 40 irp2-positive isolates were sequenced. Four isolates contained a truncated intB gene. Alignment with complete intB genes demonstrated that in these four isolates, the intB genes lacked base pairs 124 to 470. Therefore, these four intB sequences were excluded from analysis. The data of the remaining 36 intB sequences were combined with 22 sequences of intB present in GenBank to generate a phylogenetic tree ( Figure 5). Neutrality tests (Taijma's D test and Fu's F test) were not significant for positive selection (both p.0.1), indicating a neutral selection process for the intB gene. This indicates that intB sequences are suitable for phylogenetic studies since the intB gene is only affected by random mutations. Three clusters with .90% bootstrap values were evident. One intB sequence of a K. pneumoniae isolate with a low bootstrap value differed from all other isolates and fell outside the three clusters. All Enterobacteriaceae with cluster 3 type intB sequences (three E. coli, six ECC and 11 EHOS, Figure 5) contained a sequence (partly) similar to the HPI-ICEs of EHOS and ECOR31. Furthermore, one ECC isolate with a cluster 1 intB and all three E. aerogenes isolates also had a sequence (partly) similar to the HPI-ICEs of EHOS and ECOR31. The other isolates with a cluster 1 intB did not contain genes of the HPI-ICE.
All eleven EHOS isolates had identical intB sequences, showing that the mutation rate in the intB gene is low enough to study relatedness of HPIs with other isolates. As shown in Figure 5, there are two ECC isolates (03-273 and 05-349) with the same intB gene. Moreover, these isolates also had the genomic island and the three other genetic modules located in the same orientation as the EHOS. These results provide further evidence for combinatorial transfers of the HPI and the other transferable genetic elements. More extensive analysis was possible with a smaller fragment. This previously analyzed 792 bp intB gene fragment was used to compare intB sequences from one C. freundii isolate, two C. koseri isolates, eleven EHOS isolates, three E. aerogenes isolates, twelve ECC isolates (not EHOS), fifty-four E. coli isolates, three K. pneumoniae isolates, two Salmonella enterica isolates, one Shigella flexneri isolates, two Y. enterocolitica isolates, five Y. pestis isolates, and one Y. pseudotuberculosis isolates. The three clusters in the tree based on the complete intB gene were also found in the tree constructed with 97 isolates using only 792 bp of the intB gene ( Figure 6). All isolates used for the analysis of the complete intB gene clustered similarly, except for E. coli 3172/97, which formed a separate branch with a low bootstrap value. K. pneumoniae NTUHK2004 (identical in sequence to K. pneumoniae ICEKp1) formed a separate branch with a low bootstrap value, similarly to the phylogenetic tree based on the complete intB gene ( Figure 5). The intB gene fragment of the EHOS and two other ECC isolates (03-273 and 05-349) were identical with the intB fragment of ECOR31. Furthermore, this cluster 3 also contained four other E. coli isolates, five other ECC isolates and one K. pneumoniae isolate. Two Y. enterocolitica isolates formed a separate branch (cluster 2). The remaining 68 isolates are all located in cluster 1, which contains nine different species.

Expression of the HPI Iron Uptake System
Yersiniabactin production was tested in 41 of the 44 HPIpositive isolates used for ICE characterization with a GFP-reporter assay. Included were twelve ECC (not EHOS) isolates, fourteen EHOS isolates, ten E. coli isolates, three E. aerogenes isolates, and two C. freundii isolates (Table S2). Five HPI-negative ECC isolates were used as negative controls.
All 26 HPI-positive ECCs, including 14 EHOS isolates, two C. freundii isolates, eight E. coli isolates, and one E. aerogenes isolate tested positive for yersiniabactin production. Two E. coli isolates, two E. aerogenes isolates and the negative controls were negative (Table S2).
The expression of HMWP1 and HMWP2 was tested in 14 isolates by SDS-PAGE. Four pairs of EHOS isolates with and without HPI were selected, as were four HPI-positive ECC isolates belonging to different PFGE types and two HPI-positive E. aerogenes strains. All tested HPI-positive isolates were able to produce both HMWP1 and HMWP2, but only under iron-depleted conditions (data not shown). The putative HMWP2 band from one EHOS was confirmed by Edman degradation of the first eight amino acids (data not shown). No HMWP1 or HMWP2 was detected in the HPI-negative isolates or during growth in the presence of iron.
The HMWP2 of the two E. aerogenes isolates, which were negative in the GFP assay, were approximately 10 kDa larger than expected; possibly HMWP2 in these two E. aerogenes was dysfunctional, which may explain the negative GFP-assay test result (data not shown).

Discussion
The results of this study show that the E. hormaechei outbreak strain (EHOS) responsible for a nationwide outbreak in The Netherlands contained an HPI with a functional yersiniabactin-iron-uptake system, enabling the strain to obtain iron in environments with very  limited free iron such as the human body. There was circumstantial evidence that the HPI-ICEEh1 was transferred in vivo. The EHOS contained a new HPI variant, now termed HPI-ICEEh1, with an ICE element highly similar to the ICE element of HPI-ICEEc1 from E. coli ECOR31 [7,10]. The iron-uptake system of HPI-ICEEh1 was expressed, as has been previously described for other HPIs [9,45]. Whether the HPI contributed to the epidemic behavior of the EHOS cannot be concluded from these data. In several Enterobacteriacae it has been shown that HPI-containing isolates are more virulent than isolates lacking this island [7,9,[21][22][23][24]. Possibly, the increased pathogenicity increases the severity of the infections and consequently the bacterial load in patients, which possibly also increases their length of hospital stay. Both of these factors will increase the chance of cross-transmission to other patients or to the environment. Another explanation could be that optimal iron acquisition is essential for colonization of heavily colonized habitats such as the colon [46]. The 'normal' habitat of E. hormaechei is speculative because most often isolates of the E. cloacae complex isolates are not further determined than the level of E. cloacae. Recent studies showed that most infections in humans with ECC are predominantly caused by E. hormaechei [13,47] while the relative presence of E. hormaechei compared to ECC isolates in the gut was low [13]. E. hormaechei has also been isolated from animals and occasionally drinking water reservoirs [47,48]. In addition, E. hormaechei was isolated during selection studies searching for rhamnolipid-producing strains from a biodiesel facility, the search for selenium reducing bacteria from a coal mine tailing pond sediment, and the examination of salinated soil where the rhizosphere of wheat was grown [49][50][51]. This indicates that E. hormaechei is widely distributed in the environment and capable to adapt to different niches.
Analysis of the genome of an EHOS showed that the EHOS contained multiple integrations of mobile elements. Due to the putative presence of two different genomic islands together with the three putative genomic modules and the demonstrated flexibility of excision and putative integration, this part of the E. hormaechei genome may be considered an integration hotspot (IHS). This IHS increases the genomic plasticity of the bacterium. Except for HPI-ICEEh1, the functions of the genes located on the genomic islands in the IHS are largely unknown. EhGI1 may be involved in iron uptake similarly to the HPI of HPI-ICEEh1. Genes located on genomic modules three, four, and five are possibly involved in regulation and transport. Although speculative, the presence of AcrA and AcrB-like protein encoding genes indicate that a tripartite efflux pump is present. In Enterobacteriaceae, these types of efflux pumps extrude cytotoxic substances from the cell directly into the medium, bypassing the periplasm and thereby reducing susceptibility to toxic compounds [52].
One putatively truncated integrase in EhGI1 (intB2) and an integrase in HPI-ICEEh1 (intB) were present and had a sequence similarity of 95%. Most likely, the integrase of HPI-ICEEh1 is functional, because several different combinations of genomic islands and genomic modules were excised from the chromosome and formed circular structures, the integrase of EhGI1 was truncated, and EhGM3, EhGM4, and EhGM5 lack integrase genes. Therefore, we speculate that EhGI1 and the EhGMs are cotransferred by HPI-ICEEh1 or are remnants from previous recombination events. Recent studies showed the transfer of HPI-ICEKp1 to E. coli and K. pneumoniae in a conjugation experiment and possibly in vivo to other K. pneumoniae isolates [7]. Another study showed horizontal transfer of the HPI within E. coli species [11]. Here we provide evidence that multiple combinatorial transfers of the HPI and other genetic modules between ECC isolates most likely occurred. The idea of multiple combinatorial transfers of genomic islands and modules is supported by the following results. Two genomic islands (EhGI1 and HPI-ICEEh1) and three genomic modules (EhGM3, EhGM4, and EhGM5) were detected in two other ECC isolates with different genotypes in the same order. In addition, the intB gene sequences were identical, between the different intB genes of the EHOS and the two ECC isolates that have different genotypes whereas this gene shows variability. If this was not a relatively recent event, it would be expected that mutations in intB would have occurred. Therefore, it is reasonable to presume that transfers of these elements occurred relatively recently, but certainly before the diversification of the ECC species. Finally, genes encoded by the genomic islands and modules were present in different isolates, indicating transfer of those elements. Transfer of the genetic islands and modules was further supported by the finding that structures highly similar to EhGI1 and EhGI2 and partially similar to EhGM3, EhGM4, and EhGM5 were present in other Enterobacter species. Furthermore, there was a significantly higher prevalence (p,0.05) of HPIs in the non-EHOS isolates that were isolated during the EHOS outbreak period in the UMCU in comparison with those not linked to the UMCU EHOS outbreak, which could indicate transfer, although it cannot be excluded that these isolates were selected under the selective pressure in the hospital. All cluster 3 intB-containing isolates (partly) had an ICE segment, while only a fraction of the cluster 1 intB-containing isolates (partly) harbored an ICE segment, which indicates that the ICE is derived from isolates with the cluster 3 intB gene. This conclusion is further supported by the composition of the conserved segments of the HPI in different isolates. Moreover, all isolates with similar ICE elements (except one) clustered in the same branch, which indicates multiple transfer events of the HPI-ICE to other species. Furthermore, the HPI-negative isolates did not contain the ICE segment (or even part of it), supporting the hypothesis that an HPI-ICE is the progenitor of the HPI in Yersinia species [10]. The massive combinatorial transfer of the HPI-ICEEh1 and other genomic modules between Enterobacteriaceae is worrisome. It has been hypothesized that the majority of disease-causing bacteria from the intestine (e.g., Shigella spp. or Yersinia spp.) may have been derived from commensals that have acquired genes from foreign sources turning them into pathogens [53]. Incorporation of the HPI, or other genomic islands or modules, into chromosomes of other isolates may generate new pathogenic strains, which due to their different genetic backgrounds, may emerge in new niches [3]. The uptake of multiple genomic islands and modules and resistance determinants is in agreement with the hypothesis of genetic capitalism, which states that successful integration and selection of a foreign genetic element increases the number of possible genetic transfer events in the future [17]. The acquisition of the HPI, other genomic islands and multidrug resistance genes by the EHOS underscores this hypothesis [17]. Figure S1 Overview of excision and subsequent circularization of EhGI1 and HPI-ICEEh1. A) Schematic presentation of the region in the IHS where excision takes place. Black arrows depict the location and orientation of the primers for the first PCR. Red arrows depict the location of the primers for the nested PCR. B) Schematic presentation of the circular structure formed. C) The circularized fragment generated a product of 2,572 bp. The primers used were ORF77-2F and ORF7-2R (Table S1). The product was analyzed on a 1% agarose gel. D) The circularized fragment generated a product of 1,204 bp from a nested PCR. The product was amplified with the primers ORF77-2F and intB-N1 (Table S1). The product was analyzed on a 1% agarose gel. E) Box W: Schematic presentation of the PCR, nested PCR and sequenced fragment (Accession no: GU086403). Found at: doi:10.1371/journal.pone.0008662.s001 (7.75 MB TIF) Figure S2 Overview of excision and subsequent circularization of EhGI1, HPI-ICEEh1, and EhGM3. A) Schematic presentation of the region in the IHS where excision takes place. Black arrows depict the location and orientation of the primers for the first PCR. Red arrows depict the location of the primers for the nested PCR. B) Schematic presentation of the circular structure formed. C) The circularized fragment generated a product of 2,046 bp. The primers used were ORF85-2F and ORF7-2R (Table S1). the product was analyzed on a 1% agarose gel. D) The circularized fragment generated a product of 335 bp with a nested PCR. The product was amplified with the primers tRNA-N3 and intB-N1 (Table S1). The product was analyzed on a 1% agarose gel. E) Box X: Schematic presentation of the PCR, nested PCR and sequenced fragment. a Sequencing of the first PCR product with primer ORF7-2R shows 100% homology with the sequence of FN297818. b Sequencing of the first PCR product with primer tRNA-N3 (Accession no: FN556610). Found at: doi:10.1371/journal.pone.0008662.s002 (8.80 MB TIF) Figure S3 Overview of circularization of HPI-ICEEh1. A) Schematic presentation of the circular structure of HPI-ICEEh1 that is formed. B) The product amplified from the circularized fragment. This was too small and most likely an artifact, as the expected product size from the fragment was 2,651 bp. The primers used were ORF77-2F and YbtS-R (Table S1). The product was analyzed on a 1% agarose gel. C) The circularized fragment generated a product of 1,205 bp with a nested PCR. The product was amplified with the primers ORF77-2F and intB-N1 (Table S1). The product was analyzed on a 1% agarose gel. D) Box Y: Schematic presentation of the PCR, nested PCR and sequenced fragment, (Accession no: FN556611). Found at: doi:10.1371/journal.pone.0008662.s003 (8.35 MB TIF) Figure S4 Overview of excision and circularization of HPI-ICEEh1 and EhGM3. A) Schematic presentation of the region in the IHS that is excised and subsequently forms a circular structure. Black arrows depict the schematic location and orientation of the primers used for the first PCR. Red arrows depict the location of the primers for the nested PCR. B) Schematic presentation of the circular structure formed. C) The circularized fragment generated a product of 2,225 bp. The primers used were ORF85-2F and YbtS-R (Table S1). The product was analyzed on a 1% agarose gel. D) The circularized fragment generated a product of 335 bp with a nested PCR. The product was amplified with the primers tRNA-N3 and intB-N1 (Table S1). The product was analyzed on a 1% agarose gel. E) Box Z: Schematic presentation of the PCR, nested PCR and sequenced fragment (Accession no: FN556612). Found at: doi:10.1371/journal.pone.0008662.s004 (8.92 MB TIF) Figure S5 Amplified fragments overlapping the junction between the chromosome and the IHS. A) PCR amplification results from EHOS 05-545 with the different primer mixes as presented in panel C. B) PCR amplification results from the HPI-negative isolate EHOS 03-638 with the different primer mixes (panel C). a Non-specific amplified product of the overlapping region between the chromosome and the intB or intB2 gene. b Fragment after excision of EhGI1 and HPI-ICEEh1, showing that EhGM3, EhGM4 and EhGM5 are still part of the chromosome. The junction in 05-545 was sequenced to confirm the results (Accession no: FN556613). c Possible product indicating that occasionally only EhGM4 and EhGM5 are still present in the chromosome.

Table S2
Genes, open reading frames (orfs) and DNA sequences known to be located on genomic islands and modules of the EHOS. ''+'' represents that a product was amplified and ''2'' represents a negative result for the different DNA fragments investigated. a Isolated group, 1 isolates of UMCU-ECC group, 2 isolates non-UMCU-ECC group, 3 isolates of the third group, which are isolates that not fulfill the criteria of the first two groups but are HPI-positive. b Fragment size 1786 bp; orf20 present, c Fragment size 631 bp; orf20 absent. d ECC with the same intB gene as the EHOS and also similar genetic modules located in the same orieantation and sequens as EHOS. e Deletion in intB gene of 347 basepairs. Found at: doi:10.1371/journal.pone.0008662.s008 (0.07 MB XLS)