Enteroinvasive Escherichia coli O96:H19 is an Emergent Biofilm-Forming Pathogen

ABSTRACT

Enteroinvasive Escherichia coli (EIEC) is a diarrheagenic E. coli pathotype carrying a virulence plasmid that encodes a type III secretion system (TTSS) directly implicated in bacterial cell invasion. Since 2012, EIEC serotype O96:H19 has been recognized in Europe, Colombia, and most recently Uruguay. In addition to the invasion phenotype, the strains isolated from Colombian children with moderate-to-severe gastroenteritis had a strong biofilm formation phenotype, and as a result, they are referred to as biofilm-forming enteroinvasive E. coli (BF-EIEC). The objective of this study was to characterize the biofilm formation phenotype of the BF-EIEC O96:H19 strain 52.1 isolated from a child with moderate-to-severe gastroenteritis in Colombia. Random mutagenesis using Tn5 transposons identified 100 mutants unable to form biofilm; 20 of those had mutations within the pgaABCD operon. Site-directed mutagenesis of pgaB and pgaC confirmed the importance of these genes in N-acetylglucosamine-mediated biofilm formation. Both biofilm formation and TTSS-mediated host cell invasion were associated with host cell damage on the basis of cytotoxic assays comparing the wild type, invasion gene mutants, and biofilm formation mutants. Multilocus sequence typing-based phylogenetic analysis showed that BF-EIEC strain 52.1 does not cluster with classic EIEC serotype strains. Instead, BF-EIEC strain 52.1 clusters with EIEC serotype O96:H19 strains described in Europe and Uruguay. In conclusion, BF-EIEC O96:H19, an emerging pathogen associated with moderate-to-severe acute gastroenteritis in children under 5 years of age in Colombia, invades cells and has a strong biofilm formation capability. Both phenotypes are independently associated with in vitro cell cytotoxicity, and they may explain, at least in part, the higher disease severity reported in Europe and Latin America.

IMPORTANCE Enteroinvasive Escherichia coli (EIEC), a close relative of Shigella, is implicated in dysenteric diarrhea. EIEC pathogenicity involves cell invasion mediated by effector proteins delivered by a type III secretion system (TTSS) that disrupt the cell cytoskeleton. These proteins and the VirF global regulator are encoded by a large (>200 kb) invasion plasmid (pINV). This study reports an emergent EIEC possessing a cell invasion phenotype and a strong polysaccharide matrix-mediated biofilm formation phenotype. Both phenotypes contribute to host cell cytotoxicity in vitro and may contribute to the severe disease reported among children and adults in Europe and Latin America.

INTRODUCTION

Enteroinvasive Escherichia coli (EIEC) (1, 2), a close relative of Shigella, is one of the pathogens implicated in bacillary dysentery, characterized by abdominal pain, fever, and bloody mucoid diarrhea (3). The pathogenicity and invasive capabilities of both bacteria are enabled by a type III secretion system (TTSS), effector proteins, and the VirF global regulator, among others, all encoded by a large (>200 kb) invasion plasmid (pINV) (4–6). The TTSS forms an injectisome assembly that introduces effector proteins into host cells, inducing EIEC internalization and actin polymerization-mediated cell-to-cell spread (7). Studies conducted since 2012 in Italy, the United Kingdom, Spain, Sweden, Uruguay, and Colombia recognized EIEC O96:H19 as a cause of moderate-to-severe diarrhea (8–11). The serotype O96:H19 was not associated with EIEC pathotypes before 2012, and it is believed that these EIEC strains are the result of recent acquisition of the invasive plasmid by a commensal E. coli O96:H19 through plasmid conjugation (12, 13). An intriguing feature of these emergent EIEC O96:H19 strains is the ability to cause severe outbreaks and moderate-to-severe diarrhea (14). No information is available on the possible mechanisms of pathogenesis, in addition to invasion, that may explain the disease severity caused by these emergent EIEC strains or the possible improved spread potential resulting in outbreaks in multiple countries in two noncontiguous continents. It is unclear currently whether additional virulence genes are present in the chromosome or other episomal elements present in these emergent EIEC O96:H19 strains. Enhanced virulence among emerging diarrheagenic E. coli strains, because of horizontal transfer of novel virulence factors, is well established among unique organisms, including the enteroaggregative Shiga toxin-producing E. coli (EAEC-STEC) O104:H4 strain implicated in a severe outbreak of dysenteric diarrhea and hemolytic uremic syndrome in Germany in 2011. Genomic and epidemiological studies concluded that an EAEC O104:H4 strain horizontally acquired a Shiga toxin gene-containing phage, resulting in a pathogen with high Shiga toxin production and a strong biofilm formation phenotype (15, 16).

Emerging EIEC clinical isolates with the rare O96:H19 serotype were detected in two children with gastroenteritis in Colombia, South America, who participated in a case-control study of moderate-to-severe acute gastroenteritis conducted in 2012 (14). These strains were considered emergent because they expressed the unusual serotype O96:H19 not reported in association with EIEC strains before 2012. In addition, these EIEC isolates expressed a biofilm formation phenotype not previously reported among EIEC strains, which is why we designated them biofilm-forming EIEC (BF-EIEC).

The main goals of this study were to characterize the BF-EIEC O96:H19 strains isolated from two pediatric cases of moderate-to-severe gastroenteritis in Colombia, to describe the biofilm formation and invasion phenotypes of this strain, and to characterize key genes implicated in these phenotypes. Biological biofilm-forming and invasion assays, confocal microscopy, electron microscopy, and molecular biology techniques were conducted to evaluate cell invasion, biofilm formation, and cytotoxicity.

RESULTS

Cases of moderate-to-severe diarrhea associated with BF-EIEC.

In 2013, two unrelated mestizo girls, 2 years of age each, were enrolled in a case-control study of moderate-to-severe gastroenteritis in Bucaramanga, Colombia. They both had diarrhea, signs of dehydration, and abdominal pain. They received medical attention at an emergency center. Both girls were diagnosed with acute gastroenteritis on the basis of their clinical presentation and received treatment based on local guidelines. Research testing on their stools was not used to guide therapy, as complete testing results were not available at the time of their visits. Both girls recovered at home and did not require hospital admission. Stools from both girls had E. coli strains that were positive for virF and ipaH genes by PCR and were able to form biofilms. These strains were designated biofilm-forming EIEC (BF-EIEC). The child positive for BF-EIEC strain 52.1 had a fever, and her stools were positive for blood, mucus, and white blood cells, which was consistent with dysenteric diarrhea. In addition, her stool was also positive for norovirus G1. The girl positive for BF-EIEC strain 96.2 had vomiting in addition to diarrhea, and she required intravenous fluids for hydration. Her stools were positive for white blood cells and for rotavirus. BF-EIEC strains 52.1 and 96.2 were serotype O96:H19, and they were PCR positive for EIEC-defining genes, including virF, ipaH, ipaD, mxiD, and mxiH. The last three genes encode TTSS structural proteins.

BF-EIEC invades host cells.

As a proof of principle, the BF-EIEC 52.1 wild-type strain and TTSS site-directed mutants were evaluated for cell invasion by using gentamicin invasion assays. The BF-EIEC 52.1 wild-type strain invaded and survived inside HeLa cells after 1 h of exposure to gentamicin (Fig. 1A). Although there was high run-to-run variability in the quantitative invasion assay, the level of host cell invasion was comparable to that for Salmonella enterica, used as the positive control for cell invasion. In contrast, the E. coli nonpathogenic control strain DH5α failed to invade host cells. Site-directed lambda Red mutagenesis of the TTSS structural component mxiH and mxiD genes, as well as the gene for an effector protein (ipaC), resulted in diminished invasiveness but retained the biofilm phenotype (Fig. 2A and B).

FIG 1.

BF-EIEC 52.1 wild-type strain invades and adheres to HeLa cells. (A) Quantitative invasion assay with HeLa cells. HeLa cells were infected with the BF-EIEC 52.1 strain, an E. coli DH5α strain as a negative control, and a Salmonella enterica wild-type strain as a positive control. Data from 13 experiments are included in the figure. Three experiments directly compared BF-EIEC 52.1 and S. enterica. Representative images from a qualitative adherence assay using infected HeLa cells stained with Giemsa and the EIEC-EC12 wild-type strain (B), BF-EIEC 52.1 wild-type strain (C), and EAEC JM221 wild-type strain (D).

FIG 2.

Effects of invasion gene mutations in BF-EIEC 52.1 strain on cell invasion and biofilm formation. Invasion (A) and biofilm production (B) characteristics were compared between the wild-type BF-EIEC, mutants of the type III secretion system (TTSS) genes mxiD, mxiH, complemented ΔmxiD(pmxiD) and ΔmxiH(pmxiH) mutants, and the DH5α control strain. Error bars represent 95% confidence intervals from the means.

BF-EIEC exhibits strong biofilm-forming properties.

The EIEC-EC12 strain used as a control did not adhere to surfaces or host cells to any significant degree (Fig. 1B). However, the BF-EIEC strain exhibited very strong adhesion to glass and plastic surfaces (Fig. 1C) in a pattern comparable to that of enteroaggregative E. coli (EAEC JM221) (Fig. 1D). Confocal microscopy revealed vigorous biofilm production by the BF-EIEC 52.1 wild-type strain (Fig. 3A). The high biofilm formation level of the BF-EIEC 52.1 wild-type strain was similar to that of the EAEC JM221 positive control strain per quantitative biofilm assay data (Fig. 3B and C). Live-cell staining and polysaccharide staining revealed a heavy BF-EIEC biofilm consisting mostly of live cells with an abundance of extracellular polysaccharide matrix (Fig. 3A).

FIG 3.

Quantitative biofilm assays for the BF-EIEC 52.1 wild-type strain, E. coli DH5α laboratory control strain, and enteroaggregative E. coli (EAEC) JM221 strain. (A) Confocal microscopy of biofilm formed by the BF-EIEC-52.1 wild-type strain by using live cell (SYTO 9) staining and polysaccharide (calcofluor white) staining and imaged at ×600 magnification. (B) Biofilm formation measured according to the optical density of stained microplate wells. Mean biofilm production for each strain is the inverse-variance-weighted mean from five experiments. There were four direct comparisons among the BF-EIEC 52.1, E. coli DH5α, and EAEC JM221 strains. Final P values were calculated using Fisher’s combined probability test. (C) Photograph of microplate wells with biofilm formed by the BF-EIEC 52.1 wild-type strain, E. coli DH5α laboratory control strain, and EAEC JM221 strain.

Biofilm formation is exclusive to BF-EIEC and not other EIEC or Shigella strains.

The quantitative biofilm assay was also performed on several listed EIEC and Shigella clinical isolates (Fig. 4). Results from these assays show that BF-EIEC produces as much biofilm as EAEC. Shigella sp. and EIEC strains did not produce a significant level of biofilm, with a threshold optical density at 590 nm (OD₅₉₀) of 0.03, defined as three standard deviations below the mean from the E. coli DH5α negative control.

FIG 4.

Quantitative biofilm formation assay comparing EIEC, Shigella species, and BF-EIEC strains as well as a JM221 EAEC positive control (PC) and DH5α negative control (NC). A threshold OD₅₉₀ of 0.291 (three standard deviations below the mean from the DH5α control) was chosen as a cutoff for biofilm production. The EIEC control (C) is a spontaneous biofilm formation-negative mutant derived from the BF-EIEC strain. Error bars represent 95% confidence intervals from the mean.

Quantitative biofilm assay following enzymatic digestion suggests that the polysaccharide matrix of the BF-EIEC biofilm consists primarily of N-acetylglucosamine.

Confocal microscopy with calcofluor white staining revealed blue fluorescence in BF-EIEC and EAEC (JM221) samples, indicating the presence of polysaccharides in the extracellular matrix (Fig. 3A and B). With enzymatic digestion (Fig. 5), cellulase, lysozyme, DNase, RNase, hyaluronidase, and chondroitinase had negligible effects on biofilm produced by either BF-EIEC or EAEC. Among the strongest effects of the enzymes was that of N-acetylglucosaminidase on BF-EIEC biofilm (reduced to approximately 24% of the normalized control value); it had no corresponding effect on EAEC. Conversely, proteinase K had a strong effect on EAEC biofilm (reduced to approximately 22% of the normalized control value) without any appreciable corresponding effect on BF-EIEC (Fig. 5). Chitinase had similar effects on both BF-EIEC and EAEC.

FIG 5.

Biofilm degradation by extracellular enzymes. BF-EIEC and control EAEC JM221 strains were assayed for biofilm production alone as well as in the presence of several enzymes with putative effects on extracellular proteins and polysaccharides; medium with no bacteria was used as a negative control. All samples were normalized to their respective positive controls such that controls have equal means of 1.000.

Tn5 mutagenesis and lambda Red site-directed mutagenesis identified the pgaABCD operon (poly-β-1,6-N-acetylglucosamine) as critical for BF-EIEC biofilm formation.

BF-EIEC was subjected to Tn5 transposition mutagenesis to identify genes involved in biofilm formation. Of approximately 12,000 mutants screened for biofilm formation, 104 were biofilm negative relative to a threshold established by the nonpathogenic laboratory strain DH5α and using wild-type BF-EIEC as a positive control. Selected mutants B42, B46, B3, B5, and B41 had mutations in pgaB (B42 and B46), pgaC (B3 and B5), and pgaD (B41) (see Fig. S1 in the supplemental material). The pga operon encodes polysaccharide synthase (PgaC), an enzyme that polymerizes N-acetyl-d-glucosamine, which is homologous to PgaC of E. coli K-12 and structurally similar to cellulose synthase subunit A (Fig. 6). The pga operon of the BF-EIEC 52.1 strain (Fig. S1) was compared with the whole-genome sequencing of the Italian strain O96:H19 (GenBank accession number CP011416) (1). We found 99.9% identity in the pgaABD operon genes, except in pgaA. The pgaA gene in the genome of the Italian EIEC strain was disrupted (see Table S1 in the supplemental material). The pgaA encoded a protein involved in transport of polysaccharide outside of the bacterial cell.

FIG 6.

pga operon and PgaC enzyme of the BF-EIEC 52.1 strain. (A) Diagram of the pga operon with pgaD, pgaC, pgaB, and pgaA genes. Tn5 transposon insertion sites are indicated by triangles with bars. (B) Diagram of the predicted tertiary structure of the polysaccharide synthase PgaC from BF-EIEC 52.1 obtained using the known cellulose synthase subunit A tertiary structure as a model. Prediction analysis was conducted by Expasy protein structure SWISS-MODEL online software (https://swissmodel.expasy.org/interactive#).

Adherence assays evaluated the ability of the BF-EIEC wild-type strain, the BF-EIEC pgaC mutant, and the BF-EIEC pgaC mutant complemented with plasmid pBAD-pgaABCD to adhere. The results show that the wild type has strong binding to HeLa cells and biofilm formation as does as the pgaC mutant complemented with pBAD-pgaABCD (Fig. 7 and 8). The BF-EIEC TTSS mxiH mutant control also had strong binding to HeLa cells and strong biofilm formation (Fig. 8).

FIG 7.

Biofilm formation by the BF-EIEC 52.1 wild type, BF-EIEC 52.1 ΔpgaC mutant, and BF-EIEC 52.1 ΔpgaC/pBAD-pgaABCD complemented mutant strains. Biofilms from all three strains were examined by confocal microscopy after staining live cells with SYTO 9 or carbohydrates with calcofluor white and imaged at 600× magnification.

FIG 8.

Light microscopy of HeLa cell adherence assays with BF-EIEC wild-type and mutant strains. HeLa cells infected with bacteria for 2 h were subsequently stained with Giemsa stain as described in Materials and Methods before examination under a light microscope. Panels include uninfected HeLa cells and cells infected with BF-EIEC 52.1 wild-type strain (WT), BF-EIEC 52.1 ΔpgaC mutant strain, BF-EIEC 52.1 ΔmxiH mutant strain, and BF-EIEC 52.1 ΔpgaC/pBAD-pgaABCD complemented mutant strain.

Using scanning electron microscopy (SEM), wild-type BF-EIEC showed granular extracellular material, which was absent in biofilm-negative mutants (Fig. 9). When the ΔpgaC mutant was transformed with the pgaABCD operon-expressing plasmid (pBAD18-pgaABCD), the biofilms were recovered along with the granular extracellular material.

FIG 9.

Scanning electron microscopy. From left to right, the images show wild-type BF-EIEC, BF-EIEC with site-directed mutagenesis (lambda Red) of pgaC, and the same mutant complemented with pBAD18-pgaABCD.

Lambda Red site-directed mutagenesis was performed on the pgaC gene. This isogenic ΔpgaC mutant exhibited complete obliteration of carbohydrate-based biofilm production (Fig. 7). Carbohydrate-based biofilm formation was recovered by complementation with the pBAD-pgaABCD plasmid in the lambda Red ΔpgaC mutant. The PgaC protein of BF-EIEC is a homologue of E. coli K-12 PgaC, a polysaccharide synthase that polymerizes N-acetylglucosamine. The predicted tridimensional structure of PgaC was obtained by the protein structure prediction program SWISS-MODEL using cellulose synthase subunit B as a model (Fig. 6B).

pgaABCD-mediated biofilm formation and TTSS contribute to host cell damage.

Lactate dehydrogenase (LDH) release was assayed in the wild type (BF-EIEC), ΔpgaC and ΔmxiH (a TTSS mutant) mutant strains, and the ΔpgaC strain complemented with the rescue plasmid vector pBAD18-pgaABCD, with LDH acting as a surrogate marker for cytotoxicity. Both mutant strains demonstrated significant reductions in cytotoxicity (Fig. 10). The ΔpgaC mutant exhibited a remarkable 93% reduction in cytotoxicity versus that of the wild type, and the ΔmxiH mutant showed a 71% reduction in cytotoxicity. Complementation of the ΔpgaC mutants with the rescue plasmid vector restored cytotoxicity to a level indistinguishable from that of the wild type.

FIG 10.

LDH release from HeLa cells infected with BF-EIEC and mutant strains to demonstrate cytotoxicity toward host cells. The data for the ΔpgaC/pBAD18-pgaABCD complemented mutant are representative of three isolates that produced similar results. (A) Time course of LDH release, showing measurements taken at 2, 4, 6, 8, and 24 h following inoculation. Error bars represent standard errors of the means (SEMs). (B) Data for the 4-h time point included in the chart in panel A demonstrating the relative cytotoxicity in each sample as a percentage of the detergent-positive control. Error bars represent SEMs.

BF-EIEC belongs to a rare serotype and MLST cluster.

BF-EIEC is O96:H19, a rare serotype. Phylogenetic typing by PCR classifies this strain as B1. Multilocus sequence typing (MLST) further classifies this strain as sequence type 99 (ST99). The phylogenetic tree (Fig. 11) was constructed using the BF-EIEC 52.1 strain, laboratory E. coli control strains, commensal E. coli, several other EIEC strains, and several E. coli strains that have identical STs. Some ST99 strains had differing serotypes.

FIG 11.

MLST phylogenetic grouping of selected EIEC strains. According to the MLST process described in Materials and Methods, BF-EIEC is displayed (arrow) among other EIEC strains in a phylogenetic tree. Scale bar represents nucleotide substitutions per site.

DISCUSSION

This study characterized an isolate of BF-EIEC serotype O96:H19, an emergent E. coli pathotype associated with moderate-to-severe diarrhea in children in Colombia and Uruguay and in children and adults in several European countries, including Italy, United Kingdom, Spain, and Sweden (8–12, 17). We found that the Colombian BF-EIEC strain invades epithelial cells and forms biofilm, a phenotype combination that may explain, at least in part, the hypervirulent phenotype in the human host.

The foodborne outbreak of EIEC O96:H19 that occurred at a fire brigade station in Milan, Italy, in 2012 was the first reported observation of an O96:H19 serotype combined with an EIEC pathotype. In a case-control study conducted in Bucaramanga, Colombia, between 2012 and 2014 (14), there were two cases of moderate-to-severe diarrhea caused by BF-EIEC O96:H19 sequence type 99. Whole-genome sequencing of the BF-EIEC 52.1 strain (unpublished data) shows it is virtually identical to the strain isolated in Italy (18). EIEC O96:H19 was also reported as a rare agent of diarrheal disease among children in rural Uruguay in a study conducted from 2012 to 2015 (11). An MLST analysis demonstrated that the strain does not cluster with other EIECs isolated prior to 2012, which is compatible with prior phylogenetics work performed on whole E. coli and Shigella genomes (1).

BF-EIEC 52.1 is capable of invading HeLa cells by a mechanism mediated by the type III secretion system and associated effectors. The involvement of mxiD and mxiH in invasion was confirmed by site-directed mutagenesis of these genes, which eliminated the invasive phenotype. BF-EIEC 52.1 ΔmxiD and ΔmxiH mutants were still able to form biofilm yet to a lesser degree. A similar phenomenon was reported previously with the plant pathogen Dickeya dadantii, a close relative of Salmonella enterica (19, 20) and E. coli (21). Biofilm production was reduced in TTSS mutants because the bacteria were unable to secrete the required factors (e.g., HrpN).

Biofilm production is a dominant phenotype of BF-EIEC but an unusual phenotype among non-O96:H19 EIEC or Shigella strains. Conversely, pathogenic E. coli strains that produce biofilm (EAEC) generally have a distinct biofilm composition. This was confirmed by enzymatic digestion experiments, which showed that EAEC biofilm is primarily proteinaceous in nature, whereas the biofilm produced by the BF-EIEC 52.1 strain was composed of exopolysaccharides, primarily poly-β-1,6-N-acetylglucosamine (PGA). The BF-EIEC 52.1 strain produced at least as much biofilm as the reference EAEC strain JM221. Commensal E. coli produces an exopolysaccharide-associated biofilm matrix mediated by the pgaABCD operon (22–25). Mutations in the pgaABCD operon of the BF-EIEC 52.1 strain abolished biofilm formation but did not affect the invasion of epithelial cells. Instead, mutations in the TTSS decreased the invasion capacity of BF-EIEC but had a minimal effect on biofilm production. Studies conducted in E. coli K-12 showed that PgaC and PgaD are cytoplasmic membrane proteins that are required for PGA synthesis. PGA is modified by the lipoprotein PgaB facilitating its export through the outer membrane by the PgaA porin (26). The role of biofilm formation in the pathogenesis of BF-EIEC is still unknown; however, biofilm-forming E. coli is implicated in protection from adverse environmental conditions that may facilitate infection and adherence to biological and nonbiological surfaces (22, 27). Biofilm on intestinal surfaces may increase colonization, bring bacteria closer to intestinal cell surface, and facilitate cell invasion. In addition, the pgaABCD operon-mediated biofilm formation may confer E. coli the ability to develop resistance to multiple antibiotics (28). The pgaC gene in the pga operon of Klebsiella pneumonia contributes to biofilm formation and virulence (29).

The pga operon DNA sequence of the BF-EIEC 52.1 strain was identified in the Italian EIEC O96:H19 strain with a high identity level among all genes except the pgaA gene, which was disrupted. The pgaA gene encodes a porin that allows the transport of PGA outside the cell. This difference in pgaA suggests that the Colombian BF-EIEC 52.1 strain has a fully functional biofilm formation phenotype compared to that of the Italian strain.

The combination of host cell invasion and biofilm production is directly implicated in the cytotoxicity of BF-EIEC. The BF-EIEC 52.1 ΔpgaC biofilm formation mutant and the BF-EIEC 52.1 ΔmxiH invasion mutant had 93% and 73% reductions in cytotoxicity, respectively, as measured by the LDH release assay. Virulence enhancement by biofilm formation is reminiscent of a foodborne outbreak of enteroaggregative Shiga toxin-producing E. coli O104:H4 in Germany in 2011 (15). In that outbreak, it was theorized that an EAEC of human origin acquired the Shiga toxin-2 gene stx₂ via bacteriophage transduction. The strain was especially virulent; by one measure, the rate of hemolytic uremic syndrome was 20% among cases versus the expected 6% incidence in prototypical STEC O157:H7 infection. In addition to biofilm formation and TTSS-mediated cell invasion, additional virulence factors may be expressed by BF-EIEC O59:H19 strains. Evidence for additional biofilm-associated genes, including virulence genes, is under investigation. Characterization of all 100 biofilm formation-defective Tn5 mutations in the BF-EIEC 52.1 strain may reveal the nature of genes, other than the pgaABCD genes, and the role they play in biofilm formation and pathogenesis.

The two children reported in this study with moderate to severe diarrhea were positive for BF-EIEC 52.1 and norovirus G1 or BF-EIEC 96.2 and rotavirus. The coinfection of BF-EIEC with an enteric pathogenic virus raises the question of whether these coinfections may lead to a more severe disease process. EIEC and rotavirus coinfection was reported to be associated with a life-threatening severe case of dysenteric and dehydrating diarrhea (30). Similarly, the coinfection of BF-EIEC and norovirus may lead to more severe diarrheal manifestations than individual pathogens alone. Independent studies conducted in children from Ghana or Tanzania revealed that coinfections of rotavirus or norovirus with bacterial pathogens including Shigella sp. and EIEC were associated with diarrhea (31, 32). Although the mechanism by which BF-EIEC and viral coinfections leads to more severe diarrheal disease is uncertain, it is likely that the combinatorial effect of BF-EIEC-associated biofilm formation and host cell invasion plus rotavirus- or norovirus-mediated intestinal host cell lysis may be contributing factors. Biofilm-forming bacterial pathogens may have an advantage against the host if coinfected with pathogenic virus. For instance, respiratory syncytial virus (RSV) respiratory infections may promote nutrient transfer to biofilm-forming Pseudomonas aeruginosa by increasing extracellular vesicle secretion from airway host cells promoting bacterial growth and biofilm formation (33). Similarly, viral infections may contribute to worsening bacterial infections as reported in Chinese poultry farms, where Tembusu virus escalated incidence of biofilm-forming pathogenic E. coli infections resulting in high poultry mortality (34).

E. coli phylogroup ST99, many members of which are serotype O96:H19, has been reported in healthy mammalian species, including cattle, primate, and dolphin (MLST data set) (35). These strains may thus exist as commensal strains in mammalian hosts and the environment. The acquisition of an invasion plasmid from Shigella or another EIEC by horizontal transfer likely led to the emergence of this BF-EIEC with its distinctive invasive and biofilm-forming phenotype.

In summary, the BF-EIEC O96:H19 emerging pathogen, isolated from children with moderate-to-severe acute gastroenteritis in Colombia, has a virulence genotype identical to that of counterparts isolated from the high-virulence foodborne outbreak in Italy and from several sporadic cases of dysentery in Europe (United Kingdom and Spain) and Latin America (Uruguay). These findings demonstrate that BF-EIEC possesses not only an effective cell invasion phenotype mediated by the virulence plasmid-encoded TTSS but also a strong biofilm formation phenotype, both of which contribute to increased cytotoxicity. These characteristics may translate to a hypervirulent phenotype, demonstrated in severe dysenteric diarrheal outbreaks. More studies, including whole-genome sequencing analysis, are necessary to determine the origin, the diversity, and the virulence potential of this emerging pathogen. Worldwide surveillance studies are also critical to understand the epidemiology of this emergent pathogen, which has already spread to two continents.

MATERIALS AND METHODS

Bacterial strains used, growth conditions, and reagents.

Biofilm-forming enteroinvasive E. coli (BF-EIEC) 52.1 and 96.2 strains were isolated from two children with moderate-to-severe diarrhea that were enrolled in a case-control study of childhood gastroenteritis in Bucaramanga, Colombia (14, 36). For this study, we conducted all experiments using BF-EIEC strain 52.1 as the BF-EIEC prototype strain. The characteristics of all bacterial strains used in this study are listed in Table 1. Bacteria were cultured in Luria-Bertani (LB) agar or broth under aerobic conditions at 37°C. Where noted, growth medium was supplemented with 25 μg/mL kanamycin, 30 μg/mL chloramphenicol, 100 μg/mL ampicillin, or 10 mM l-arabinose. All enzymes, reagents, and media were purchased from Sigma-Aldrich Corp. (St. Louis, MO). O antigen was typed using antisera against E. coli O antigens O1 to O187 (except for cancelled groups O31, O47, O67, O72, O94, and O122) (18). H antigen typing was performed by PCR of fliC, followed by restriction fragment length polymorphism (RFLP) analysis (19). All serotyping was conducted by the E. coli Reference Center of Pennsylvania State University.

TABLE 1.

Bacterial strains and their characteristics^a

Strain	Feature(s)	Reference or source
BF-EIEC	Biofilm-forming emergent EIEC strain isolated from patients with diarrhea	14
BF-EIEC(pKD46)	pKD46 plasmid electroporated into BF-EIEC; Ampr	This study
BF-EIEC ΔmxiH	Isogenic mxiH gene deletion mutant	This study
BF-EIEC ΔmxiH(pmxiH)	Isogenic mxiH mutant complemented with wild-type mxiH gene in pBAD18 plasmid	This study
BF-EIEC ΔmxiD	Isogenic mxiD gene deletion mutant	This study
BF-EIEC ΔmxiD(pmxiD)	Isogenic mxiD mutant complemented with wild-type mxiD gene in pBAD18 plasmid	This study
BF-EIEC ΔipaC	Isogenic ipaC gene deletion mutant	This study
BF-EIEC ΔpgaC::cat	pgaC::cat insertion mutant; Chlr	This study
BF-EIEC ΔpgaB::Tn5	pgaB::Tn5 insertion mutant; Kanr	This study
BF-EIEC ΔpgaC::Tn5	pgaC::Tn5 insertion mutant; Kanr	This study
BF-EIEC ΔpgaD::Tn5	pgaD::Tn5 insertion mutant; Kanr	This study
BF-EIEC ΔpgaC::cat/pBAD-pgaABCD	cat insertion pgaC mutant transformed with pBAD18 containing a pgaABCD operon cassette; Chlr, Ampr	This study
BF-EIEC ΔpgaB::Tn5/pBAD-pgaABCD	Tn5 insertion pgaB mutant transformed with pBAD18 containing a pgaABCD operon cassette; Kanr, Ampr	This study
BF-EIEC ΔpgaC::Tn5/pBAD-pgaABCD	Tn5 insertion pgaC mutant transformed with pBAD18 containing a pgaABCD operon cassette; Kanr, Ampr	This study
BF-EIEC ΔpgaD::Tn5/pBAD-pgaABCD	Tn5 insertion pgaD mutant transformed with pBAD18 containing a pgaABCD operon cassette; Kanr, Ampr	This study
EAEC JM221	Clinical isolate	43
EIEC-CR43	Clinical isolate	44
EIEC-CR82	Clinical isolate	44
EIEC-CR121	Clinical isolate	44
EIEC-CR128	Clinical isolate	44
EIEC-CR135	Clinical isolate	44
EIEC-CR144	Clinical isolate	44
EIEC-CR156	Clinical isolate	44
EIEC-Cg414	Clinical isolate	45
EIEC-EC12	Clinical isolate	Gomez lab strain collection
Shigella flexneri	Clinical isolate	Gomez lab strain collection
Shigella boydii	Clinical isolate	Gomez lab strain collection
Shigella sonnei	Clinical isolate	Gomez lab strain collection
Shigella sp. SH11	Clinical isolate	Gomez lab strain collection
Shigella sp. SH16	Clinical isolate	Gomez lab strain collection
Shigella sp. SH18	Clinical isolate	Gomez lab strain collection
E. coli TOP10	Laboratory nonpathogenic strain	Gomez lab strain collection
E. coli DH5α	Laboratory nonpathogenic strain	Gomez lab strain collection

^aChl^r, chloramphenicol resistant; Kan^r, kanamycin resistant; Amp^r, ampicillin resistant; BF-EIEC, biofilm-forming enteroinvasive E. coli; EAEC, enteroaggregative E. coli.

Quantitative biofilm assay.

The microplate biofilm formation assay was performed as described previously (37). Briefly, bacterial cultures were grown overnight in LB broth with aeration at 37°C. On the following day, 5 μL of overnight culture per replicate was transferred to a 96-well plate (Immulon 2 HB; Thermo Fisher Scientific Inc., Waltham, MA), previously loaded with 195 μL/well of Dulbecco’s modified Eagle’s medium (DMEM) with 0.45% dextrose. The plates were incubated under aerobic conditions at 37°C for 18 to 20 h. On the following day, the plates were washed four to five times with water and stained with 0.1% aqueous crystal violet for 15 min. The plates were then washed again, and the remaining stain on the wells was dissolved in 95% ethanol. The resulting solution was transferred to a new plate, and the OD₅₉₀ was measured with a microplate spectrophotometer reader. The threshold to determine if one strain was positive for biofilm was defined as three standard deviations above the mean from the E. coli DH5α negative control.

We investigated the effect of several enzymes on biofilm formation, including 100 U/mL DNase I (bovine pancreas), 20 μg/mL RNase A (bovine pancreas), 20 μg/mL proteinase K (Tritirachium album), 20 μg/mL cellulase (Trichoderma sp.), 0.1 U/mL chitinase (Trichoderma viride), 0.1 U/mL chondroitinase ABC (Proteus vulgaris), 100 μg/mL lysozyme (chicken egg white), 1 U/mL β-N-acetylglucosaminidase (Canavalia ensiformis), and 40 U/mL hyaluronidase (bovine testes). Enzymes were added to the wells immediately before bacterial inoculation.

DNA extraction and virulence gene identification.

Bacterial genomic DNA was extracted from 2-mL overnight cultures using the Gene Elute bacterial DNA kit (Sigma-Aldrich Corp., St. Louis, MO). The extracted DNA was used for the PCR assay containing forward and reverse primers for either ipaH, ipaD, or mxiH (see Table 2 in the supplemental material) and GoTaq G2 Hot Start Green master mix (Promega Corp., Fitchburg, WI) according to the manufacturer’s instructions. All PCR products were loaded onto a 2% Tris-acetate-EDTA (TAE) agarose gel and visualized with SYBR safe stain.

TABLE 2.

Oligonucleotides employed in this study

Primer	Sequence (5′→3′)	Target	Reference or source
ipaH-F	CTCGGCACGTTTTAATAGTCTGG	ipaH	46
ipaH-R	GTGGAGAGCTGAAGTTTCTCTGC
ipaD-F	GATCTCTCCCGGAGGTAACGA	ipaD	36
ipaD-R	CTCGGCACGTTTTAATAGTCTGG
mxiH-F	CTGAAACTTTTGATGATGGAACTCAAACA	mxiH	36
mxiH-R	GATTGCGCGTTCCTATATAATGTATATTCA
pKD3/mxiD-F	ATGAAAAAATTTAATATTAAATCTTTGACTCTCTTGATTGTATTGTGTAGGCTGGAGCTGCTTC	cat	This study
pKD3/mxiD-R	TTAGTAATTTAAGTATGAAACCAATGATTTTTCGTCTTCCAGCAATGGGAATTAGCCATGGTCC
pKD3/mxiH-F	ATGAGTGTTACAGTACCGGATAAAGATTGGACTCTGAGTTCATTGTGTAGGCTGGAGCTGCTTC	cat	This study
pKD3/mxiH-R	TTATCTGAAGTTTTGAATAATTGCAGCATCAACATCCTTAATCAATGGGAATTAGCCATGGTCC
pKD3/pgaC-F	ATGATGAGGTTCGTTTTCTTCTGGCCGTTTTTTATGTCCATTATGTGTAG GCTGGAGCTGCTTC	cat	This study
pKD3/pgaC-R	TTAACCTCTCAGAATCCCGCGATCGGGACTTACCCAACGGGCGCATGGGAATTAGCC ATGGTCC
pgaABCD-F	GATCGATATCGATACAGAGAGAGATTTTGGCAATACATGGAG	pgaABCD	This study
pgaABCD-R	GATCGGATCCATGCATACAGTTAAGTGTATTATCGGTGCAGAG

Random and site-directed mutagenesis.

Tn5 random mutagenesis was performed with the EZ-Tn5 KAN-2 kit (Epicentre, Madison, WI) according to the manufacturer’s instructions. Briefly, the electrocompetent BF-EIEC 52.1 strain was transformed with 1 μL EZ-Tn5 transposome, and transformant colonies were selected on LB agar-kanamycin plates. Transformants were screened for biofilm formation as described above. Tn5 insertion sites in biofilm-negative mutants were mapped using random amplification of transposon ends (RATE) using single-primer PCR as defined by Ducey and Dyer (38).

Type III secretion system (TTSS) mxiH, mxiD, ipaC, and pgaC mutants were constructed using a lambda Red recombination system as described by Datsenko and Wanner (39). Briefly, pKD3/mxiH, pKD3/mxiD, pKD3/ipaC, and pKD3/pgaC forward and reverse primers (Table 2) were used to amplify the chloramphenicol acetyltransferase gene (cat) using pKD3 as a template in a standard PCR. Purified PCR products were used to transform the BF-EIEC 52.1 strain harboring plasmid pKD46. Deletion mutants were then selected on LB agar-chloramphenicol plates. Mutants later had the CHL/cat gene cassette removed by transformation with pCP20. Mutants for mxiH and mxiD were rescued for the invasive phenotype by cloning mxiH and mxiD into pBR322.

Lambda Red mutagenesis was conducted as described for pga operon genes pgaB, pgaC, and pgaD. Those Tn5 mutants in pgaB, pgaC, and pgaD were complemented with a plasmid containing the entire pgaABCD operon. The pgaABCD operon with its Shine-Dalgarno element was PCR amplified using forward and reverse primers for pgaABCD (Table 2). The resulting PCR gene product was digested with EcoRV and BamHI and ligated into the pBAD18 expression vector under the control of the araC promoter. The resulting recombinant plasmids initially transformed into E. coli DH5α were subsequently isolated, purified, and transformed into each of the BF-EIEC 52.1 Tn5 pgaB, pgaC, and pgaD lambda Red gene mutants. For quantitative biofilm assays, these transformants were grown in DMEM with 0.45% dextrose, 100 μg/mL ampicillin, and 10 mM l-arabinose (as an araC promoter inducer).

Multilocus sequence typing, phylogenetic analysis, and DNA sequencing analysis.

Multilocus sequence typing (MLST) was performed as described by Alikhan et al. (40) and EnteroBase. Briefly, the E. coli housekeeping genes adk, fumC, gyrB, icd, mdh, purA, and recA were PCR amplified and sequenced. Sequence type (ST) was obtained by comparing the above-mentioned gene sequences with those of the EnteroBase database (http://enterobase.warwick.ac.uk). A phylogenetic tree was constructed with CLC Genomics Workbench 11.0.1 (Qiagen) using a concatenated sequence of the seven above-listed genes for each strain. The phylogenetic tree was constructed using the neighbor-joining method with Jukes-Cantor distance measurement. The pga operon genes from BF-EIEC 52.1 strain were compared with the whole-genome sequence of the Italian EIEC O96:H19 strain (GenBank accession number CP011416) using nucleotide BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastn&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome).

HeLa cell adhesion and invasion assay.

HeLa cell adhesion and invasion assays were performed as described by Beloin et al. (24) and Scaletsky et al. (41) and Elsinghorst (42), respectively. Briefly, HeLa cells were grown in DMEM with 10% fetal bovine serum, 2 mM l-glutamine, and 1× pen-strep (100 U/mL penicillin and 100 μg/mL streptomycin) on coverslips in 24-well plates to 70% to 80% confluence for adhesion assays and 100% confluence for invasion assays. The cells were washed twice with phosphate-buffered saline (PBS), and fresh DMEM (without pen-strep) supplemented with 2% mannose was added for the adhesion assay. Ten microliters of overnight bacterial culture was added to each well, and the plates were incubated at 37°C in 5% CO₂ for 1 h. After incubation, the medium was aspirated, and the cells were washed five times with PBS.

For the qualitative adhesion assay, the cells were fixed with chilled 100% methanol for 5 min, followed by 30 min of staining with 10% Giemsa. The wells were then washed three times with PBS. After air drying, the coverslips were mounted onto glass slides with mounting medium and observed under a light microscope. For the quantitative adhesion assays, infected cells previously washed with PBS were lysed with 500 μL 1% Triton X-100, diluted, and plated onto LB agar plates. The plates were incubated under aerobic conditions at 37°C overnight. Bacterial CFU were enumerated the following day. For the quantitative invasion assay, after washing, the cells were incubated in DMEM with 600 μg/mL gentamicin for an additional hour at 37°C. The medium was then aspirated, and the cells were washed three times with PBS and subsequently lysed with 500 μL 1% Triton X-100, diluted, and plated onto LB agar plates. The plates were incubated under aerobic conditions at 37°C overnight. Bacterial CFU were enumerated the following day.

Cytotoxicity assay.

Cytotoxicity assays were performed with a cytotoxicity detection kit (Roche Applied Sciences, Mannheim, Germany) per the manufacturer’s instructions. Briefly, 10 μL overnight bacterial culture was added to a confluent monolayer of HeLa cells submerged in 490 μL fresh DMEM and incubated at 37°C for 4 h. The supernatant was collected and centrifuged to remove cellular debris. Bacterial cytotoxicity to HeLa cells was determined by indirectly measuring the release of cytoplasmic lactate dehydrogenase (LDH) into the medium. Background control values were measured in wells inoculated only with 10 μL medium. The total lysis LDH value was obtained by incubating a monolayer with 500 μL 1% Triton X-100.

Confocal microscopy.

Bacterial biofilms were grown on glass coverslips and stained with either SYTO 9, propidium iodide, or calcofluor white for 1 h. The samples were washed three times with PBS, mounted with ProLong Gold antifade (Thermo Fisher Scientific Inc., Waltham, MA), and imaged by confocal laser scanning microscopy (Zeiss 710 confocal laser scanning microscope).

Scanning electron microscopy.

Bacterial biofilms were grown on glass coverslips in 24-well plates. For scanning electron microscopy (SEM), samples were fixed with 2.0% paraformaldehyde and 2.5% glutaraldehyde in 0.05 M sodium cacodylate buffer (pH 7.4) for 24 h. After primary fixation, the samples were washed three times with 0.05 M sodium cacodylate buffer before dehydration with increasing concentrations of ethanol (25%, 50%, 75%, and 100%). The samples were then dried using a Tousimis critical point dryer, mounted onto aluminum SEM sample stubs (Electron Microscopy Sciences), and sputter coated with a 5-nm layer of gold-palladium. Finally, the samples were painted with a thin strip of colloidal silver at the edge to facilitate charge dissipation and imaged with an FEI Quanta 250 field-emission gun scanning electron microscope.

Statistical analyses.

Except where indicated otherwise, comparisons between groups were made with Student’s t tests. All results are representative of three experiments.

For the quantitative biofilm assay with enzymes, blank wells were used as negative controls. All results for each strain (EAEC JM221 strain and BF-EIEC 52.1 strain) were normalized to their respective positive controls in which no enzyme was added, such that both sets had controls of equal means.

For LDH release assays, the relative cytotoxicity was calculated by subtracting the average from blank wells from the value for each well and dividing the resulting value by the difference between the average from blank wells and the average from positive controls. The resulting sets of percentages (at 4 h) were compared by a Student's t test.