Shiga-Toxin-Producing Strains of Escherichia coli O104:H4 and a Strain of O157:H7, Which Can Cause Human Hemolytic Uremic Syndrome, Differ in Biofilm Formation in the Presence of CO2 and in Their Ability to Grow in a Novel Cell Culture Medium

One pathogen that commonly causes gastrointestinal illnesses from the consumption of contaminated food is Escherichia coli O157:H7. In 2011 in Germany, however, there was a prominent outbreak of bloody diarrhea with a high incidence of hemolytic uremic syndrome (HUS) caused by an atypical, more virulent E. coli O104:H4 strain. To facilitate the identification of this lesser-known, atypical E. coli O104:H4 strain, we wanted to identify phenotypic differences between it and a strain of O157:H7 in different media and culture conditions. We found that E. coli O104:H4 strains produced considerably more biofilm than the strain of O157:H7 at 37 °C (p = 0.0470–0.0182) Biofilm production was significantly enhanced by the presence of 5% CO2 (p = 0.0348–0.0320). In our study on the innate immune response to the E. coli strains, we used HEK293 cells that express Toll-like receptors (TLRs) 2 or 4. We found that E. coli O104:H4 strains had the ability to grow in a novel HEK293 cell culture medium, while the E. coli O157:H7 strain could not. Thus, we uncovered previously unknown phenotypic properties of E. coli O104:H4 to further differentiate this pathogen from E. coli O157:H7.


Introduction
One pathogen commonly associated with small outbreaks of gastrointestinal illnesses from the consumption of contaminated food is Escherichia coli O157:H7, which is a Shigatoxin-producing E. coli (STEC) [1,2]. Clinical presentations of E. coli O157:H7 infections can include bloody diarrhea, hemorrhagic colitis, and hemolytic-uremic syndrome (HUS) [3,4]. HUS is a sequela of three conditions: hemolytic anemia, thrombocytopenia, and acute renal failure [4,5]. Because antibiotics can exacerbate the toxicity of infections leading to HUS, currently, there are only supportive treatments for patients with STEC-caused bloody diarrhea [6][7][8]. There are two major types of Shiga toxins (Stxs) with subtypes of each produced by E. coli O157:H7: Stx1 and Stx2 [9,10]. Based on epidemiological studies of outbreaks and animal model studies, Stx2 appears to be the primary subtype responsible for causing bloody diarrhea and HUS [11][12][13].
Less frequent and smaller outbreaks of food-borne bacterial contamination have been attributed to non-O157:H7 serotypes. However, in the spring of 2011 in Germany, an unusually large number of cases (~3800) of bloody diarrhea and HUS were caused by responds to lipopolysaccharides (LPS). TLR4 also requires accessory proteins, such as the LPS binding protein (LBP), myeloid differentiation factor 2 (MD2), and CD14, for optimal activity [46,50,51]. TLR5 is a homodimer and can be activated by flagella or its flagellin subunit [45,[52][53][54].
Since less is known about the atypical E. coli O104:H4 pathotype than the more common E. coli O157:H7, we wanted to identify additional phenotypic differences between these two HUS-causing STECs. We directly compared the two pathogens under different growth or culture conditions to facilitate the identification of the atypical E. coli O104:H4 versus the O157:H7 pathogen. We report here that CO 2 significantly enhanced biofilm formation by the E. coli O104:H4 strains at 37 • C but not by the E. coli O157:H7 strain. In our examination of the interaction of the E. coli strains with a host's innate immune response using HEK293 cells that expressed TLR2 or 4, we found that the E. coli O104:H4 strains but not the O157:H7 strain had the ability to grow in a novel TLR cell culture medium.

Bacterial Strains, Reagents, Growth Conditions, and Antibiotic Susceptibility
The E. coli strains used in the present study are listed in Table 1 Table 1 will be referred to as Ec 01, Ec 02, or Ec 03, respectively. Stock cultures of strains were stored at −80 • C in 50% glycerol in Mueller-Hinton (MH) II broth, cation-adjusted (BD Difco, Thermo Fisher Scientific, Waltham, MA, USA).
For growth studies, stock cultures were streaked out onto sheep-blood agar (SBA) (Remel, Thermo Fisher Scientific) or MH II agar (1.5%) (BD Difco, Thermo Fisher Scientific) plates, and the plates were incubated overnight at 37 • C. A single colony was used to inoculate 2-5 mL of cation-adjusted MH II broth, and the culture was incubated at 37 • C overnight with shaking at 180 rpm. The following day, the overnight culture was diluted 1/50-1/100 in MH II broth or other desired media for further studies. Alternatively, fresh overnight colonies on a solid medium were used to make a suspension of cells for an inoculum in 96-well plates or 5 mL tubes. For the quantitation of E. coli bacterial cells, the conversion factor of 1.0 OD 600 nm, equivalent to 6.83 × 10 8 colony-forming units (CFUs) was used (Henry Heine, USAMRIID). To determine CFUs in desired cultures, 10-fold dilutions of the culture were made in 1X PBS, and 100 µL of each dilution was plated onto triplicate SBA plates. The SBA plates were incubated at 37 • C overnight and the number of colonies was counted to determine the original CFU/mL.
In our initial assessment of the E. coli strains used in the present study, we used Biolog GEN III microplates (Biolog, Hayward, CA, USA) to obtain a phenotypic display of common or differing characteristics. After bacterial strains were grown overnight on SBA plates at 37 • C, bacterial cells and GEN III microplates (carbon utilization) were prepared as per the manufacturer's directions. We also evaluated the antibiotic susceptibilities of the E. coli strains used in the present study. This was determined as previously described by the broth dilution method following the Clinical and Laboratory Standards Institute method [55]. The results from the Biolog GENIII plates and antibiotic susceptibilities studies of the E. coli strains are shown in Tables S1 and S2, respectively, and in part in Table 1. Generally, the E. coli O104:H4 strains exhibited the same phenotypes or antibiotic susceptibilities. The results for the O157:H7 strain were either the same or different from the O104:H4 strains depending on the carbohydrate/chemical or antibiotic being examined. To further characterize the E. coli strains, we examined the phenotype of the E. coli strains on MacConkey agar with sorbitol and a tryptone bile x-glucuronide medium (Remel, ThemoFisher Scientific). To examine the motility of the E. coli strains, we inoculated MH II agar (0.4%) plates with an overnight culture of the strains. The plates were examined for bacterial motility after incubation overnight at 37 • C. Difco Heart Infusion agar (ThermoFisher Scientific, Waltham, MA, USA) for Congo Red (Sigma-Aldrich, St. Louis, MO, USA) (100 µg/mL) plates were used for dye utilization/uptake studies (see Figure S1A) [56]. The presence of Stx2 (List Labs, Campbell, CA, USA) was determined by Western slot-blot analysis of filtered culture supernates with an anti-Stx2 monoclonal antibody [57] that was obtained from Dr. Alison D. O'Brian (Department of Microbiology and Immunology, Uniformed Services University of the Health Sciences, Bethesda, MD, USA). The results of these studies and other characteristics [58][59][60] are presented in Table 1.

Biofilm Formation
Biofilm formation was examined in 96-well round-bottom plates (Corning Falcon, ThermoFisher) and 5 mL polypropylene tubes (12 × 75 mm, Corning Falcon, Thermo Fisher). Bacterial strains were grown on SBA plates overnight at 37 • C, and cells were suspended in MH II broth and OD 600 adjusted to 1.0, as stated previously. For 96-well ELISA plates, 10-20 µL of adjusted cell suspension was used to inoculate 180-190 µL of MH II broth or 200-500 µL MH II broth in 5 mL cultures tubes with snap-cap lids (to safely wash and process the pellets). Bacterial cultures were incubated overnight at 37 • C with or without 5% CO 2 . When cultures were incubated in 5 mL culture tubes, the lids were loosely closed to allow for air/CO 2 exchange. Biofilm formation was quantitated in the 5 mL tubes using a modified method [61] with 0.1% crystal violet (in water and filtered through a 0.45 µm filter) (Sigma-Aldrich). All washing and processing of biofilm pellets by E. coli strains were performed under a biosafety hood under Biosafety Level 2 conditions. All centrifugation steps were conducted at 2000 rpm for 5 min in covered buckets, and buckets were loaded or emptied under the biosafety hood. After overnight incubation, 1.0 mL of PBS (ThermoFisher) was added to each sample, and samples were vortexed and centrifuged. A tube with only MH II medium was included as a blank/reagent control. The samples were decanted and washed twice with 1.0 mL of PBS. After the last centrifugation step, 0.1-0.2 mL of crystal violet was added to each sample, including the blank/reagent control, and samples were left at room temperature for 15-30 min. Then, 1.0 mL of PBS was added to each sample, and the samples were vortexed and centrifuged. After centrifugation, 2.0 mL of PBS was added to each sample, and the samples were vortexed. This wash step was repeated twice. After the last wash, the tubes were left to dry under the biosafety hood for at least 3 h. To the dry pellets, 1.0 mL of 95% ethanol was added, and the samples were vortexed vigorously to dissolve the pellets and then left for 15-30 min. In some cases, the samples were left in the solvent overnight with the caps of the tubes snapped closed to ensure complete dissolution of the pellet. Then, 200 µL of each sample was transferred to a 96-well ELISA plate, which included the blank/reagent control sample, and the plates were read at 600 nm. The absorbance of the blank/reagent control sample (absorbance at 600 nm < 0.10) was not subtracted from the readings from the wells that contained the bacterial cells, but it was plotted with the results of the test samples.

Human Embryonic Kidney (HEK) 293 Toll-like Receptor (TLR) Cell Cultures
HEK293 cells expressing TLR2 or 4 used for the evaluation of TLR activation by E. coli strains were obtained from Invivogen (San Diego, CA, USA). The TLR-expressing HEK293 cells were also transfected with an embryonic alkaline phosphatase reporter gene. The expression of the reporter gene was dependent on the binding of multiple NF-kB and AP-1 proteins in the enhancer region of the reporter gene. Once expressed, the alkaline phosphatase was secreted (SEAP) into the HEK Blue Detection (HBD) cell culture medium, where its activity could be detected (Invivogen). SEAP cleaved the color indicator 5-bromo4chloro-3-indolyl phosphate in the HBD cell culture medium. Once the color indicator was cleaved by SEAP, a blue-purple product was generated. The growth and maintenance of the HEK293 TLR2/4 cells and the examination of TLR2/4 activation by live bacterial strains were as previously described [45]. The activation of TLR2/4 by E. coli strains in the HBD cell culture medium (Invivogen) was performed in 96-well flat-bottom Immunlon 2HB plates (ThermoFisher, Waltham, MA, USA). The HBD cell culture medium was prepared as per the manufacturer's directions. Briefly, the contents of a single pouch of HBD powder were poured into a 125 mL sterile flask (Corning, Fisher Scientific), and 50 mL of endotoxin-free water was added to solubilize the powder. The dissolved HBD medium was incubated at 37 • C for 30-60 min with occasional stirring and passed through a 0.22 µm filter. The sterile medium was stored at 4 • C until use for up to 2 weeks. HEK293 TLR2 and TLR4 cells were adjusted to 2.5 × 10 5 cells/mL and 2.0 × 10 5 cells/mL, respectively, in an HBD cell culture medium, as per the manufacturer's directions before use. In the activation studies, 180 µL of adjusted HEK293 TLR2 or 4 cells were inoculated in triplicate wells with 2.5 × 10 4 bacterial cells per well. For the control wells, 2 × 10 6 heat-killed Listeria monocytogenes (HKLM) (Invivogen) was used for the TLR2 positive control and 0.2 ng E. coli lipopolysaccharide (LPS) (Invivogen) was used for the TLR4 positive control. The inoculated HEK293 TLR plates were incubated at 37 • C with 5% CO 2 overnight for 18-24 h. The results were read at 630 nm with a BioTek Elx808 spectrophotometer (BioTek, Winooski, VT, USA). Similar E. coli cultures were performed in an HBD cell culture or DMEM cell culture maintenance medium without HEK293 TLR cells in 96-well plates or 15 mL culture tubes. All readings of cultures were used to evaluate the amount of TLR2 or 4 activation by live bacterial cells in 96-well plates and were used to plot the data, including the media-only wells, which contained only media, unless noted. The latter readings (<0.10 absorbance at 630 nm) were not subtracted from the wells containing E. coli cells, but they were plotted with the readings of the other wells.

E. coli Strains in an HBD Cell Culture Medium
To examine the growth of the E. coli strains in the HBD cell culture medium, single colonies from SBA plates were used to inoculate 2 mL of MH II broth, and the cultures were incubated overnight at 37 • C with shaking at 180 rpm. The following day, 0.25 mL of the overnight culture was used to inoculate 12.5 mL of the HBD cell culture medium (1/50 dilution) in a 125 mL Corning flask (Fisher Scientific), and the cultures were incubated at 37 • C with shaking at 180 rpm. Growth of the cells was followed by an increase in absorbance at 600 nm for up to 48 h. The HBD cell culture medium was used as a blank for the culture readings. For comparison, E. coli strains were also grown in 12 mL of MH II broth at 37 • C with shaking at 180 rpm after inoculation with 0.12 mL of an overnight culture (1/100 dilution), and the growth was followed by an increase in absorbance at Microorganisms 2023, 11, 1744 6 of 18 600 nm. MH II broth was used as a blank for the culture readings. In other cases, where noted, single colonies of E. coli strains grown on SBA plates were directly used to inoculate 2 mL of the HBD culture medium, and the cell cultures were incubated overnight at 37 • C with 5% CO 2 .

Comparison of Amino Acid Sequence of AAF/I Protein Components
The amino acid sequence of the AggA, AggB, AggC, and AggD proteins for AAF/I was obtained from the National Center for Biotechnology Information (NCBI). For E. coli O104:H4 2011C-3493, the amino acid sequence for plasmid pAA proteins was under NCBI reference sequence NC_018666.1. For strain 2009EL-2050, the amino acid sequence for plasmid pAA proteins was under NCBI reference sequence NC_018654.1. For strain 2009EL-2071, the amino acid sequence for plasmid pAA proteins was under NCBI reference sequence NC_018662.1.

Statistical Analysis
Values were log transformed prior to analysis, with results summarized as the geometric mean and geometric standard error of the mean. Comparisons between groups were made by Welch's t-test, applied to the log-transformed values. The comparisons between groups were based on a Welch's t-test of the appropriate single degree of freedom contrast in the two-way ANOVA, with factors corresponding to treatment groups. Analysis was performed in SAS ® version 9.4 (SAS Institute Inc., Cary, NC, USA). The results were considered significant when p < 0.05.

Enhanced Biofilm Formation by E. coli O104:H4 in 5% CO 2
We found clear differences in biofilm formation among E. coli strains grown during our studies and identified differences in growth between E. coli O104:H4 and O157:H7. We grew the E. coli strains in MH II broth in 96-well plates. We noticed a difference in their pellets after overnight incubation at 37 • C ( Figure 1A, top panel). E. coli O157:H7 and the control E. coli ATCC 25922 had well-defined circular pellets, while the E. coli O104:H4 strains Ec 01, Ec 02, and Ec 03 had pellets that covered the bottom of the wells. In some instances, closer examination of the pellets of the E. coli O104:H4 strains revealed small pocked-marked clearings on the peripheral of the pellets reminiscent of phage plaques ( Figure S1B). We also found that incubation of the 96-well plates at 37 • C with 5% CO 2 noticeably altered the appearance of the pellets of the E. coli O104:H4 strains. They appeared as wrinkled or folded membranous structures. However, this was not observed in the pellets of the E. coli ATCC 25922 or O157:H7 strains ( Figure 1A, bottom panel).
We used a crystal violet assay to quantitate the change in pellet formation by the E. coli strains with or without CO 2 . E. coli strains were grown in an MH II medium (0.2-0.5 mL) in 5 mL culture tubes and cultures were incubated overnight at 37 • C with or without CO 2 . Figure 1B,C present representative results from one of three studies we conducted under these conditions. The pellets of the E. coli strains incubated in the 5 mL tubes looked similar to those observed in the 96-well plates with and without CO 2 ( Figure 1B). However, we did not observe plaque-like clearings as we did in the 96-well plates. Figure 1C shows that there was more pellet produced by the E. coli O104:H4 strains than the E. coli ATCC 25922 or O157:H7 strains under either condition. There was a significant difference in the amount of pellet formed by the E. coli O104:H4 strains in the presence of CO 2 compared to its absence: Ec 01 (p = 0.0326), Ec 02 (p = 0.0022), and Ec 03 (p = 0.0498) ( Figure 1C). This difference in the amount of pellet produced in the absence or presence of CO 2 by E. coli O104:H strains was also supported when we compared the amount of pellet formed without CO 2 between the E. coli O157:H7 and O104:H4 strains. The latter strains produced more than 3-to 7-fold more pellets than E. coli O157:H7 (Ec 01, p = 0.0270; Ec 02, p = 0.0470; Ec 03, p = 0.0182). When we compared the amount of pellet produced by the E. coli O157:H7 and O104:H4 strains grown in the presence of CO 2 , the difference in the amount of pellet formed increased to 10-to 15-fold (Ec 01, p = 0.0348; Ec 02, p = 0.0320; Ec 02, p = 0.0343). Because of previous reports linking E. coli O104:H4 with biofilm formation [34,62,63], the pellets produced by the E. coli O104:H4 strains will henceforth be referred to as biofilms in this report. We used a crystal violet assay to quantitate the change in pellet formation by the E. coli strains with or without CO2. E. coli strains were grown in an MH II medium (0.2-0.5 mL) in 5 mL culture tubes and cultures were incubated overnight at 37 °C with or without CO2. Figure 1B,C present representative results from one of three studies we conducted under these conditions. The pellets of the E. coli strains incubated in the 5 mL tubes looked similar to those observed in the 96-well plates with and without CO2 ( Figure 1B). However, we did not observe plaque-like clearings as we did in the 96-well plates. Figure 1C shows that there was more pellet produced by the E. coli O104:H4 strains than the E. coli ATCC 25922 or O157:H7 strains under either condition. There was a significant difference in the amount of pellet formed by the E. coli O104:H4 strains in the presence of CO2 compared to its absence: Ec 01 (p = 0.0326), Ec 02 (p = 0.0022), and Ec 03 (p = 0.0498) ( Figure  1C). This difference in the amount of pellet produced in the absence or presence of CO2 by E. coli O104:H strains was also supported when we compared the amount of pellet formed without CO2 between the E. coli O157:H7 and O104:H4 strains. The latter strains produced more than 3-to 7-fold more pellets than E. coli O157:H7 (Ec 01, p = 0.0270; Ec 02, p = 0.0470; Ec 03, p = 0.0182). When we compared the amount of pellet produced by the E. coli O157:H7 and O104:H4 strains grown in the presence of CO2, the difference in the amount of pellet formed increased to 10-to 15-fold (Ec 01, p = 0.0348; Ec 02, p = 0.0320; Ec

Interaction of the Live E. coli Strains with Innate Immune TLR2 and 4
We used HEK293 cells that expressed TLR2 or 4 to evaluate whether they interacted with the E. coli strains. Besides examining differences in the growth between the pathogenic E. coli O104:H4 strain and the O157:H7 strain, we also asked if there were possible differences in E. coli O104:H4 and O157:H7 interactions with a host's innate immune system. is activated by acylated lipoproteins, while TLR4 is activated by lipopolysaccharides [46]. The activation of TLRs stimulates the expression of NF-ebb that is required for the enhanced expression of the reporter protein SEAP from stimulated HEK293 cells. This gives a blue/purple color after cleavage of the indicator 5-bromo-4-chloro-3-indolyl phosphate by secretory alkaline phosphatase. We inoculated HEK293 TLR2 or 4 cells in a cell culture medium in 96-well plates (triplicate wells) with the E. coli strains and incubated the cultures at 37 • C with 5% CO 2 overnight. Representative results from one of three TLR2 and 4 activation studies are presented. The results of this study are shown in Figure 2A and are representative results from one of three studies we conducted to examine the activation of TLR2 or 4 by the E. coli strains. We observed a notable significant difference in the amount of TLR4 activated compared to the amount of TLR2 activated (p = 0.0041-0.0001) by all E. coli strains that we evaluated except E. coli O157:H7 ( Figure 2B). Nevertheless, the E. coli O157:H7 strain in this case still activated more TLR4 than TLR2, but the difference in the activation between the two TLRs by E. coli O157:H7 was not significant. [46]. The activation of TLRs stimulates the expression of NF-ebb that is required for the enhanced expression of the reporter protein SEAP from stimulated HEK293 cells. This gives a blue/purple color after cleavage of the indicator 5-bromo-4-chloro-3-indolyl phosphate by secretory alkaline phosphatase. We inoculated HEK293 TLR2 or 4 cells in a cell culture medium in 96-well plates (triplicate wells) with the E. coli strains and incubated the cultures at 37 °C with 5% CO2 overnight. Representative results from one of three TLR2 and 4 activation studies are presented. The results of this study are shown in Figure 2A and are representative results from one of three studies we conducted to examine the activation of TLR2 or 4 by the E. coli strains. We observed a notable significant difference in the amount of TLR4 activated compared to the amount of TLR2 activated (p = 0.0041-0.0001) by all E. coli strains that we evaluated except E. coli O157:H7 ( Figure 2B). Nevertheless, the E. coli O157:H7 strain in this case still activated more TLR4 than TLR2, but the difference in the activation between the two TLRs by E. coli O157:H7 was not significant.  When we compared the amount of TLR2 or 4 activated by the E. coli O157:H7 strain with the amount activated by the other E. coli strains, we found that E. coli O157:H7 activated significantly less TLR2 (p = 0.0114-0.0007), but there was no difference in TLR4 activation.

Growth of the E. coli O104:H4 Strains but Not the O157:H7 Strain in an HBD Cell Culture Medium
When examining the activation of TLR2/4 by the E. coli strains, we found a color reaction difference between the E. coli ATCC 25922 and O157:H7 strains (blue-purple) and the E. coli O104:H4 Ec 01, Ec 02, and Ec 03 strains (yellow-green) (see Figure 2A). To determine if the difference depended on live E. coli strains, we inoculated HEK293 TLR2/4 cells with irradiated-inactivated E. coli strains. After overnight incubation, all the E. coli strains produced a similar blue-purple color reaction, with a higher amount of TLR2 than TLR4 activated. These results suggested that the yellow-green color reaction generated by the E. coli O104:H4 strains with HEK293 TLR2/4 cells was dependent on live bacterial cells. We also examined other cell lines, such as RAW 264.7 and Vero cells in an HBD cell culture medium with the live E. coli strains. We observed the same color reaction that we observed with E. coli strains using HEK293 TLR2/4 cells. To determine if the presence of TLRs on the HEK293 TLR2/4 cells was responsible for the different color reactions, we inoculated HEK293 TLR null cells that did not express TLR2 or 4 in an HBD cell culture medium with live E. coli strains. Figure 2C presents representative results from one of three studies we conducted with HEK293 TLR null cells. We observed no activity with the E. coli ATCC 25922 or O157:H7 strains or the positive controls, but the wells containing the E. coli O104:H4 strains were greenish in color after incubation ( Figure 2C). Because the yellow-greenish color reaction by the E. coli O104:H4 strains in the HBD cell culture medium did not appear to depend specifically on HEK293 cells or the presence of TLR2 or 4, we inoculated the HBD cell culture medium (without the HEK293 cells) with the E. coli strains. This study was performed three times and Figure 2D is representative of one of these studies. We observed no color change with the E. coli ATCC 25922 or O157:H7 strains, but we observed a yellowish color reaction with the E. coli O104:H4 strains ( Figure 2D). Microscopic examination of the wells containing the E. coli O104:H4 strains revealed the presence of small rod-shaped, bacterial-like cells. These cells were absent in wells that contained the E. coli ATCC 25922 or O157:H7 cells. Thus, we made dilutions of all the wells and plated aliquots onto SBA plates. After incubation overnight at 37 • C, we detected more than 1.0 × 10 8 CFU/mL from the E. coli O104:H4 cultures, but we detected no CFUs from the E. coli ATCC 25922 or O157:H7 cultures. This color reaction difference between the E. coli strains was not observed with a DMEM cell maintenance medium that is normally used to maintain the HEK293 TLR2/4 cells ( Figure 2D). Taken together, these results suggest that the E. coli O104:H4 strains but not the E. coli ATCC 25922 or O157:H7 strains can grow in an HBD cell culture medium independent of HEK293 TLR2/4 cells.
To further examine the potential of E. coli O104:H4 strains to grow in the HBD cell culture medium, we examined cultures of E. coli O104:H4 Ec 01, Ec 02, and Ec 03 strains and the E. coli ATCC 25922 and O157:H7 strains in 125 mL flasks containing an HBD cell culture medium (12.5 mL). We inoculated the HBD cell culture medium (0.25 mL) with an overnight culture of the E. coli strains grown in an MH II medium, and the cultures were incubated at 37 • C with shaking at 180 rpm. We followed growth by tracking the change in absorbance (A) 600 nm, the cell cultures were followed for 48 h. Figure 3A presents representative results from one of three growth studies of the E. coli strains in HBD cell culture media in 125 mL flasks. After 48 h, the E. coli O104:H4 cultures reached an A 600 of 1.29-1.38, and we detected 1.6 × 10 8 -4.3 × 10 8 CFUs/mL present in the E. coli O104:H4 cultures. The E. coli O104:H4 Ec 03 strain appeared to grow slightly slower than the E. coli O104:H4 Ec 01 and Ec 02 strains at the beginning, but it reached close to the same density as the other two E. coli O104:H4 strains after 48 h. We observed no change in culture density nor did we recover any CFUs from the E. coli ATCC 25922 or O157:H7 cultures after the same time period. Figure 3B shows the color of the E. coli cultures in the HBD cell culture medium after 48 h. The E. coli O104:H4 Ec 01, Ec 02, and Ec 03 cultures were green, while the E. coli ATCC 25922 and O157:H7 cultures remained pink (unchanged). In contrast, when we grew the five E. coli strains in MH II broth (12.0 mL) inoculated with 0.12 mL of an overnight culture, we observed that all E. coli strains grew at an almost identical rate. They reached a similar A 600 (~1.80) after 7 h of incubation at 37 • C with shaking at 180 rpm ( Figure 3B). The cell densities of the cultures were 5.8 × 10 9 -1.2 × 10 10 CFU/mL after 7 h of incubation, which was~10-fold higher than the growth of the E. coli O104:H4 strains in the HBD cell culture medium. It should be noted that no supplemental CO 2 was present during the incubation of these cultures. Thus, the growth of the E. coli O104:H4 strains in the HBD cell culture medium was shown by an increase in cell density and a change in the color of the medium.
trast, when we grew the five E. coli strains in MH II broth (12.0 mL) inoculated with 0.12 mL of an overnight culture, we observed that all E. coli strains grew at an almost identical rate. They reached a similar A600 (~1.80) after 7 h of incubation at 37 °C with shaking at 180 rpm ( Figure 3B). The cell densities of the cultures were 5.8 × 10 9 -1.2 × 10 10 CFU/mL after 7 h of incubation, which was ~10-fold higher than the growth of the E. coli O104:H4 strains in the HBD cell culture medium. It should be noted that no supplemental CO2 was present during the incubation of these cultures. Thus, the growth of the E. coli O104:H4 strains in the HBD cell culture medium was shown by an increase in cell density and a change in the color of the medium.

Growth of the E. coli O104:H4 Strains in an HBD Cell Culture Medium Starting with a Single Colony
We further examined our observation that the E. coli O104:H4 strains but not the E. coli ATCC 25922 or O157:H7 strains could grow in an HBD cell culture medium. We took single colonies of the five E. coli strains grown overnight on SBA plates and inoculated each into 2 mL of an HBD cell culture medium in 15 mL culture tubes. The cultures were then incubated at 37 • C with 5% CO 2 overnight (without shaking). A single overnight E. coli colony on sheep blood agar (~2.5 mm diameter) consists of~2 × 10 8 CFUs (Amemiya, unpublished). Figure 4 presents representative results from one of the three independent studies we conducted. After overnight incubation, the E. coli O104:H4 Ec 01, Ec 02, and Ec 03 cultures were greenish in color with occasional bubbles on top of the liquid culture. The E. coli ATCC 25922 and O157:H7 cultures were light blue-purple in color with no bubbles. The media control stayed pinkish. We checked the pH of the cultures to determine if any metabolic activity occurred. The media control and E. coli ATCC 25922 and O157:H7 samples had a pH between 8.5 and 9.0, while the E. coli O104:H4 cultures had a pH of 5.0. The color change in the E. coli ATCC 25922 and O157:H7 cultures when compared to the media-only sample may have been due to the presence of alkaline phosphatase, which is found in the periplasmic space of gram-negative bacteria. It may have leaked from the bacterial cells in the original inoculum [63,64]. Thus, we were able to repeat the color reaction change and growth of the E. coli O104:H4 cultures that we observed in the larger HBD cell cultures in a different format to differentiate their growth from the E. coli O157:H7 strain.
The E. coli ATCC 25922 and O157:H7 cultures were light blue-purple in color with no bubbles. The media control stayed pinkish. We checked the pH of the cultures to determine if any metabolic activity occurred. The media control and E. coli ATCC 25922 and O157:H7 samples had a pH between 8.5 and 9.0, while the E. coli O104:H4 cultures had a pH of ~5.0. The color change in the E. coli ATCC 25922 and O157:H7 cultures when compared to the media-only sample may have been due to the presence of alkaline phosphatase, which is found in the periplasmic space of gram-negative bacteria. It may have leaked from the bacterial cells in the original inoculum [63,64]. Thus, we were able to repeat the color reaction change and growth of the E. coli O104:H4 cultures that we observed in the larger HBD cell cultures in a different format to differentiate their growth from the E. coli O157:H7 strain. Figure 4. Growth of E. coli O104:H4 strains in HBD cell culture medium starting with single colonies. A single colony from an overnight growth on SBA plates was used to inoculate 2 mL of HBD medium and cultures were incubated overnight at 37 °C with 5% CO2. The figure presents representative results from one of three independent studies. There were color differences between the E. coli O104:H4 strains and the ATCC 25922 and O157:H7 strains. This, along with the pH differences (pH ~5.0 vs. 8.5-9.0), reflected the growth of the E. coli O104:H4 strains. The purple color of the E. coli ATCC and O157:H7 cultures may have been derived from endogenous alkaline phosphatase that came from the original inoculum. A single colony from an overnight growth on SBA plates was used to inoculate 2 mL of HBD medium and cultures were incubated overnight at 37 • C with 5% CO 2 . The figure presents representative results from one of three independent studies. There were color differences between the E. coli O104:H4 strains and the ATCC 25922 and O157:H7 strains. This, along with the pH differences (pH~5.0 vs. 8.5-9.0), reflected the growth of the E. coli O104:H4 strains. The purple color of the E. coli ATCC and O157:H7 cultures may have been derived from endogenous alkaline phosphatase that came from the original inoculum.

Discussion
In our observations on the biofilms produced by the EHEC/EAEC pathotype E. coli O104:H4, we found that biofilms produced by the O104:H4 strains could be affected by the presence of CO 2 . In our initial study, there was a 3-to 7-fold difference in the substantial amount of biofilm produced by the E. coli O104:H4 strains compared to the EHEC pathogen E. coli O157:H7 in the absence of CO 2 . We observed a significant enhancement (10-to 15-fold) of the amount of biofilm formed by E. coli O104:H4 when grown in the presence of CO 2 . We observed no response of the O157:H7 strain. The effect of CO 2 on biofilm formation and the response to changes in CO 2 levels in local environments were found to be critical in some pathogens' ability to upregulate the expression of virulence factors required to successfully infect a mammalian host [65]. The early stages of biofilm formation begin with the ability of E. coli O104:H4 to adhere to and colonize intestinal epithelial cells. This is mediated in part by the biosynthesis of at least two types of fimbriae, which are long, extracellular, polymerized protein structures: AAF from the 74 kb pAA virulence plasmid and LpF from the pathogen's chromosomal genes [31,33,38]. The presence of AAF has been shown to be required for adherence and biofilm formation by E. coli O104:H4 [20,32,62]. It has been suggested that adherence of the pathogen to epithelial cells facilitates the translocation of Stx2a to the intestinal epithelium [66,67]. The expression of genes that encode for AAF is under the control of the AggR regulator, which is also encoded on the pAA plasmid. AggR belongs to the family of AraC-like transcriptional activators and regulates the expression of select genes on the pAA plasmid and EAEC chromosome [31,68]. The upregulation of AggR expression appears to be (p)ppGpp-dependent [69]. There are at least five subtypes of AAF (I-V), with their operons organized as aggDCBA, where the D gene encodes a chaperone, the C gene encodes an usher protein, the B gene encodes a minor pilin, and the A gene encodes the major pilin subunit of the AAF fimbriae [33,70]. Thus, the assembly of AAF appears to occur through the chaperone-usher pathway that is ubiquitous in Gram-negative bacteria [71]. In a study to evaluate the effect of AAF types (I, III, IV, and V) in an EAEC 55989 agg3 − deletion mutant, it was reported that the AAF/I recombinant formed the largest bacterial aggregates and adhered more to Hep-2 cells than the EAEC 55989 agg3 − recombinants expressing III, IV, or V fimbriae. These results suggested that the AAF/I subtype may be partly responsible for the extreme pathotype of the outbreak E. coli O104:H4 strain [33]. However, the AAF/I-expressing recombinant also exhibited the highest auto-aggregation rate and the lowest biofilm formation of the AAF III, IV, or V recombinants [33]. It was not clear why there was no correlation between an increase in adherence of the different recombinants to Hep2 cells and the amount of biofilm formed with the expression of AAF/I. It had been previously shown that the presence of AAF/I coincides with the ability to form biofilms in E. coli O104:H4 [32,62,68]. It may be that biofilm formation is a more temporal, controlled process that requires the biosynthesis of certain molecules in an ordered manner. Biofilm formation may not start with an aggregate of cells formed in a haphazard manner. Additionally, in an EAEC 55989 agg3 − bacteria, AAF/I may not mediate biofilm formation, nor in the LB medium used in this study. A recent report on EAEC strains also found a variant AAF (agg3A and agg5A genes) that coincided with the presence of a novel pAA variant [72].
To identify the type of AAF present in the E. coli O104:H4 strains used in the present study, we examined the annotated amino acid proteins of the pAA plasmids from the E. coli O104:H4 strains 2011C-3493 (Ec01), 2009EL-2050 (Ec02), and 2009E-2071 (Ec03) that had been deposited in GenBank [59]. They are annotated under the NCBI reference sequence nos. NC_018666.1, NC_018654.1, and NC_018662.1, respectively. All the protein subunits AggA, AggB, AggC, and AggD were identified as the aggregative adherence fimbriae I (AAF/I) type. To determine how close the AAF/I subunits were to each other from the different E. coli O104:H4 strains, we compared the amino acid sequence of the AggABCD protein subunits from 2011C-3493, 2009E-2050, and 2009E-2071 (see Figure S2). The alignment of the individual subunits showed that there was 100% identity in the amino acid sequences of the AggA, AggC, and AggD proteins. In AggB, there was only one amino acid difference out of 145 amino acids, which was at amino acid 95 (G or W). We also compared the amino acid sequences of the AggR regulator (265 amino acids in length) from the three pAA plasmids. We found them to be identical. In all three pAA plasmids, the aggR gene was 10,359 nucleotides from the aggA gene. Since AggR positively regulates the expression of the AAF/I operon, we proposed that the direction of transcription is likely to be → aggABCD, which we found in the arrangement for the AAF/I subunit proteins in the 2011C-3493 and 2009EL-2071 strains. In these two cases, the AggR was upstream from the aggABCD operon. However, for strain 2009EL-2050, the AAF/I subunits were reported as DCBA; yet, AggR was downstream from the aggABCD operon. In this latter case, the direction of transcription may be DCBAagg ←. Bielaszewska [73]. Additionally, Berger et al. (2016) reported the presence of aggA, aggB, aggC, and aggD genes for AAF/I on the pAA plasmid from E. coli O104:H4 LB26692 [31]. Thus, the E. coli O104:H4 strains used in our study appear to express AAF/I fimbriae and may be in part responsible for the robust biofilm formation by the E. coli O104:H4 strains observed in our studies.
Other fimbriae involved in biofilm formation included lpf, which was first noted in the outbreak E. coli O104:H4 strains by Bielaszewska et al. (2011) [66] and Rasko et al. (2011) [21]. It was subsequently found that the E. coli O104:H4 strains had two lpf operons (lpf1 and lpf2). They were highly homologous to the lpf1 and lpf2 genes in the EAEC 55989 strain but not in the E. coli O157:H7 EDL933 strain [38]. The lpf1 operon consists of genes A-E, where lpf1A is the major fimbriae subunit, lpf1B is a chaperone, lpf1C is an outer membrane usher protein, lpf1D is a minor fimbrial subunit, and lpf1E may be a regulator or another fimbrial subunit [38]. The lpf2 operon consists of genes A-D, with the possible presence of a duplicate D gene, and the protein products are analogous to those found in the lpf1 operon [38]. Thus, like AAF/I, the assembly of the LpF fimbriae appears to be carried out through the chaperone-usher pathway [71].
It had been previously noted that biofilm formation by E. coli O104:H4 was greater than that by E. coli O157:H7, both in vitro and in vivo [62]. The in vitro studies, however, were performed without CO 2 in DMEM with 0.45% glucose. It was reported that E. coli O104:H4 produced 7.3-fold greater amounts of biofilm compared with the E. coli O157:H7 strain. The authors also examined biofilm formation in a flow cell in DMEM after 16 h and found that biofilm formation only occurred with E. coli O104:H4 and not with E. coli O157:H7 under these conditions [62]. We did not observe biofilm formation by the E. coli O104:H4 strains in an HBD cell culture medium, only in an MH II medium, nor did we observe biofilm formation by the E. coli O104:H4 strains in a DMEM maintenance medium, which we used to maintain the HEK293 TLR2/4 cells. However, the maintenance medium also contained other proprietary inhibitory agents. Additionally, our finding that E. coli O157:H7 was a weak biofilm producer was previously noted for most isolates of this EHEC strain [28,[74][75][76]. Weak biofilm and curli fimbriae production also coincided with the weak binding of Congo Red dye [74,77]. Chen et al. (2013) [28] proposed that there are three factors that influence biofilm formation in E. coli strains: (1) prophage insertion into the MerR-like (mlrA or previously labeled yehV) regulator gene; (2) mutation in RpoS, the stress-response sigma factor required for the expression of stress response genes; and (3) impaired bacterial motility. MlrA and RpoS are both required for biofilm formation and curli fimbriae biosynthesis [78,79]. MlrA is a DNA-binding transcriptional activator that is required for the expression of csgD which is responsible for the expression of the curli csgBAC genes in E. coli. RpoS activity, on the other hand, is required for the expression of mlrA [78][79][80]. Excision of the prophage from the mlrA gene was shown to restore curli biosynthesis and biofilm formation [75]. In some non-O157:H7 strains that were limited in biofilm formation and curli expression, they found that prophage insertions in mlrA could not be complemented. These strains shared a common phenotype that had a defect in motility. Thus, it was suggested that motility was another factor that limited biofilm formation [28]. We surmise in our studies with E. coli ATCC 25922 that the limited biofilm formation and Congo Red binding that we observed may similarly result from a defect in mlrA or RpoS, as reported for O157:H7 and other non-O157:H7 strains [28]. E. coli ATCC 25922 has been previously reported to be a strong biofilm producer [81,82]. It has also been found that inactivation of the purL gene in E. coli ATCC 25922 can result in limited biofilm formation [83]. This finding suggests that there may be other uncharacterized components or genetic elements that could affect biofilm formation in E. coli besides defects in mlrA, RpoS, or a lack of motility.
The differential growth between the E. coli O104:H4 and O157:H7 strains in the HBD cell culture medium was unexpected. Normally, this cell culture medium is used for detecting the activation of TLR(s) that are present in specialized HRK293 cells. HEK293-TLR cells have been used previously to evaluate the potential host innate immune response to several other live Gram-negative pathogens. The growth of these microorganisms in an HBD cell culture medium, however, was not observed [45]. In the present study, we examined the activation of TLR2 and 4, which are two TLRs that are exogenously expressed on the surface of mammalian cells [46], by live E. coli strains. We found that the live E. coli strains activated both TLR2 and 4, with TLR4 activation being more prevalent than TLR2 activation. We have previously observed that under certain conditions, TLR2 or 4 activation can vary widely in some Gram-negative pathogens. TLR2 and 4 could both be activated with one being more dominant than the other, TLR2 or 4 alone could be activated, or neither TLR2 nor 4 was activated [45]. The activation of these TLRs depends on the microorganism being examined or the conditions of growth of the microorganism [45]. Because the HBD cell culture medium was a proprietary medium (Invivogen), we can only speculate that there may have been at least two categories of inhibitory compounds (or antibiotics) present in the HBD cell culture medium. At least one of these allowed the E. coli O104:H4 strains to grow but not the O157:H7 strain. In comparison, in a rich, complex medium, such as the MH II medium, the growth of the E. coli strains was essentially identical (see Figure 3C). The first type of inhibitory compound that may have been present in the HBD cell culture medium is an antibiotic(s) to maintain the sterility of the HEK293 TLR cell line, for example, a combination of penicillin/streptomycin and/or another proprietary anti-microbial agent. The second type of inhibitory compound that may have been present is an antibiotic (or proprietary reagent), for example, ampicillin, to help maintain or select the SEAP expression plasmid in the transfected HEK293 TLR cell line [84,85]. The inhibitory reagent may also have been a combination of antibiotics or inhibitory components in the HBD cell culture medium. From our studies, we know that E. coli O157:H7 was relatively sensitive to streptomycin and ampicillin, while the E. coli O104:H4 strains we used in the present study were resistant to both antibiotics (see Table S2) [58][59][60]. Additionally, the E. coli O157:H7 and O104:H4 strains that were used in the present study were resistant to penicillin G (Table S2), which eliminate it as the differential or selective agent in the HBD cell culture medium. A mixture of HBD cell culture medium and Congo Red agar was also evaluated. The differential growth of the E. coli strains on a solid HBD cell culture medium and Congo Red agar was consistent with what we observed with the HBD cell culture medium. The colonies of E. coli O104:H4 on this solid medium, however, were black in color.
Further studies are needed to analyze the culture conditions that allow the E. coli O104:H4 strains to grow in the HBD cell culture medium but not the E. coli O157:H7 strain. Additionally, other strains of E. coli O104:H4 and O157:H7 as well as other non-O157:H7 strains associated with HUS must be examined to evaluate the specificity of the medium. That the E. coli O104:H4 strains were able to grow in flasks with an HBD cell culture medium when shaken without CO 2 was noted. However, smaller (0.2-2.0 mL), stationary cultures (96-well plates or small tubes) of O104:H4 strains always grew with CO 2 . In addition, the effect of CO 2 on virulence and the possible augmented expression of AAF/I fimbriae and other pathogen virulence proteins, such as Stx2a, must be investigated. Our report of new findings on the difference between E. coli O104:H4 and O157:H7 will further help differentiate between these two pathogens. In the future, it may assist in establishing the type of appropriate treatment for patients presenting with HUS. The involvement of alternate metabolic pathways that allow for the growth of E. coli O104:H4 and enhance virulence may offer potential targets for possible therapeutic intervention for HUS caused by this pathogen.