Initially, Vero cell monolayer (2D) and THP-1 cell suspensions in a 48-well tissue culture plate were tested for their sensitivity to purified Stx (1 μg/ml) after 16 h of exposure. In 2D setup, THP-1 cells showed significantly higher sensitivity to purified commercial Stx toxin subtypes than the Vero cells, and the corresponding Trypan blue stained microscopic cell images confirmed cell death characterized by cell rounding, detachment, and dye uptake (Figure 1A,B). Both cell lines were then embedded in collagen matrix using rat-tail collagen type I in a 48-well plate as before to create a 3D cell structure (Banerjee et al., 2008) and tested against purified Stx preparations. Data show that the 3D Vero cells were more sensitive to Stx than the 3D THP-1 cells (Figure 1A,B). In contrast, 2D THP-1 is more sensitive to Stx than the 2D Vero (Figure 1A,B). Next, we determined the limit of detection for each platform to purified toxins diluted serially and found that cytotoxicity response is concentration-dependent (Figure 1C–F). In both 2D Vero and 3D THP-1 assays, the detection limit was 0.5–1 μg/ml and cytotoxicity values varied from 20 to 30% after 16 h exposure. While the 3D Vero and 2D THP-1 cell platforms detected the Stx subtypes as low as 62.5–125 ng/ml. Collectively, these data indicate that the 3D Vero cell platform provides the best response to Stx thus was selected for further assay development.

Next, we tested the sensitivity of 2D and 3D Vero cell-platforms after exposure to stx⁺ and stx^- STEC cells after 16 h exposure at different cell concentrations. In both 2D and 3D platforms, a minimum of 10⁷ CFU/ml STEC was required to yield a positive response (Figure 2A–D and Table 1). The cytotoxic response from the 3D Vero platform (38 ± 7%, P < 0.05) was significantly higher (P < 0.05) than the 2D Vero platform (26 ± 2%) after exposure to 10⁷ CFU/ml of STEC (Figure 2A–C). Exposure to a higher concentration of STEC (10⁸ CFU/ml) induced significantly (P < 0.05) higher levels of cytotoxic response (72 ± 2 vs. 76 ± 4%) (Figure 2C). Cytotoxicity values of stx^- strain (489) tested in either 2D or 3D platform (0–5 ± 1%) were significantly lower than the stx⁺ (204P) strain (P < 0.05). Microscopic images of Trypan blue stained Vero cells confirmed cytotoxicity response of STEC cells (Figure 2D). Furthermore, an increase in the concentration of STEC exposure increased dye uptake and cell rounding and cell damage.

Next, we determined an optimal time required to show a definite difference in cytotoxicity between stx⁺ (204P) and stx^- (489) strains. At 2 h, there was no difference, while at 6 and 16 h post-infection (hpi), cytotoxicity values were found to be significantly (P < 0.05) different between stx⁺ and stx^- strains (Figure 2E). Furthermore, the 3D Vero showed a higher response than the 2D Vero toward stx⁺ strain and consequent microscopic images confirmed this observation (Figure 2F).

We verified the sensitivity of the 3D Vero cell platform to 27 STEC and non-STEC strains at 2, 6, and 16 hpi. Data show that all STEC strains tested (including serovars O157, O26, O121, O103, O45, and O145) showed significantly higher cytotoxicity (P < 0.05) than the non-STEC strains at 6 and 16 h while no discernable cytotoxicity response at 2 h (Table 2). Since no significant difference was observed among negative controls at each time point on a 2D or 3D platform (P < 0.05), a cutoff value was determined by taking three standard deviations above the mean of all negative controls (Table 3), and the assay threshold is established to be 15% (Zhang et al., 1999; Moodie et al., 2010). Microscopic images of Vero cells captured after infection with stx⁺ STEC (204P) and non-STEC bacterial isolates (Salmonella enterica serovar Tennessee, Salmonella Heidelberg, Salmonella Kentucky, Listeria innocua F4248, Listeria monocytogenes F4244, Citrobacter freundii, Serratia marcescens, Hafnia alvei, and non-pathogenic E. coli isolates) were in agreement with the cytotoxicity data (Supplementary Figure S1).

Our ultimate goal is to use an Stx-induction agent, such as antibiotics, together with the traditional bacterial enrichment broth before testing of food samples on a 3D cell-based assay platform without an intermediate sample-processing step. Antibiotics-induced Stx production has been reported before (Yee et al., 1993; Shimizu et al., 2003). Therefore, we wanted to examine the effect of several antibiotics, such as mitomycin C (2 μg/ml), ciprofloxacin (100 ng/ml), polymyxin B (2 mg/ml) and ultraviolet light (UV, exposure time 1.5 min) on Stx production and corresponding cytotoxicity on 2D and 3D Vero cell platforms. Data show that mitomycin C and UV induced the highest amount of Stx production than the ciprofloxacin or polymyxin B (Table 3). Toxin production was quantified in a dot blot immunoassay using anti-Stx1 and anti-Stx2 antibodies, and commercial Stx1 and Stx2 toxins as standards (Figure 3). Mitomycin C and UV induced 316 ± 475–906 ± 305 μg/ml Stx1 and 1687 ± 0.42–3298 ± 0.30 μg/ml Stx2, respectively. Ciprofloxacin induced 1,260 ± 12 μg/ml Stx1 and 314 ± 0.04 μg/ml Stx2 while polymyxin B induced undetectable levels of Stx1 and 417 ± 0.06 μg/ml Stx2. The crude toxin preparations (10× concentrated or unconcentrated) were then tested against 2D and 3D Vero cell platforms, and both showed positive effects. Notably, the effect was significantly higher on 3D than the 2D Vero cells (Figure 4A,B).

A 3-h pre-induction of stx⁺ strains (EDL933 and 204P) with mitomycin C (2 μg/ml) yielded 65 ± 4% and 81 ± 8% cytotoxicity, respectively. A similar trend was observed for ciprofloxacin, achieving 60 ± 11% and 71 ± 15% cytotoxicity, respectively, for these two strains. For uninduced bacteria, exposure to the same isolates had cytotoxicity values of 62 ± 9% and 76 ± 7%, respectively (Figure 4C). As expected, stx^- strains (489 and 490) showed negligible cytotoxicity (3–7 ± 2%) with or without antibiotics. We also tested the cytotoxic effects of all antibiotics, if any, without the bacteria and only polymyxin B caused cytotoxicity on Vero cells (Figure 4D). Cryo-scanning electron microscopy (SEM) images also showed cell damage induced by Stx preparation from 204P strain after mitomycin C induction (Figure 4E). Stx-induced cell damage is mostly characterized by porous honeycomb-like cell architecture. Collectively, these data indicate that mitomycin C, ciprofloxacin and the UV treatment induced Stx production, and the cell-free toxin preparation showed a significantly higher cytotoxic effect than the toxin preparation from the uninduced cells. However, differences in cytotoxicity were not significant for STEC cells with or without induction when tested against the 3D Vero cells; therefore, antibiotics were not used with our food sample testing experiment described below.

Escherichia coli 204P (stx⁺) and 489 (stx^-) – inoculated raw ground beef samples (3 samples × 9 technical replicates = 27), procured from local grocery stores were enriched in mTSB at 42°C for 15 h (FDA, 2001). An aliquot of the enriched food samples was centrifuged and resuspended in LB and then applied to 3D Vero cell platform and incubated for 6 h. Note, mTSB alone exhibited cytotoxic response against Vero, and thus this medium was replaced with LB before Vero assay (Supplementary Figure S1). A 15-h enrichment time was previously determined to yield a cell concentration of at least 10⁷ CFU/ml (Figure 2 and Table 1) required for positive cytotoxicity on the 3D Vero platform. The 3D Vero cell assay positively detected STEC form all artificially contaminated ground beef samples tested with cytotoxicity values varied from 36 to 61% (Figure 5A). Strain 489 (stx^-) served as a negative control for the assay and yielded 3–7% cytotoxicity, which is significantly lower than the stx⁺ samples (P < 0.05). Trypan blue staining of Vero cells confirmed dye uptake, cell rounding and characteristic cytotoxic effect, which was in good agreement with LDH-based cytotoxicity assay (Figure 5B). The presence of STEC in the enrichment broth (mTSB) was re-confirmed after subculturing on SMAC plates and multiplex PCR analysis of colonies for the presence of stx1 and stx2 genes (Figure 5C). These data clearly indicate that the 3D Vero cell platform is suitable for the detection of STEC cells from raw ground beef samples.

Frozen stock cultures (Table 1) were grown in BHI (Brain Heart Infusion) broth and then maintained on BHI agar (BHI, Becton Dickinson) plate for 1 month at 4°C. For fresh cultures, isolated colonies were inoculated into modified tryptone soy broth (mTSB) containing 0.15% bile salt, 0.4% dipotassium hydrogen phosphate, and 0.25% glucose (Becton Dickinson, Franklin Lakes, NJ, United States) at 42°C for 15 h with shaking at 120 rpm. Bacteria were also grown in Luria Bertani (LB) broth (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl), and EC broth (Becton Dickinson). To confirm and verify cultures, isolates were streaked on Modified Oxford Agar (MOX, Neogen, Lansing, MI, United States), Xylose Lysine Tergitol 4 (XLT4, Becton Dickinson), Sorbitol McConkey Agar (SMAC, Becton Dickinson) and RAPID’Enterobacteriaceae Medium (Bio-Rad, Hercules, CA, United States).

African green monkey kidney (Vero) cell line (ATCC CCL-81) were purchased from the American Type Culture Collections (ATCC) and maintained in Dulbecco’s modified Eagles medium (DMEM) (Sigma, St. Louis, MO, United States) with 10% fetal bovine serum. For all cytotoxicity assays, 3.2 × 10⁴ Vero or undifferentiated THP-1 (human monocyte cell line, ATCC) cells were grown as monolayers (2D) or suspensions (2D), respectively, in 48 well plates at 37°C with 5% CO₂ under humidity for 24 h. Vero cell monolayers were trypsinized with 0.25% of trypsin (Sigma) as described by the vendor and cell counts were determined by Trypan blue (0.4%) staining (Sigma). For 3D cell culture, cell suspensions were centrifuged and about 3.2 × 10⁴ Vero or undifferentiated THP-1 cells were embedded with collagen (0.7 mg/ml) containing 50 μl PBS, 113 μl collagen I, 2.5 μl NaOH, and 335 μl DMEM (Sigma) as described (Banerjee et al., 2008) in 48 well plates (Corning). DMEM or RPMI supplemented with 10% fetal bovine serum (FBS) was added after a 30-min incubation at 37°C in 5% CO₂ to allow for complete gelation.

Lactate dehydrogenase release was measured as described before (Roberts et al., 2001; Maldonado et al., 2005). THP-1 suspension cells were centrifuged (1,800 × g for 3 min), washed with serum-free RPMI and seeded at 3.2 × 10⁴ cells in 48 well plates with an initial volume adjusted to 200 μl/well. Vero cell monolayers and Vero embedded cells were washed with serum-free DMEM before the addition of DMEM (200 μl/well). Cells were exposed to 300 μl of STEC cells (∼10⁸ CFU/ml) for 2 to 16 h or serial dilutions of crude or commercial Stx1a, Stx2a, and Stx2c toxins (Toxin Technologies, Sarasota, FL, United States or BEI Resources, Manassas, VA, United States) for 16 h. Vero or THP-1 cell supernatants were collected after centrifugation (1,800 × g for 3 min) and loaded onto a sterile 96 well plate. Samples (50 μl) were mixed with LDH reaction reagent (50 μl) containing diaphorase, NAD⁺, sodium lactate, and iodophenyl-nitrophenyl-phenyltetrazolium chloride (Pierce Biotechnology, Rockford, IL, United States). Plates were incubated for 15–20 min in the dark at room temperature before taking measurements at absorbance 490/680 nm for LDH release using a microplate reader (Epoch, BioTek, Winooski, VT, United States). High controls treated with Triton X-100 (2%) for 45 min and low controls treated with serum-free DMEM+LB were used to calculate the percent cytotoxicity. Crude toxin preparations, STEC cells (204P) or commercial Stx toxins were used as positive controls for all LDH assays while cell-free supernatant or cells from stx^- strain (O157:H12 strain 489) were used as negative control. To visualize cytopathic effect from treatments, cells were fixed with formaldehyde (4% in PBS, Sigma) and stained with Trypan blue (4%, 1:2 dilution, Corning, Waltham, MA, United States) solution for 3 min. Images were captured at 400× magnification using an Olympus Inverted Microscope.

Four different growth media were evaluated to test for potential interference with the assay. Three hundred microliters of each LB broth, EC broth, and DMEM were added to 200 μl of DMEM before exposure to Vero cells and incubated for 16 h at 37°C in 5% CO₂. LDH assay of cell supernatants was performed as above.

For antibiotic induction, Vero cells were exposed to 500 μl of DMEM containing polymyxin B (50 μg/ml, Sigma), mitomycin C (24 ng/ml, Sigma), or ciprofloxacin (1.2 ng/ml, Sigma) for 16 h at 37°C in 5% CO₂. To verify the activity of antibiotics, Vero cell supernatants were tested against EDL933 (10⁸ CFU/ml) on BHI. Since polymyxin B can potentially interfere with the assay and can give a high background signal, mitomycin C and ciprofloxacin were used for further experiments. Since LB and DMEM had the least cytotoxicity response, these two media were then used as resuspension media to evaluate cytotoxicity response of stx+ positive and stx^- samples after a 3 h pre-induction with mitomycin C (2 μg/ml) and ciprofloxacin (100 ng/ml) at 37°C with shaking at 120 rpm. All samples were analyzed for LDH release after 16 h as described above. Vero cells were imaged under light microscopy (Olympus) at 400× magnification.

Crude toxins were prepared from stx+ O157:H7 strains (204P, EDL933, SEA13A72), O5:NM (467), and stx^- O157:H12 strain 489. Bacteria were grown in mTSB at 37°C for 15 h with shaking at 120 rpm. An aliquot of each culture was centrifuged and re-suspended to a final volume of 300 μl in LB to reach a concentration of 10⁸ CFU/ml and used immediately for LDH assay as above.

Overnight cultures were diluted to 1:50 in 50 ml LB and incubated for 3 h at 37°C with shaking at 120 rpm before the addition of mitomycin C (2 μg/ml) or ciprofloxacin (100 μg/ml, Sigma). For UV light treatment, overnight cultures were pelleted at 10,000 × g for 3 min, resuspended in LB, and dispensed into a sterile petri dish which was placed under a UV lamp (115 V, 68 Amps, UVP, Upland, CA, United States) for 1 min maintaining a distance of 12 inches between the lamp and sample. All cultures were further incubated for 5 h at 37°C with shaking at 120 rpm before collecting supernatant after centrifugation at 10,000 × g for 3 min. For polymyxin B treatment, 50 ml overnight culture was pelleted, re-suspended in 500 μl of PBS, supplemented with polymyxin (1 mg/ml), and further incubated for 30 min before collecting supernatant after centrifugation at 10,000 × g for 3 min. Pellets from all treatments were incubated with 100 μl of chloroform at room temperature for 30 min to improve Stx release and toxin yield. Cell-free supernatants of all treatments were concentrated to about 3–5 ml (about 10-fold concentrated) using Amicon Ultra-15 Centrifugal Filter Units using a 50 kDa cut-off cellulose membrane at 5,000 × g for 15 min (Millipore Sigma, Billerica, MA, United States).

Nitrocellulose membranes (0.2 μm, Biotrace NT, Thermo Scientific, Rochester, NY, United States) were pre-wetted in Tris-buffered saline with 0.1% tween 20 (TBST) for 1 min and placed in Bio-Dot apparatus (Bio-Rad, Hercules, CA, United States) under vacuum. Twofold diluted crude toxin samples and commercial Stx standards (Toxin Technologies, Sarasota, FL, United States, or BEI Resources, Manassas, VA, United States) were loaded into each well and application of gentle vacuum allowed fluid passage through the membrane. Membranes were washed with 100 μl of TBST/well, blocked with TBST containing 5% non-fat dry milk (NFDM), and incubated at room temperature for 45 min. Membranes were then washed with TBST three times at room temperature for 3 min with gentle agitation. Primary antibody solution containing antibody to Stx1-1 or Stx2-5 (Skinner et al., 2014, 2015; Wang et al., 2016) at dilutions 1:1000 in TBST with 5% NFDM was added to blots and allowed to incubate overnight at 4°C.

Twenty-seven bacterial isolates were screened for cytotoxicity potential after a 15 h enrichment in mTSB at 42°C with shaking. Vero cells (2D and 3D) were washed with serum-free DMEM (SF-DMEM) before exposure to bacteria (300 μl ≈ 10⁸ CFU/ml). The total volume of wells was adjusted to 500 μl with SF-DMEM before incubation at 37°C in 5% CO₂. To determine the limit of detection of the assay, overnight cultures of stx⁺ (204P) and stx^- (489) were resuspended in LB and serial dilutions of bacterial preparations were added to Vero cells and incubated for 16 h at 37°C in 5% CO₂. To determine optimal detection time, Vero cell cytotoxicity was assayed after exposure to stx⁺ (204P) and stx^- (489) at 10⁸ CFU/ml) for 2 h, 6 h, and 16 h. Salmonella, Listeria spp., Citrobacter, Hafnia, and Serratia spp. were also tested to investigate assay specificity. All samples were centrifuged at 1,800 × g for 3 min and assayed for LDH. LDH values were acquired from three independent sets of experiments and analyzed in duplicates. After LDH analysis, cells were fixed with formaldehyde (4%) before Trypan blue (4%, 1:2 dilution) staining for 3 min to determine Vero cell morphology after treatments at 400× magnification.

Minimally dehydrated collagen gel embedded Vero cells were mounted on slotted Gatan holders with 1 mm of the sample above the holder. Samples were cryo-transferred after plunge freezing with nitrogen slush into the Gatan preparation chamber, which is held under high vacuum. Frozen samples were fractured with a cold scalpel in the Gatan Alto 2500 preparation chamber at -185°C (Pleasanton, CA, United States) before being transferred into the SEM chamber. Samples were mounted on the cryo-stage set for -90°C sublimation and imaged until a structure was observed before sputter coating for 120 s using a platinum target at -185°C. SEM Cryo-stage was lowered to -140°C during sputter coating and was reinserted onto the NovaNano SEM cryo-stage before capturing final images of fractured surfaces. An accelerating voltage of 5 kV and spot size 3 with a working distance of approximately 5 mm at magnifications ranging ×1,000 and ×10,000 were used in the NovaNano SEM.

3D Vero cell platform was used to screen and detect STEC from artificially contaminated ground beef samples. Three brands of ground beef (5 lb each), purchased from local grocery stores, were artificially inoculated with stx⁺ 204P (O157:H7) and stx^- 489 (O157:H12) strain under aseptic conditions. Samples were processed following the USDA guidelines for screening STEC in ground beef samples (FSIS, 2016). In brief, ground beef samples (32 ± 3.2 g each, 3 samples × 9 technical replicates = 27) were placed in sterile strainer bags (Seward, Islandia, NY, United States) inoculated with 100 μl of 204P (6.7 × 10² ± 2.1 × 10² CFU/ml) and 489 (7.4 × 10² ± 5.9 × 10² CFU/ml). Samples were then diluted with 97 ± 1.9 ml of mTSB and hand massaged for 1 min to homogenize before incubating at 42°C for 15 h with shaking at 120 rpm. To determine a final cell density after enrichment, samples were serially diluted (1:10 in PBS) and plated on SMAC plates and incubated at 37°C for 24 h. For 3D Vero cell assay, 300 μl samples were centrifuged at 10,000 × g for 3 min and resuspended in the same volume of LB before exposure to 2D or 3D Vero cells (3.2 × 10⁴ cells) in a 48-well plate format. After incubation at 37°C for 6 h in 5% CO₂, samples were centrifuged at 1,800 × g for 3 min and analyzed for LDH release. Uninoculated meat samples were also tested to check for background response and interference. Suspected colonies identified on the SMAC were confirmed by PCR, targeting stx1 and stx2 genes.

DNA templates were extracted following a boiling and colony PCR method (Feng et al., 2002; Hoang Minh et al., 2015). Stx-specific primers targeting stx1 (5′-ATCCTATTCCCGGGAGTTTACG-3′ and 5′-GCGT CATCGTATACACAGGAGC-3′) and stx2 (5′-CACCAGACAATGTAACCGCTG-3′ and 5′-TIACCATTTCAGTACCTTCTGGTAA-3′) genes were used to confirm the Stx presence in the 3D Vero-identified STEC positive samples. DNA templates from non-STEC isolates, uninoculated meat samples, and water were used as negative controls. The reaction mixture (25 μl) contained 200 ng of DNA template, 1× GoTaq Flexi buffer, 1 U of GoTaq, 2.5 mM MgCl₂, 200 μM deoxynucleoside triphosphates (dNTP), and 4 μM of stx1 and stx2 forward and reverse primers. The PCR was carried out with an initial denaturation step at 94°C for 120 s, followed by 35 cycles of denaturation at 95°C for 60 s, annealing at 57°C for 60 s, 72°C for 60 s and a final extension at 72°C for 35 s.

Values are presented as mean ± SD. GraphPad Prism (version 6) software was used to perform ordinary one-way or two-way ANOVA followed by Tukey’s multiple comparisons analysis to evaluate whether there are a significant difference among media (LB, DMEM, EC, mTSB) and antibiotics (mitomycin C and ciprofloxacin) induced cytotoxicity response. ImageJ (NIH) was used to assess Stx yield after induction. SAS software was used to perform 2 × 2 × 3 ANOVA followed by Tukey’s post hoc analysis to assess the interaction between stx⁺ and stx^- strains, dimension (2D vs. 3D), and time of incubation (2 h, 6 h, and 16 h).