Received 8 August 2004/Accepted 1 November 2004

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Apr. 2005, p. 1856 1864 Vol. 71, No. 4 0099-2240/05/$08.00 0 doi:10.1128/aem.71.4.1856 1864.2005 Copyright 2005, American Society for Microbiology. All Rights Reserved. Development of an Immunomagnetic Bead-Immunoliposome Fluorescence Assay for Rapid Detection of Escherichia coli O157:H7 in Aqueous Samples and Comparison of the Assay with a Standard Microbiological Method Thomas R. DeCory, 1 Richard A. Durst, 1 Scott J. Zimmerman, 2,3 Linda A. Garringer, 2 Gary Paluca, 2 Heleen H. DeCory, 1 and Richard A. Montagna 4 * Department of Food Science and Technology, Cornell University, Geneva, 1 Clinical Center, Public Health Laboratory, Department of Health, County of Erie, 2 and Department of Microbiology, School of Medicine, and School of Public Health and Related Professions, State University of New York at Buffalo, 3 Buffalo, and Innovative Biotechnologies International, Inc., Grand Island, 4 New York Received 8 August 2004/Accepted 1 November 2004 The objective of this study was to develop and optimize a protocol for the rapid detection of Escherichia coli O157:H7 in aqueous samples by a combined immunomagnetic bead-immunoliposome (IMB/IL) fluorescence assay. The protocol consisted of the filtration or centrifugation of 30- to 100-ml samples followed by incubation of the filter membranes or pellet with anti-e. coli O157:H7 immunomagnetic beads in growth medium specific for E. coli O157:H7. The resulting E. coli O157:H7-immunomagnetic bead complexes were isolated by magnetic separation, washed, and incubated with sulforhodamine B-containing immunoliposomes specific for E. coli O157:H7; the final immunomagnetic bead-e. coli O157:H7-immunoliposome complexes were again isolated by magnetic separation, washed, and lysed with a n-octyl- -D-glucopyranoside to release sulforhodamine B. The final protocol took less than 8htocomplete and had a detection limit of less than 1 CFU of E. coli O157:H7 per ml in various aqueous matrices, including apple juice and cider. To validate the protocol at an independent facility, 100-ml samples of groundwater with and without E. coli O157:H7 (15 CFU) were analyzed by a public health laboratory using the optimized protocol and a standard microbiological method. While the IMB/IL fluorescence assay was able to identify E. coli O157:H7-containing samples with 100% accuracy, the standard microbiological method was unable to distinguish E. coli O157:H7-spiked samples from negative controls without further extensive workup. These results demonstrate the feasibility of using immunomagnetic beads in combination with sulforhodamine B-encapsulating immunoliposomes for the rapid detection of E. coli O157:H7 in aqueous samples. Among the five recognized classes of enterovirulent Escherichia coli (collectively referred to as the EEC group), which often cause gastroenteritis in humans, the enterohemorrhagic E. coli strain O157:H7 was first recognized as a cause of illness in humans in 1975. Since then, numerous outbreaks of E. coli O157:H7-induced disease have occurred, with most associated with the consumption of undercooked ground beef (11, 24). Outbreaks have also been associated with drinking unpasteurized milk (15); swimming in or drinking sewage-contaminated water (25); and consumption of contaminated apple juice or cider (4, 8, 9), fruits, vegetables, or salads (19, 30). The pathogenicity of E. coli O157:H7 appears to be associated with a number of virulence factors, including the production of several cytotoxins collectively referred to as Shiga-like toxins (SLTs) because SLT-1 of E. coli O157:H7 closely resembles the Shiga toxin of Shigella dysenteriae type 1 (21). An outbreak of E. coli O157:H7 infection, including at least two deaths, in the Albany, N.Y., area has further revealed the susceptibility of * Corresponding author. Mailing address: Innovative Biotechnologies International, Inc., 335 Lang Blvd., Grand Island, NY 14072. Phone: (716) 773-4232. Fax: (716) 773-4257. E-mail: rmontagna @ibi.cc. Present address: Celltech Americas, Inc., Rochester, NY 14623. the human population to infection under circumstances whereby contaminated fecal material made its way into a water supply (10, 13, 20). The contaminated water subsequently was ingested at a county fair, not only causing a high infection rate but also raising concerns about the use of untreated groundwater for human consumption. A major problem associated with E. coli O157:H7 contamination of food and water-based products for human consumption is the inability to rapidly, accurately, and inexpensively detect such contamination. Two major approaches currently being utilized are based on either immunoassay formats (12, 16) or gene amplification methods, using PCR approaches (14, 18). While immunoassay and PCR-based tests offer exquisite sensitivity, they both require culture-based enrichment of the target organism in order to increase it to detectable levels at the required sensitivities. Antibody-tagged magnetic beads have also been successfully used to detect E. coli O157:H7, but a culture-based enrichment step is still required in the assay format (3, 31). While culture-based enrichment steps are able to enhance the signal resulting from immunologically based tests, the use of liposome nanovesicles encapsulating fluorescent dyes, by virtue of their signal amplification capabilities, can dramatically reduce the time required for this step. Liposome 1856

VOL. 71, 2005 E. COLI O157:H7 IMB/IL FLUORESCENCE ASSAY 1857 nanovesicles represent a significant improvement in the diagnostic arsenal against human pathogens because of two main features: (i) the ability to bind various biorecognition elements to their exterior, thereby permitting an array of assay formats, and (ii) the ability to encapsulate high concentrations of detectable marker molecules, including fluorescent molecules. Unlike classical fluorescence assays whereby a specific biorecognition element, such as an antibody or nucleic acid probe, is coupled to a single signal-generating molecule, the use of liposome nanovesicles provides an opportunity to couple the same biorecognition element to a vehicle that carries hundreds of thousands of signal-generating molecules. The objective of this study was to develop an assay for the detection of E. coli O157:H7 using immunomagnetic beads (IMBs) specific for E. coli O157:H7 in combination with sulforhodamine B-containing liposome nanovesicles coated with anti-e. coli O157:H7 antibodies, thereby applying the advantages of liposome nanovesicles to an immunoassay format. We report key findings in the development and optimization of the final immunomagnetic bead-immunoliposome (IMB/IL) fluorescence assay, present data on the detection of E. coli O157:H7 in various aqueous matrices, and report on the preliminary validation of the final protocol at a public health laboratory in comparison to a standard microbiological method. MATERIALS AND METHODS Materials. Dipalmitoylphosphatidylcholine, dipalmitoylphosphatidylethanolamine (DPPE), dipalmitoylphosphatidylglycerol, and the Mini Extruder were purchased from Avanti Polar Lipids, Inc. (Alabaster, Ala.). Cholesterol, triethylamine, chloroform, methanol, HEPES, EDTA, dimethyl sulfoxide, Sephadex G-50, Sepharose CL-4B, sulforhodamine B (SRB), Tween 20, and n-octyl- -Dglucopyranoside were purchased from Sigma Chemical Co. (St. Louis, Mo.). N-Succinimidyl-S-acetylthiopropionate (SATA), N-[ -maleimidoundecanoyloxy]sulfosuccinimide ester (sulfo-kmus), hydroxylamine hydrochloride, N-ethylmaleimide, and blocker casein (1% casein in Tris-buffered saline) were purchased from Pierce (Rockford, Ill.). A DispoDialyzer membrane (molecular weight cutoff, 15,000) was purchased from Spectrum Lab, Inc. (Rancho Dominguez, Calif.). Durapore membrane filters (0.22- and 0.45- m pore sizes), potassium phosphate, and 12- by 75-mm polystyrene snap cap tubes were purchased from Fisher Scientific (Pittsburgh, Pa.), while 8- by 40-mm glass vials were purchased from National Scientific Company (Duluth, Ga.). The 0.45- m-poresize gridded membrane filters were purchased from Pall Gelman Sciences (Ann Arbor, Mich.), and the 0.2- and 0.4- m-pore-size Nuclepore Track-Etch membranes were purchased from Whatman (Clifton, N.J.). Sorbitol-containing Mac- Conkey agar (SMAC) was purchased from Becton Dickinson (Sparks, Md.). Affinity-purified polyclonal goat anti-e. coli O157:H7 antibodies and heatkilled E. coli O157:H7 positive control antigen were purchased from Kirkegaard & Perry Laboratories Inc. (Gaithersburg, Md.). E. coli O157:H7 immunomagnetic beads were purchased from Neogen Corp. (Lansing, Mich.) and Dynal Biotech Inc. (Brown Deer, Wis.). Medi-8 accelerated-growth medium for E. coli O157:H7 was purchased from Eichrom Technologies (Darien, Ill.). Tryptic soy broth (TSB) and tryptic soy agar (TSA) were purchased from Difco Laboratories (Detroit, Mich.). E. coli O157:H7 (ATCC 43895), an E. coli laboratory strain (ATCC 25922), and a Bacillus subtilis strain (ATCC 6633) were purchased from the American Type Culture Collection (Manassas, Va.). The magnetic separation stand was purchased from Promega (Madison, Wis.). The handheld Aquafluor fluorometer was purchased from Turner Designs (Sunnyvale, Calif.). Pasteurized apple juice and pasteurized apple cider were purchased from Wegman s Food Markets (Rochester, N.Y.). Bottled spring water was purchased from Aqua Valley (Edmeston, N.Y.) or Clearwater Springs (Syracuse, N.Y.), tap water was acquired from the City of Geneva, N.Y., and groundwater used in optimization experiments was collected from a nonchlorinated well in Scottsville, N.Y. Groundwater for experiments conducted at the Erie County Public Health Laboratories were collected from nonchlorinated wells in Hamburg, N.Y. Preparation of anti-e. coli O157:H7 antibody-tagged immunoliposomes. Liposome nanovesicles were prepared by a method involving hydration and freezing and thawing and extrusion (1). Briefly, 28.9 mg of dipalmitoylphosphatidylcholine, 3.7 mg of dipalmitoylphosphatidylglycerol, 16.8 mg of cholesterol, and 0.5 ml of a solution of DPPE-ATA (thiolated phospholipid) in chloroform were added to 10 ml of chloroform and 4 ml of methanol (the DPPE-ATA was prepared in advance by reacting 5 mg of DPPE and 3.5 mg of SATA in 1.0 ml of 0.7% [vol/vol] triethylamine in chloroform for 30 min at room temperature, followed by evaporation under vacuum and reconstitution in 1.0 ml of chloroform). After being mixed thoroughly, the solution was evaporated under vacuum on a rotary evaporator, resulting in a thin lipid film. To this was added 4 ml of a 150 mm solution of SRB in 20 mm HEPES buffer (ph 7.5), and the resulting solution was subjected to five freeze-thaw cycles as described previously (1) to produce large multilammelar liposome nanovesicles. The liposomes were subsequently extruded sequentially through a 0.4- and 0.2- m Nuclepore Track- Etch membranes, and any unencapsulated SRB was removed by applying the liposomes to a Sephadex G-50 size exclusion column equilibrated with Trisbuffered saline (ph 7.4, same osmolality as SRB encapsulant). Preparation of sulforhodamine-containing liposomes via this procedure generally resulted in the formation of vesicles with a mean ( standard deviation [SD]) diameter of 203 14.8 nm, as shown by particle size analysis (LS130 particle size analyzer; Coulter, Hialeah, Fla.). Assuming the formation of unilammelar vesicles and a width of 2 nm for the phospholipid bilayer, the amount of SRB contained in each liposome is approximately 6 10 10 nmol. Anti-E. coli O157:H7 antibody-tagged liposomes were prepared by reacting activated liposomes with sulfo-kmus-derivatized antibodies. Briefly, anti-e. coli O157:H7 antibodies were derivatized by reacting 7.2 l of a 1/50 (wt/vol) solution of sulfo-kmus in dimethyl sulfoxide with 3 mg ( 20 nmol) of anti-e. coli O157:H7 antibodies in 0.8 ml of 0.05 M potassium phosphate and 1 mm EDTA (ph 7.8) for 3hinthedark at room temperature. The reaction was quenched by the addition of 12 l of 0.5 M Tris (ph 7.8), and the derivatized antibody was dialyzed overnight in a DispoDialyzer (molecular weight cutoff, 15,000) against Tris-buffered saline (ph 7.4) at 4 C. Based on the results of a Bartlett assay (2) conducted on the liposome nanovesicles prepared as described above and a desire to tag 0.4 mol of antibody per 100 mol of total lipid, 250 l (125 mol) of 0.5 M hydroxylamine hydrochloride in 0.1 M HEPES and 25 mm EDTA (ph 7.5) were added to 2.5 ml of liposome solution (equal to 100 nmol of surface DPPE-ATA) and allowed to react at room temperature for 2 h in the dark. The ph of the reaction was adjusted to 7.0 with 0.5 M potassium phosphate, and the dialyzed antibodies were added. The solution was allowed to react at room temperature in the dark for 3 h and then quenched by the addition of 20 l (2 mol) of N-ethylmaleimide in phosphate-buffered saline (PBS; ph 7.0). Finally, unconjugated antibodies were separated from antibody-tagged liposomes by means of a Sepharose CL-4B size exclusion column equilibrated with Trisbuffered saline (ph 7.4, same osmolality as SRB encapsulant). E. coli O157:H7 growth conditions. E. coli O157:H7 was cultured in TSB for 18 h at 37 C with shaking. Cultures were serially diluted with TSB, and cell concentrations were determined by applying 100 l of the 10 6 and 10 7 dilutions to 10 TSA plates each and visually counting colonies after incubating the plates for 24 h at 37 C. Typically, a 100- l aliquot of the 10 6 dilution of the overnight culture resulted in 100 to 200 CFU while a 100- l aliquot of a 10 7 dilution resulted in 10 to 20 CFU. Detection of E. coli O157:H7 in water samples and apple juice and cider. Water samples (100 ml) were spiked with 100 to 1,000 l ofa10 7 serially diluted overnight culture of E. coli O157:H7 for a final concentration range of 0.1 to 0.2 CFU/ml to 1 to 2 CFU/ml. Spiked and unspiked samples (used as negative controls) were filtered through 0.45- m-pore-size, 25-mm-diameter Durapore membrane filters with a water aspirator (29 in of Hg, 30 to 40 s/100 ml of water). The membrane filters were placed in 12- by 75-mm polystyrene snap cap tubes to which were added 20 l ofe. coli O157:H7 immunomagnetic beads (1:10 dilution in blocker casein) and 2 ml of Medi-8 accelerated-growth medium for E. coli O157:H7 prepared according to the manufacturer s directions. Apple juice samples (100 ml) were prepared identically to the water samples (above) except that samples were spiked with 100 to 1,000 l of a serially 10 5 -diluted stock solution of E. coli O157:H7 in apple juice (the stock solution was prepared in advance by adding 20 ml of an overnight culture to 200 ml of apple juice and placing the solution in a refrigerator [4 C] for 5 to 6 days). Apple cider samples (30 ml) were likewise spiked with 100 to 1,000 l of a serially 10 5 -diluted stock solution of E. coli. However, spiked and unspiked samples were centrifuged at 3,800 g for 15 min at 4 C rather than filtered. Following centrifugation, the supernatant was removed, and 2 ml of Medi-8 acceleratedgrowth medium for E. coli O157:H7 was added to each pellet. After brief vortexing (5 s), the resulting suspensions were transferred to 12- by 75-mm

1858 DECORY ET AL. APPL. ENVIRON. MICROBIOL. polystyrene snap cap tubes to which was added 20 l ofe. coli 157:H7 immunomagnetic beads (1:10 dilution in blocker casein). The snap cap tubes containing the filter membrane or pellet (apple cider), immunomagnetic beads, and Medi-8 accelerated-growth medium were placed on a Labquake rotating mixer (8 rpm) in an incubator at 42 C and rotated for 4 h to both enrich and capture the E. coli O157:H7 with the immunomagnetic beads. At the end of the simultaneous enrichment and immunomagnetic capture steps, the tubes were removed from the incubator and positioned in the magnetic separation stand. After 3 min (10 min for apple cider), the supernatant was aspirated from the tubes with a vacuum aspirator and 2 ml of blocker casein was added to the magnetic bead-e. coli O157:H7 pellet. The tubes were returned to the rotating mixer and rotated at room temperature for 5 min (10 min for apple cider). After 5 min, the tubes were once again positioned in the magnetic separation stand, and the magnetic separation was repeated and the supernatant was removed as described above. Following the second magnetic separation (above), the filter membranes were removed (for water and apple juice samples) and 200 l of the anti-e. coli O157:H7 immunoglobulin G-tagged liposomes (1:40 dilution in blocker casein) was added to each tube. The tubes were replaced on the rotating mixer and rotated at room temperature for 1 h, followed by magnetic separation and supernatant removal (as described above). The resulting pellet was washed again with 2 ml of blocker casein for 5 min as described above and subjected to magnetic separation as described above, and finally 200 l of30mmn-octyl- - D-glucopyranoside in Tris-buffered saline was added to each of the tubes, followed by vigorous vortexing for 5 s. After a final magnetic separation, the resulting supernatant containing released SRB was transferred to 8- by 40-mm glass vials, and the fluorescence of each sample was measured with a handheld Aquafluor fluorometer (Turner Design) at an excitation wavelength of 540 nm and an emission wavelength of 570 nm. To enumerate the actual number of E. coli O157:H7 cells added to the various water and apple juice and cider samples, 10 replicate 100- l samples of the dilution used to spike each sample matrix were plated onto TSA plates. The plates were incubated overnight at 37 C ( 20 h), and colonies were counted the next day with the aid of a touch counter (Fisher Scientific). Validation of the assay at the Erie County Regional Public Health Laboratories and comparison to a standard microbiological assay. To both validate the IMB/IL fluorescence assay and determine the preliminary specificity of the assay, 100-ml samples of commercial spring water were spiked with various concentrations of B. subtilis, an E. coli laboratory strain, or E. coli O157:H7 at Cornell University and transferred along with unspiked samples to the Erie County (N.Y.) Public Health Laboratories for a blinded analysis. Laboratory staff at the public health laboratory were given detailed protocols with minimal training beforehand. Samples were processed by the IMB/IL fluorescence assay by staff at the public health laboratory as described above for water matrices with the exception that the enrichment-capture step was carried out for 3.5 h and the incubation with anti-e. coli O157:H7 liposomes was carried out for only 1/2 h. In addition, negative and positive controls were included in the analysis. The positive control consisted of a 1-ml sample of a 1:10,000 dilution of heat-killed E. coli O157:H7 positive control antigen in Tris-buffered saline ( 7 10 5 cells/ml) incubated directly with the immunomagnetic beads and processed as described above; the negative control consisted of a 50-ml sample of Tris-buffered saline, filtered through a 0.45- m Durapore membrane filter, which was further incubated with immunomagnetic beads and processed as described before. The IMB/IL fluorescence assay was also compared to a standard microbiological method utilized in the Erie County Public Health Laboratories. Briefly, 1-liter samples of groundwater were collected by the staff at the public health laboratory. From each liter, three 100-ml samples were spiked with 100 l ofa 10 7 overnight dilution of E. coli O157:H7 containing approximately 15 CFU, and three 100-ml samples were prepared without spiking. One set of spiked and unspiked samples from each 1-liter sample was assayed by the IMB/IL fluorescence assay, one set was assayed by a standard microbiological assay employed by the public health laboratory, and the third set was assayed by an additional assay (modeled after the procedure of Bopp et al. [5]) intended to confirm the absence of E. coli O157:H7 in the groundwater samples utilized in this validation experiment. Samples assayed by the IMB/IL fluorescence assay were processed as described above for water matrixes with the exception that the enrichmentcapture step was carried out for 5 h and the incubation with anti-e. coli O157:H7 immunoliposomes was carried out for 30 min. Positive and negative controls were included as described above. The standard microbiological method employed at the Erie County Public Health Laboratories consisted of filtration of the spiked and unspiked samples with 0.45- m-pore-size gridded membrane filters onto SMAC plates. The SMAC plates were incubated for 18 to 24 h at 35 C, and the resulting sorbitol-nonfermenting colonies were counted under 10 FIG. 1. Summary of the E. coli O157:H7 assay protocol. magnification with the aid of a Quebec colony counter. Any sorbitol-nonfermenting colonies were considered suspect because most O157 Shiga toxin-producing E. coli isolates do not ferment the carbohydrate D-sorbitol overnight. The third set of spiked and unspiked groundwater samples from each 1-liter sample was processed as follows. Samples were filtered, and any E. coli O157:H7 organisms present were captured as described for the IMB/IL fluorescence assay except that samples were filtered with 0.22- m-pore-size 25-mm-diameter Durapore membrane filters prewetted with PBS 0.02% Tween 20, and the filter membranes were incubated with anti-e. coli O157:H7 immunomagnetic beads in the presence of 1 ml of PBS 0.02% Tween 20 for 2 h at room temperature. The resulting immunomagnetic bead-e. coli O157:H7 complexes were then isolated by magnetic separation and applied to SMAC plates after suspension in 100 l of PBS 0.02% Tween 20. Finally, any sorbitol-nonfermenting colonies appearing after 18 to 24 h at 35 C were tested by the API 20E biochemical test method (6, 7, 29) (biomérieux, Inc., Durham, N.C.) according to the manufacturer s instructions. The API 20E biochemical test method is a standardized identification system for Enterobacteriaceae and other nonfastidious, gram-negative rods that utilizes 21 miniaturized biochemical tests. It can confirm the presence of E. coli but not the exact serotype. Statistics. Descriptive statistics (means SD) were reported where appropriate. t tests were used to compare data presented in Figures 5 to 8. Linear regression analyses were conducted using individual data points rather than means (Prism, version 3.02; GraphPad Prism Software, Inc.). Standard deviations for signal to noise and percent control in the experiment testing different filters were calculated using the propagation of error equation z/z [( x/x) 2 ( y/y) 2 ] 1/2, where z, x, and y are the SDs of z, x, and y, respectively. Finally, the limits of detection for E. coli O157:H7 in various aqueous matrices were determined by interpolation from the linear regression equations fitted to the respective dose-response curves with the mean (plus 3 SDs) fluorescence signal of the blank sample. RESULTS Assay optimization. The protocol for the IMB/IL fluorescence assay for E. coli O157:H7 is illustrated in Fig. 1. Briefly, the final protocol consisted of the filtration or centrifugation of 30- to 100-ml aqueous samples followed by incubation of the filter membranes or pellet with anti-e. coli O157:H7 immunomagnetic beads in Medi-8 accelerated-growth medium for E. coli O157:H7 for 3 to 4 h. The resulting E. coli O157:H7-

VOL. 71, 2005 E. COLI O157:H7 IMB/IL FLUORESCENCE ASSAY 1859 FIG. 2. Release of fluorescence signal with n-octyl-d-glucopyranoside. au, arbitrary units. immunomagnetic bead complexes were isolated by magnetic separation, washed, and incubated with SRB-containing immunoliposomes specific for E. coli O157:H7; the final immunomagnetic bead-e. coli O157:H7-immunoliposome complexes were isolated by magnetic separation once again, washed, and then lysed with n-octyl- -D-glucopyranoside to release the fluorescent marker. The release of fluorescence signal upon addition of n-octyl- -D-glucopyranoside to immunoliposomes is shown in Fig. 2. In this experiment, the addition of 200 lof30 mm n-octyl- -D-glucopyranoside to 5 l of immunoliposomes ( 10 9 liposomes) diluted in 1 ml of Tris-buffered saline resulted in a 3,000-fold increase in fluorescence signal. The original protocol called for a brief culture enrichment step followed by incubation of the enriched culture with immunomagnetic beads. During early experiments, it was determined that the total assay time could be reduced by simultaneously incubating the immunomagnetic beads specific for E. coli O157:H7 in the enrichment medium (data not shown), thereby allowing the enriched organisms to be captured by the immunomagnetic beads during the enrichment period. This combined enrichment-capture step has the added benefit of providing information on the viability of the targets (i.e., nonviable organisms would not propagate and increase in number). Because our objective was to develop an assay system that could easily be used in field settings, initial experiments focused on the choice of filter membrane utilized in the filtration of samples. Three different types of filter materials and/or pore sizes were investigated including 0.45- m nitrocellulose membranes, and 0.45- and 0.22- m Durapore membranes. Briefly, triplicate 1-ml samples of a 10 7 dilution of an overnight culture of E. coli O157:H7 (containing approximately 100 to 200 CFU/ml) and triplicate 1-ml samples containing no E. coli O157:H7 were filtered through the membranes via vacuum filtration. The resulting filters were then incubated with 2 ml of TSB and 20 l of immunomagnetic beads for 3 h prior to immunomagnetic separation, incubation with immunoliposomes, and lysis of the final immunomagnetic bead-e. coli O157:H7-immunoliposome complex to release the fluorescent marker. Control samples consisted of triplicate 1-ml samples of the 10 7 dilution incubated directly with the immunomagnetic beads and 1 ml of TSB without filtration. The percentages of FIG. 3. Comparison of filter membranes used in the filtration of E. coli O157:H7 samples in the IMB/IL fluorescence assay. Control was a nonfiltered E. coli O157:H7 sample. Data represent the means SDs. CV, coefficient of variation. control signal (fluorescence signals of filtered spiked sample/ fluorescence signal of the nonfiltered control sample 100), signal-to-noise ratios (fluorescence signals of filtered spiked samples/fluorescence signal of filtered unspiked sample), and percentages of coefficient of variation (SDs/means) of the final fluorescent samples are shown in Fig. 3. The 0.45- m Durapore filter had the greatest percent control signal (101% 13.1%), the greatest signal-to-noise ratio (47.0 4.94), and the least percent coefficient of variation (5.90%) of the three filters tested and was therefore chosen for the final assay format. Once the optimal filter membrane was identified, we focused our attention on the choice and optimal concentration of immunomagnetic beads to utilize. Our goal was to not only further concentrate the test sample but also selectively isolate E. coli O157:H7 for detection and quantitation. Two different commercially available immunomagnetic beads specific for E. coli O157:H7 were evaluated, namely, Dynal and Neogen beads, containing 6.6 10 6 and 8.0 10 6 beads per 20 l, respectively. While similar dose-response fluorescence signals were obtained after incubating 20 l of each of the undiluted immunomagnetic beads with increasing amounts of E. coli

1860 DECORY ET AL. APPL. ENVIRON. MICROBIOL. FIG. 4. Signal-to-noise ratios of undiluted Neogen immunomagnetic beads and various dilutions thereof in an E. coli O157:H7 IMB/IL fluorescence assay. Data represent the means of two determinations. FIG. 5. Effect of enrichment-capture media and temperature on fluorescence signal in the detection of E. coli O157:H7 in an IMB/IL fluorescence assay. Data represent the means SDs. EC broth, E. coli broth; Medi-8, Medi-8 accelerated-growth medium for E. coli O157: H7. au, arbitrary units. O157:H7 (data not shown), Neogen beads were chosen for their economy. Further analysis of the Neogen beads demonstrated that diluting the stock immunomagnetic beads by a factor of 10 resulted in a signal-to-noise ratio 40-fold greater than that achieved with undiluted stock immunomagnetic beads. Figure 4 presents the signal-to-noise ratios of undiluted immunomagnetic beads and dilutions thereof in Tris-buffered saline in the E. coli O157:H7 IMB/IL assay. Briefly, 1-ml dilutions of heat-killed E. coli O157:H7 in Tris-buffered saline containing approximately 7 10 5 cells were incubated directly in duplicate with 20 l of immunomagnetic beads at the concentrations indicated for 2 h. The resulting IMB-E. coli complexes were subjected to immunomagnetic separation and incubated for 1 h with immunoliposomes, followed by further immunomagnetic separation and lysis of the final IMB-E. coli O157:H7-immunoliposome complex to release the fluorescent marker. Again the 1:10 dilution displayed the greatest signalto-noise ratio, due to the substantial decrease in background signal relative to that for undiluted beads. However, diluting the immunomagnetic beads further resulted in a drop-off in fluorescence signal, possibly due to the inability of the beads to capture all of the E. coli in the heat-killed sample. Therefore, to both reduce the cost of materials required to perform the assay and obtain the best signal-to-noise ratio, we elected to use a 1:10 dilution of the Neogen immunomagnetic beads in the optimized assay. Several enrichment-capture media were tested in the IMB/IL assay, including TSB, 1.5 TSB, Medi-8 acceleratedgrowth medium for E. coli O157:H7, and E. coli broth. Triplicate 1-ml samples of a 10 8 dilution of an overnight sample of E. coli O157:H7 (containing approximately 15 CFU) were incubated directly with 2 ml of each of the media for 4hinthe presence of immunomagnetic beads at either 37 or 42 C. Figure 5 presents the means SDs of the final fluorescence signals following completion of the remaining protocol steps using each of these media. The highest fluorescence signal was noted for samples incubated with the Medi-8 acceleratedgrowth medium for E. coli O157:H7 at 42 C (16.8 2.3), and samples incubated with Medi-8 media at this temperature gave significantly higher fluorescence readings than any other medium at either 37 or 42 C (P 0.05). The effect of enrichment time on the final signal in the detection of E. coli O157:H7 in IMB/IL fluorescence assay is shown in Fig. 6. For the purpose of this experiment, 1-ml samples of a 10 7 overnight dilution of E. coli O157:H7 ( 150 CFU/ml) were filtered directly through 0.45- m Durapore membranes and subsequently incubated with 2 ml of Medi-8 medium at 42 C for different time periods in the presence of immunomagnetic beads prior to the next assay steps. As expected, the resulting data showed that E. coli O157:H7 samples underwent exponential growth during enrichment, with a mean doubling time of 27.7 min (95% confidence interval 23.9 to 32.9 min), consistent with reported literature values (26). The fluorescence signal achieved with the spiked sample with a 2-h enrichment was significantly different (P 0.0104) from that for the unspiked (negative) control, while the fluorescence signal achieved with the 3-h enrichment was highly significantly different from the control (P 0.0001), indicating that 3hwas more than sufficient for the detection of 1.5 CFU/ml in a real-world sample (assuming a 100-ml real-world sample had been analyzed). Based on these results, an incubation period of 4 h was subsequently chosen as adequate for the detection of E. coli O157:H7 in the range of 0.15 to 1.5 CFU/ml. During the course of the assay optimization, other assay FIG. 6. Effect of enrichment-capture time on fluorescence signal in the detection of E. coli O157:H7 in an IMB/IL fluorescence assay. Data represent the means SDs. au, arbitrary units.

VOL. 71, 2005 E. COLI O157:H7 IMB/IL FLUORESCENCE ASSAY 1861 FIG. 7. Dose-response curves for groundwater, tap water, and spring water (A) and apple juice and apple cider (B) spiked with E. coli O157:H7 and assayed with an IMB/IL fluorescence assay. Data represent the mean ( SD) fluorescence signals. au, arbitrary units. parameters, such as the number and volume of wash steps, formulation of washing buffer, times of various steps, and choice of fluorescence detector were evaluated (data not shown). The final assay format used for the remainder of this study is as outlined in Materials and Methods. Evaluation of the assay with different aqueous matrices. Using the optimized method, we evaluated a number of different aqueous matrices. Three different types of water (bottled spring water, tap water, and groundwater) were spiked with various dilutions of an overnight culture of E. coli O157:H7 (representing approximately 15 to 150 CFU) to test the sensitivity of the assay in each water matrix. The mean ( SD) fluorescence signals for samples of spring water, tap water, and groundwater spiked with E. coli O157:H7 and assayed by the optimized IMB/IL fluorescence assay are shown in Fig. 7A. There was a direct, linear relationship between fluorescence intensity and E. coli O157 CFU in the samples (R 2 0.979, 0.957, and 0.974 for spring water, tap water, and groundwater, respectively), indicating that the IMB/IL fluorescence assay can be used quantitatively. In all cases, evaluation of 100-ml samples demonstrated that final concentrations less than 1 CFU/ml were readily detected by using the optimized assay format. By linear regression analysis, the limits of detection (defined as the mean for a blank or unspiked sample plus 3 SDs) were calculated to be 0.028, 0.23, and 0.055 CFU/ml for spring water, tap water, and groundwater, respectively. In agreement, spring water, tap water, and groundwater containing 0.27, 0.38, and 0.14 CFU of E. coli O157:H7/ml, respectively, were statistically significantly different from unspiked control samples (P 0.0001, 0.0038, and 0.0018, respectively). Of note, we consistently observed that groundwater samples yielded stronger signals than either chlorinated tap water or bottled spring water. This finding may be explained by the low viability of E. coli O157 strains in chlorinated water. At doses of 0.2 g of residual chlorine/ml, the microbes have been reported to die within 30 s (17). Likewise, ozonation is generally used to disinfect spring water during the bottling process and may also affect the viability of exogenously added E. coli O157:H7. We also used our optimized assay format to determine if known concentrations of E. coli O157:H7 could be readily detected in apple cider and apple juice. To acclimate the exogenously added E. coli O157:H7 to the lower-ph environment of the apple juice and cider, a stock solution of E. coli O157:H7 was prepared in pasteurized apple juice stored for several days. Next, fresh pasteurized apple juice and cider samples were spiked with various dilutions of the of E. coli O157:H7 stock solution (containing 15 to 150 CFU) and analyzed via the IMB/IL fluorescence assay. Of note, due to the presence of higher levels of particulate matter in apple cider, we found it necessary to replace the filtration step with a simple centrifugation step to concentrate the target organisms prior to analysis. Following centrifugation, the enrichment-capture media and immunomagnetic beads were directly added to the pellet and the rest of the assay was performed as described above. Figure 7B shows the mean ( SD) fluorescence signals for samples of apple juice and cider spiked with increasing amounts of E. coli O157:H7 and analyzed via the IMB/IL fluorescence assay. Excellent dose-response curves were obtained with both apple juice and cider (R 2 0.971 and 0.988, respectively), with limits of detection even lower than those observed for the water matrices. By linear regression analysis, the limit of detection was calculated at 0.031 CFU/ml for apple cider and is expected to be similar for apple juice; however, determination of a formal limit of detection for the later matrix was not possible due to the extremely low fluorescence signal in the unspiked sample. In agreement, apple juice and cider samples containing 0.02 and 0.04 CFU of E. coli O157:H7/ml were found to be statistically significantly different from unspiked controls (P 0.0002 and 0.0023, respectively). Assay validation and comparison against a standard microbiological method. A preliminary evaluation of the specificity of the assay was performed as a study in the blind by the Erie County Regional Public Health Laboratories (Buffalo, N.Y.) by determining the strength of the signal resulting from the evaluation of known concentrations of E. coli O157:H7, a laboratory strain (ATCC 25922) of E. coli, and B. subtilis in bottled spring water. As noted in Table 1, low concentrations of E. coli O157:H7 (i.e., 1 to 2 CFU/ml) yielded fluorescence signals 4- to 10-fold greater than that for the negative control (n 2). In contrast, 10-fold-higher concentrations of E. coli (laboratory strain) or B. subtilis (i.e., 10 to 20 CFU/ml) yielded fluorescence signals similar to that for the negative control. Defining a positive result as a sample with a fluorescence signal more than three times the background noise, staff at the Erie County Public Health Laboratories were able to identify E. coli O157: H7-containing samples with 100% accuracy. Unspiked samples

1862 DECORY ET AL. APPL. ENVIRON. MICROBIOL. TABLE 1. Specificity of the E. coli O157:H7 IMB/IL fluorescence assay a Sample (concn [CFU/ml]) Fluorescence signal (au) c Result b Unspiked 47.5 Negative Unspiked 46.4 Negative Unspiked 92.0 Negative B. subtilis (10 20) 91.8 Negative B. subtilis (10 20) 80.4 Negative E. coli laboratory strain (1 2) 30.1 Negative E. coli laboratory strain (5 10) 43.4 Negative E. coli laboratory strain (10 20) 59.5 Negative E. coli O157:H7 (1 2) 384.1 Positive E. coli O157:H7 (1 2) 160.0 Positive E. coli O157:H7 (5 10) 627.3 Positive E. coli O157:H7 (10 20) 438.6 Positive Negative control 37.1 Negative Positive control 542.2 Positive a Tests performed by the Erie County Regional Public Health Laboratories. b A positive result for the detection of E. coli O157:H7 was defined as a fluorescence signal greater than three times that for the negative control. c au, arbitrary units. and those samples spiked with the E. coli laboratory strain or B. subtilis all gave negative results. The performance of the IMB/IL fluorescence assay was also compared to that of a standard microbiological method used at the Erie County Regional Public Health Laboratories. Briefly, sets of spiked (0.15 CFU of E. coli O157H7:H7/ml) and unspiked groundwater samples were analyzed by each method. The IMB/IL fluorescence assay was performed as described before with the exception that the enrichment-capture time was increased to 5 h (to ensure the detection of the low concentration of E. coli O157:H7) and the incubation of the resulting immunomagnetic bead-e. coli O157:H7 complexes with anti-e. coli O157H7 immunoliposomes was decreased to 30 min. The standard microbiological method entailed filtration of spiked and unspiked samples with 0.45- m gridded membrane filters and application of the filters directly onto SMAC plates. Any sorbitol-nonfermenting colonies present the next day following overnight incubation were considered suspect because most O157 Shiga toxin-producing E. coli isolates do not ferment D-sorbitol overnight. To confirm the absence of any endogenous E. coli O157:H7 in the groundwater used in this experiment, a third set of spiked and unspiked samples was analyzed via the API 20E enteric identification system (6, 7, 29). To prepare samples for the API 20E system, samples were filtered and the resulting filters were incubated with anti-e. coli O157:H7 immunomagnetic beads in PBS-Tween buffer. After 2 h, any immunomagnetic bead-e. coli O157:H7 complexes were isolated by magnetic separation and applied to SMAC plates; the following day, any sorbitol-nonfermenting colonies were further tested with the API 20E enteric identification system. While the API 20E system does not identify E. coli O157:H7 specifically, a negative reading for E. coli was interpreted as indirect evidence for the absence of E. coli O157:H7. Figure 8 shows the mean ( SD) fluorescence signals for E. coli O157:H7 spiked and unspiked groundwater samples analyzed via the optimized IMB/IL fluorescence assay and the standard microbiological method. Only spiked samples yielded positive fluorescence signals in the IMB/IL assay. In fact, the FIG. 8. Mean ( SD) fluorescence signals for E. coli O157:H7- spiked groundwater and nonspiked samples at a public health laboratory using the IMB/IL fluorescence assay (A) and mean ( SD) CFU results by a laboratory using a standard microbiological method (B)., P 0.05. au, arbitrary units. IMB/IL fluorescence assay was able to accurately detect all 10 groundwater samples spiked with 0.15 CFU of E. coli O157: H7/ml with a mean fluorescence signal approximately 10-fold greater than that for unspiked controls (109 35.8 and 14.8 3.47 for spiked and unspiked samples, respectively; P 0.0079). This result for the fluorescence signal of the spiked samples is consistent with the dose-response curve for groundwater generated previously (Fig. 7). In contrast, when membrane-filtered samples were analyzed on SMAC plates via the standard microbiological method used in the public health laboratory, all samples regardless of whether or not they were spiked with E. coli O157:H7 yielded significant plate counts of suspect colonies. While it is possible that the groundwater used in these experiments contained endogenous E. coli O157:H7, results of the API 20E enteric identification system used in the analysis of the third set of spiked and unspiked samples confirmed that, in fact, there was no E. coli present, and therefore no E. coli O157:H7, in the groundwater used to prepare samples in this experiment. In agreement with results obtained with the IMB/IL fluorescence assay, the API 20E enteric identification system detected a mean of 6.20 2.28 E. coli CFU in E. coli O157:H7-spiked samples. DISCUSSION The objective of this study was to develop and optimize an assay for the detection of E. coli O157:H7 using immunomagnetic beads specific for E. coli O157:H7 in combination with

VOL. 71, 2005 E. COLI O157:H7 IMB/IL FLUORESCENCE ASSAY 1863 SRB-containing liposome nanovesicles coated with anti-e. coli O157:H7 antibodies, thereby applying the advantages of liposome nanovesicles to an immunoassay format. The final optimized protocol (Fig. 1) consisted of the filtration or centrifugation of 30- to 100-ml samples followed by incubation of the filter membranes or pellet with anti-e. coli O157:H7 immunomagnetic beads in a special growth medium for E. coli O157:H7 for 4 h. The resulting E. coli O157:H7-immunomagnetic bead complexes were then isolated by magnetic separation, washed, and incubated with SRBcontaining immunoliposomes specific for E. coli O157:H7; the final immunomagnetic bead-e. coli O157:H7-immunoliposome complexes were again isolated by magnetic separation, washed, and lysed to release SRB. The final IMB/IL assay took less than 8 h to complete and had a detection limit of less than 1 CFU of E. coli O157:H7 per ml in various aqueous matrices as well as apple juice and cider. The extreme sensitivity of the assay resulted from the encapsulation of high concentrations of fluorescent dye molecules within the anti-e. coli O157:H7 immunoliposomes. Thus, by immunomagnetically capturing the E. coli O157:H7 in an immunomagnetic bead-e. coli O157:H7-immunoliposome complex and subsequently lysing the immunoliposomes to release the fluorescent marker, it was possible to significantly enhance the detection signal while keeping the total assay time to a single work day. A number of other investigators have reported excellent limits of detection (5, 22, 23, 27, 28) similar to those obtained with the IMB/IL assay. By encapsulating large numbers of fluorescent dye molecules within the core of the immunoliposomes, however, we have been able to significantly reduce the enrichment times over those previously reported. In addition, the use of a simple and handheld fluorometer provides a means to obtain quantitative results in an inexpensive, easy-to-perform, and portable format well suited for field use. As expected, we observed a direct relationship between time of enrichment-incubation with immumomagnetic beads in the IMB/IL assay and limit of detection. For example, while the limit of detection of E. coli O157:H7 in groundwater was 0.055 CFU/ml following a 4 h enrichment-incubation with immunomagnetic beads in Medi-8 growth medium, the exponential growth curve fitted to the data predicted a limit of detection of about 0.014 CFU/ml following a 5-h enrichment-incubation, and, in fact, we were easily able to detect 0.15 CFU/ml by simply increasing the enrichment-incubation time by 1 h, while still completing the assay within 8 h (Fig. 8); the fluorescence signal of the E. coli O157:H7-containing sample was approximately 10 times that of the unspiked control. Thus, the IMB/IL assay described herein offers the added flexibility of increasing (or decreasing) the limit of detection by adjusting the time of enrichment-incubation to meet state and local public health laboratory requirements. Alternately, it is expected that one could improve the limit of detection by increasing the volume of sample matrix analyzed, although this was not formally tested. Following optimization of the IMB/IL fluorescence assay, the protocol was tested at a public health laboratory to demonstrate that the newly developed assay could be easily performed by laboratory personnel not familiar with the assay. In the hands of staff blind to sample content and with minimal formal training in the IMB/IL fluorescence assay procedure, the assay was found to be 100% accurate in distinguishing E. coli O157:H7-containing water samples from water samples spiked with B. subtilis, a laboratory strain of E. coli, or unspiked controls. Staff at the public health laboratory also compared the IMB/IL fluorescence assay against their standard microbiological method. Results obtained with the standard microbiological method were inconclusive, with a number of suspect colonies identified in both spiked and unspiked samples. While further workup of the suspect colonies may have allowed for the identification of E. coli O157:H7-containing samples, the public health laboratory did not analyze these samples further since, according to their standard procedure, groundwater samples containing 500 plate counts are not analyzed further. In contrast, the IMB/IL fluorescence assay was able to identify samples containing E. coli O157:H7 with 100% accuracy. Furthermore, and perhaps more important in the public health setting, identification of E. coli O157:H7-containing samples with the IMB/IL fluorescence assay was possible within a single 8-h work shift in contrast to the several days necessary to perform the conventional microbiological method and additional workup to elucidate the microbial status of any suspect colonies (had plate counts been greater than 500 CFU). In conclusion, the results presented herein demonstrate the feasibility of using immunomagnetic beads in combination with SRB-encapsulating immunoliposomes for the rapid detection of E. coli O157:H7 in aqueous samples. Due to the extreme sensitivity and specificity of immunoliposomes, only a few hours of enrichment were required to detect, quantitate, and identify E. coli O157:H7-containing samples, substantially less time than is required for previous immunoassays. Finally, use of this procedure is expected to increase the speed by which E. coli O157:H7 may be identified in aqueous specimens linked to epidemiological investigations and may thus increase the speed by which health care facilities and the population at large may be alerted of possible health threats. ACKNOWLEDGMENTS This work was supported by funding from the New York State Energy Research and Development Authority and Innovative Biotechnologies International, Inc., Grand Island, N.Y. REFERENCES 1. Avanti Polar Lipids, Inc. 2000. Products catalog, ed. VI. Avanti Polar Lipids, Inc., Alabaster, Ala. 2. Bartlett, G. R. 1959. Phosphorus assay in column chromatography. J. Biol. Chem. 234:466 468. 3. Bennett, A. R., S. MacPhee, and R. P. Betts. 1996. The isolation and detection of Escherichia coli O157 by use of immunomagnetic separation and immunoassay procedures. Lett. Appl. Microbiol. 22:237 243. 4. Besser, R. E., S. Lett, J. Weber, M. P. Doyle, T. Barrett, J. G. Wells, and P. Griffin. 1993. An outbreak of diarrhea and hemolytic uremic syndrome from Escherichia coli O157:H7 in fresh pressed apple cider. JAMA 269:2217 2220. 5. Bopp, D. J., B. D. Sauders, A. L. Waring, J. Ackelsberg, N. Dumas, W. Braun-Howland, D. Dziewulski, B. J. Wallace, M. Kelly, T. Halse, K. A. Musser, P. F. Smith, D. L. Morse, and R. J. Limberger. 2003. Detection, isolation, and molecular subtyping of Escherichia coli O157:H7 and Campylobacter jejuni associated with a large waterborne outbreak. J. Clin. Microbiol. 41:174 180. 6. Brooks, K. A., M. Jens, and T. M. Sodeman. 1974. A clinical evaluation of the API microtube system for identification of Enterobacteriaceae. Am. J. Med. Technol. 40:55 61. 7. Castillo, C. B., and D. A. Bruckner. 1984. Comparative evaluation of the Eiken and API 20E systems and conventional methods for identification of members of the family Enterobacteriaceae. J. Clin. Microbiol. 20:754 757. 8. Centers for Disease Control and Prevention. 1996. Outbreaks of Escherichia coli O157:H7 infections associated with drinking unpasteurized commercial apple juice British Columbia, California, Colorado, and Washington, October 1996. Morb. Mortal. Wkly. Rep. 45:975.