Rapid Separation and Concentration of Food-Borne Pathogens in Food Samples Prior to Quantification by Viable-Cell Counting and Real-Time PCR

Transcription

1 APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Jan. 2007, p Vol. 73, No /07/$ doi: /aem Copyright 2007, American Society for Microbiology. All Rights Reserved. Rapid Separation and Concentration of Food-Borne Pathogens in Food Samples Prior to Quantification by Viable-Cell Counting and Real-Time PCR Hiroshi Fukushima,* Kazunori Katsube, Yukiko Hata, Ryoko Kishi, and Satomi Fujiwara Shimane Prefectural Institute of Public Health and Environmental Science, 582 Nishihamasada, Matsue, Shimane , Japan Received 26 July 2006/Accepted 11 October 2006 Buoyant density gradient centrifugation has been used to separate bacteria from complex food matrices, as well as to remove compounds that inhibit rapid detection methods, such as PCR, and to prevent false-positive results due to DNA originating from dead cells. Applying a principle of buoyant density gradient centrifugation, we developed a method for rapid separation and concentration following filtration and low- and highspeed centrifugation, as well as flotation and sedimentation buoyant density centrifugation, for 12 food-borne pathogens (Salmonella enterica, Escherichia coli, Yersinia enterocolitica, Campylobacter jejuni, Vibrio cholerae O139, Vibrio parahaemolyticus O3K6, Vibrio vulnificus, Providencia alcalifaciens, Aeromonas hydrophila, Bacillus cereus, Staphylococcus aureus, and Clostridium perfringens) in 13 different food homogenates. This method can be used prior to real-time quantitative PCR (RTi-qPCR) and viable-cell counting. Using this combined method, the target organisms in the food samples theoretically could be concentrated 250-fold and detected at cell concentrations as low as 10 1 to 10 3 CFU/g using the RTi-qPCR assay, and amounts as small as 10 0 to 10 1 CFU/g could be isolated using plate counting. The combined separation and concentration methods and RTi-qPCR confirmed within 3 h the presence of 10 1 to 10 2 CFU/g of Salmonella and C. jejuni directly in naturally contaminated chicken and the presence of S. aureus directly in remaining food items in a poisoning outbreak. These results illustrated the feasibility of using these assays for rapid inspection of bacterial food contamination during a real-world outbreak. Rapid detection of pathogenic organisms that cause foodborne illness is needed to ensure food safety (1). Even with improved methods for detecting pathogens in foods and environmental samples, microbiologists often face a needle-in-ahaystack challenge (8). It is very difficult to detect small numbers of food-borne pathogens amid large numbers of harmless background microflora in a complex sample matrix. Traditional culture techniques for direct isolation and identification of food-borne pathogens in food samples in poisoning outbreaks are time-consuming and laborious; therefore, efforts have been made to reduce the time required to identify these pathogens. Over the years, plating methods have been replaced by more rapid methods, such as DNA hybridization, enzyme immunoassays, and real-time PCR (RTi-PCR) (8, 22). However, at best, these methods detect 10 3 to 10 4 CFU/g of target pathogens, meaning that culture enrichment steps are still necessary, as is confirmation of presumptively positive results (8). Methods for separating bacteria from a food matrix and then concentrating them depend on several chemical, physical, and biological principles. Filtration and centrifugation are physical methods that are commonly used to separate and concentrate microorganisms from a complex sample matrix. However, there is still a lack of suitable separation and concentration methods that allow rapid quantification of the nucleic acids and removal of PCR inhibitors. Buoyant density centrifugation * Corresponding author. Mailing address: Shimane Prefectural Institute of Public Health and Environmental Science, 582 Nishihamasada, Matsue, Shimane , Japan. Phone: Fax: fukushima-hiroshi@pref.shimane.lg.jp. Published ahead of print on 20 October (BDC) has been used for rapid detection of food pathogens, such as Shigella flexneri (13), Escherichia coli (12), and Yersinia enterocolitica (10, 11, 13), by a sedimentation method and for rapid detection of Y. enterocolitica (22) and Campylobacter jejuni (23) by the flotation method. The benefits of BDC as a sample pretreatment method are well established and include (i) the possibility of separating biological matrix particles and microorganisms with different buoyant densities (12); (ii) elimination of parts of the PCR-inhibiting food substances (10); (iii) prevention of false-positive results due to DNA originating from dead cells, which has limited the use of quantitative PCR (qpcr) (22); (iv) the possibility of direct quantification of target organisms in the presence of a large background flora (22); (v) maintenance of cell viability, which allows isolation and analysis of the microorganisms (18); and (vi) speed and easy handling. However, even PCR methods after BDC detect at best 10 3 to 10 4 CFU/g of target pathogens (11, 12, 22, 23). In this study we examined the use of a combination of several sample preparation methods, including filtration, lowspeed centrifugation, high-speed centrifugation, and finally two BDC methods, called (i) flotation, in which the top layer consists of a low-density solution and the bottom layer consists of a high-density solution of density gradient medium mixed with the sample, and (ii) sedimentation, in which density gradient solutions are rapidly and easily prepared without contamination from other food matrix in tubes. The aim of this study was to develop a rapid separation and concentration method that works within 3 h for 12 food-borne pathogens in food samples prior to quantification by viable-cell counting and RTi-PCR. Finally, this method was tested with naturally con- 92

2 VOL. 73, 2007 RAPID SEPARATION AND CONCENTRATION OF PATHOGENS 93 TABLE 1. Buoyant densities of Enterobacteriaceae strains, C. jejuni strains, Vibrionaceae strains, S. aureus strains, B. cereus strains, C. perfringens strains, and 13 food homogenates Bacterium (no. of strains) or food homogenate(s) Strains Buoyant density (g/ml) Bacteria Escherichia coli (8) EC-2736, EC-2649, EC-3515, EC-4725, EC-4131, SE-02005, SE-02025, SE Salmonella spp. (3) Sal-2339, Sal-2340, Sal Yersinia enterocolitica (5) Pa177, Pa241, Pa2718, Pa9346, Pa Providencia alcalifaciens (3) ATCC 9886, 112, Campylobacter jejuni (4) SC009, SC010, SC011, SC Vibrio cholerae (4) ATCC 14035, NIID63-93, NIID169-68, SVP TDH-positive V. parahaemolyticus (4) SVP02, SVP03, SVP04, NIIDK Vibrio vulnificus (3) SVV1526, SVV04001, SVV Aeromonas hydrophila (2) ATCC 7966, M Staphylococcus aureus (3) SS 05, FB0501, FB Bacillus cereus (10) 127, 128, 129, 130, 131, 132, 133, 135, 136, Clostridium perfringens (2) CW2, H Food homogenates Minced beef, bovine liver, minced chicken, processed cheese, scrambled egg, tofu, Chinese noodle, bread, raw chopped jack horse mackerel, short-neck clam Minced pork, ready-to-eat hamburger steak Milk taminated chicken samples, as well as with food samples remaining from a poisoning outbreak. MATERIALS AND METHODS Bacterial strains. The 12 food-borne pathogens used in this study are shown in Table 1. Bacterial cultures and viable-cell counting were described in a previous report (6). Determination of buoyant densities of bacteria. The buoyant densities (BD) of the different bacterial strains in the stationary phase and in food were determined by centrifugation using a step density gradient of Percoll (Pharmacia Biotech, Sweden) as described by Pertoft (17). The following nine concentrations of Percoll were prepared by diluting stock isotonic Percoll (SIP) (Percoll 1.5 M NaCl, 9:1) with 0.15 M NaCl: g/ml (SIP 0.15 M NaCl, 1,000:0), 1.10 g/ml (SIP 0.15 M NaCl, 811:189), 1.09 g/ml (SIP 0.15 M NaCl, 721:279), g/ml (SIP 0.15 M NaCl, 595:405), g/ml (SIP 0.15 M NaCl, 552:448), g/ml (SIP 0.15 M NaCl, 468:532), g/ml (SIP 0.15 M NaCl, 384:616), g/ml (SIP 0.15 M NaCl, 215:785), and g/ml (SIP 0.15 M NaCl, 88:912). For Vibrio parahaemolyticus and Vibrio vulnificus in this study, a 3% NaCl solution was used for dilution. The step density gradient was prepared by carefully layering 1-ml portions of nine Percoll concentrations (from high concentrations to low concentrations) in a plastic 15-ml conical tube. Then 0.2 ml of 0.15 M NaCl plus 5 l of each density marker bead (DMB) (Pharmacia Biotech, Sweden) preparation were placed on the gradient medium. After this, 1 ml of the bacterial strain was carefully layered on top. Centrifugation was performed with a Hitachi himac CR22E (Hitachi) at 16,000 g for 30 min at room temperature. A plot describing the densities of different locations in the step density gradient was prepared by measuring the distances from the bottom of the tube to the different layers of DMB and plotting the known densities of the different DMB (in grams per milliliter) versus distance (in centimeters). Food samples. Thirteen food samples were used for examination of the recovery of 12 food-borne pathogens (Table 1). To confirm the absence of natural contamination from small numbers of food-borne pathogens in the 13 samples, each 10 g was analyzed using enrichment cultures. Members of the Enterobacteriaceae, Staphylococcus aureus, and Bacillus cereus were cultured overnight at 37 C (or 30 C for Yersinia) in buffered peptone water (BPW) (BBL). Members of the Vibrionaceae were cultured overnight at 37 C in alkali peptone water (Nissui). C. jejuni was cultured overnight at 42 C using microaerobic incubation in Preston medium (Oxoid). Clostridium perfringens was cultured overnight at 37 C using anaerobic incubation in thioglycolate medium. Each 1 ml of an enrichment culture was centrifuged at 1,880 g for 5 min at room temperature. The upper portion was layered on top of 0.7 ml of 50% (vol/vol) SIP in a 1.5-ml microtube. The contents were centrifuged at 14,500 g for 5 min. The supernatant was carefully removed and discarded, leaving 0.1 ml, which was vortexed with 1.4 ml of 0.15 M NaCl. After centrifugation at 14,500 g for 5 min, the pellet was treated with 50 l of Instagene matrix (Bio-Rad) to obtain a DNA sample (according to the manufacturer s instructions). The DNA samples were analyzed by RTi-PCR. Protocol for rapid separation and concentration of food-borne pathogens in food samples. The optimal protocol is shown in Fig. 1. The first step was filtration, followed by low- and high-speed centrifugation. Each 25-g food sample and 1 ml of a bacterial solution were placed in a small plastic bag (Stomafilter P type; Gunze, Tokyo, Japan), which was sealed with Teflon cloth (40 mesh) on the inside 3 cm from the end by heat treatment, and they were mixed with 225 ml of 0.02% Tween 20 (Wako)-buffered peptone water (Tween20-BPW) (BBL) and homogenized in a stomacher for 1 min. The Tween 20 was added to BPW to emulsify the fat in the samples. Tween20-BPW containing 2% NaCl was used for V. parahaemolyticus and V. vulnificus. Approximately 220-ml portions of filtered solutions of the homogenates were removed and placed into sterilized 350-ml glass tubes and then were centrifuged at 1,880 g for 5 min at room temperature, using a swing router. The upper portion was removed and placed in a sterilized 500-ml plastic tube and then centrifuged at 16,000 g with a Hitachi himac CR22E for 5 min at room temperature. The upper portion was then removed, the pellet was suspended in 1.5 ml of 0.15 M NaCl, and a viable-cell count was obtained. The contents were centrifuged at 14,500 g with a benchtop centrifuge for 5 min at room temperature. The upper portion (1 ml) was removed, and the pellet (0.5 ml) was used for sampling. The second step was flotation and sedimentation buoyant density centrifugation for purification of food-borne pathogens. In the flotation assay, 0.5-ml portions of sample suspensions were mixed with 1 ml of a g/ml Percoll solution and centrifuged at 4,500 g for 15 min at room temperature. The upper portion, including the food matrix, was carefully removed. For the sedimentation assay, the bottom portion (about 0.5 ml), including organisms, food particles, and the mass with the highest buoyant density, was homogenized and then placed on top of two layers (0.6 ml of a g/ml Percoll solution and 0.6 ml of a g/ml Percoll solution) in a 1.5-ml microtube to which two density markers (orange for g/ml and green for g/ml or red for g/ml for S. aureus and B. cereus) were added. The preparations were centrifuged at 14,500 g for 5 min at room temperature, and then using sterile 1-ml pipettes, about 1 ml was taken from the interface between the two density makers and divided into two samples. The samples were added to 1 ml of 0.15 M NaCl in a 1.5-ml microtube. The preparations were then centrifuged at 14,500 g for 5 min. The bottom portions (0.5 ml) were resuspended with 1 ml of 0.15 M NaCl and then centrifuged at 14,500 g for 5 min. Each pellet was used for viable-cell counting and DNA extraction with Instagene matrix. One portion of the sample was resolved with 50 l of 0.15 M NaCl, and then viable-cell counts (CFU/g), which were obtained by culturing each dilution (10 l) (for each suitable culture) using

3 94 FUKUSHIMA ET AL. APPL. ENVIRON. MICROBIOL. FIG. 1. Optimal protocol for rapid separation and concentration of food-borne pathogens in food samples using filtration, centrifugation, and BDC prior to quantification by viable-cell counting and real-time PCR. suitable selective agar plates, were determined for these BDC-lysate pellets (50 l). The other portion was treated with 50 l of Instagene matrix for DNA extraction prior to RTi-qPCR by using specific primers for each food-borne pathogen. The total volume of each 25-g sample could be reduced to 0.1 ml, and the target organisms in the sample were theoretically concentrated 250-fold. RTi-PCR conditions. For real-time PCR, we used glass capillary tubes, SYBR Premix EX Taq (Takara, Japan), and a LightCycler instrument (Roche Diagnostics, Mannheim, Germany) as described by the manufacturer. Each reaction tube contained 10 l of SYBR Premix EX Taq, 7.2 l of PCR-grade H 2 O, 0.4 l ofa 10 M primer set, and 2 l of template DNA in a 20- l PCR mixture. The assay cycling profile was 95 C for 10 min, followed by 28 to 32 cycles of denaturation at 95 C for 5 s and then annealing at 60 C for 20 s. Fluorescence signals were measured once in each cycle at the end of the extension step. After PCR amplification, a melting temperature curve analysis was done. The LightCycler PCR products were cooled to 65 C and then heated to 95 C at a rate of 0.1 C per s. The fluorescence signals obtained were continuously monitored to confirm amplification specificity during 1 h of analysis. The melting temperature peaks for the products were calculated by performing 10 or more assays for each sample and were based on the initial fluorescence curve by plotting the negative derivative of fluorescence over temperature versus temperature. The two primer sets used for detection of Salmonella enterica and enterohemorrhagic E. coli were primers Salm16S-F and R1 (4) and primers eae-1 and 2 (7), respectively. Primers CPE554-F2 (5 -CTGCAGATAGCTTAGGAAATATTGATCA-3 ) and CPE665-R2 (5 -GCAGCTAAATCAAGGATTTCTTTTTCT-3 ) amplifying a 112-bp fragment for detection of C. perfringens were constructed using the CPE gene (GenBank accession no. AF81849). Nine other LightCycler PCR primer sets used in this study have been described previously (5, 6). Viable-cell counting and qpcr of spiked food-borne pathogens in 13 food samples. In experiment 1 (see Table 3), 10 5 to 10 6 CFU of 12 food-borne pathogens was added to 12 different food samples (25 g each) mixed with 225 ml of Tween20-BPW and 25 ml of whole milk. Viable-cell counts in 1.5-ml concentrated samples were determined after BDC using the protocol shown in Fig. 1. Then RTi-qPCR was performed with DNA extracts from the BDC-lysate pellets along with 50 l of Instagene matrix. In experiment 2 (see Table 4), 12 foodborne pathogens were serially diluted 10-fold to obtain between 10 0 and 10 5 CFU/ml, and 1-ml portions of the dilutions were added to 25 g of ready-to-eat hamburger steak mixed with 225 ml of Tween20-BPW. These samples were treated according to the protocol described above. Use of rapid separation and concentration for chicken samples for quantification by viable-cell counting and RTi-qPCR. Thirty-two chicken samples (16 chopped meat samples and 16 flat wing samples) were bought at eight local stores in the eastern part of Shimane Prefecture on 15 to 29 May Chicken rinse samples were treated using the protocol described above. Viable-cell counts were obtained by culturing samples on Skirrow agar for C. jejuni and on MLCB

4 VOL. 73, 2007 RAPID SEPARATION AND CONCENTRATION OF PATHOGENS 95 agar for S. enterica. RTi-qPCR was performed by using specific primers for C. jejuni and S. enterica. Other 25-g samples were diluted 10-fold with Preston medium and Tween20-BPW, and then the numbers of CFU and the most probable numbers (MPN) of C. jejuni and S. enterica were calculated for all samples. Use of rapid separation and concentration for instances of poisoning. The method was experimentally tested for a poisoning outbreak involving 22 (40.7%) of 90 individuals who ate vegetable dishes at meals after a funeral in Matsue City in Shimane Prefecture on 16 May In 22 patients, poisoning symptoms, such as acute vomiting (13 patients [59.1%]), diarrhea (18 patients [81.8%]), fever (1 patient [4.5%]), and headache (2 patients [9.1%]), occurred 2 to 18 h (average, 5.16 h) after cooked vegetable dishes were eaten. Stool samples were collected from nine patients with serious cases at Matsue Red Cross Hospital on 16 May, and then pathogenic organisms were cultivated in the hospital laboratory. To investigate the possibility of infection by toxic organisms such as S. aureus or B. cereus, seven remnants of cooked food, including two samples of fried bean curd, one sample of shiitake mushrooms, one sample of konjac, one sample of butterbur, one sample of Japanese freeze-dried tofu, and one sample of chikuwa, were collected from the kitchen of the meeting place along with one sample of shiitake mushrooms from a cookshop and brought to our laboratory at 10:30 a.m. on 17 May. Within 2 h, 250-fold concentrates of these eight samples were obtained. One half of each sample was treated with Instagene matrix and then tested by RTi-qPCR in order to screen for S. aureus and B. cereus. The virulence or specific genes of other food-borne pathogens were analyzed by using duplex RTi-PCR as described by Fukushima et al. (6). The other half of each sample was used for viable-cell counting by culture on mannitol salt agar for S. aureus and on NGKG agar for B. cereus. RESULTS BD of the bacterial strains and food homogenates. The BD of the food-borne pathogens and food homogenates are shown in Table 1. The BD of the pathogenic species of Enterobacteriaceae and C. perfringens were similar and ranged from to g/ml. The BD of members of the Vibrionaceae were similar and ranged from to g/ml. The BD of C. jejuni ranged from to g/ml, the BD of B. cereus ranged from to g/ml, and the BD of S. aureus was to g/ml. The BD of almost all food homogenates were g/ml; the exceptions were g/ml for pork and hamburger steak and g/ml for whole milk. Filtration and low- and high-speed centrifugation. The cell concentrations and rates of recovery for 12 food-borne pathogens added to 13 food samples at a level of 10 5 CFU/25 g are shown in Tables 2 and 3. Filtration through Stomafilter mesh could remove solid food particles from 10-fold-diluted food samples. The average recovery rate was 106.6%, and the values ranged from 100.3% for B. cereus to 140.3% for S. aureus, except for the 48.5% for C. perfringens. Almost all the inoculated bacteria belonging to the Enterobacteriaceae, Vibrionaceae, and C. jejuni were in the supernatant fluid after low-speed centrifugation, and almost all food components were deposited on the bottom. The average rate of recovery of organisms from food samples was 67.4%, and the values ranged from 49.2% for S. enterica to 98.2% for Providencia alcalifaciens, except for 5.9% for C. perfringens and 41.4% for B. cereus, which were significantly different from the values for the other pathogens [t 3.78, t 0 t (0.01)]. High-speed centrifugation resulted in deposition of approximately one-quarter of the food-borne pathogens, undefined background flora, and some food components on the bottom. The average rate of recovery of organisms from food samples was 24.4%, and the values ranged from 19.1% for V. parahaemolyticus to 56.8% for E. coli O157, except for 6.4% for B. cereus and 3.2% for C. perfringens, which d Average standard deviation. a All values are averages for at least duplicate analyses through filtration to qpcr after BDC. b Cell concentrations and recovery rates were calculated using the positive samples. c 1, minced beef; 2, bovine liver; 3, minced chicken; 4, minced pork; 5, processed cheese; 6, scrambled egg; 7, tofu; 9, bread; 10, raw chopped jack horse mackerel; 11, short-neck clam; 13, whole milk. Avg S. enterica d E. coli O Y. enterocolitica P. alcalifaciens C. jejuni C. perfringens V. cholerae O V. vulnificus V. parahaemolyticus ,11 A. hydrophila ,3,5,9,10, 11,13 B. cereus ,2,3,4,6,7 S. aureus Organism Inoculum (log 10 CFU/ 25 g) Selective plate count % Recovery Selective plate count % Recovery Selective plate count % Recovery Selective plate count % Recovery Real-time quantitative PCR count % Negative Recovery sample(s) c Filtration Low-speed centrifugation High-speed centrifugation Buoyant density centrifugation Buoyant density centrifugation b Concn of cells and recovery rates after the following treatments a : TABLE 2. Quantification and rates of recovery of inoculated organisms from 13 different food samples after filtration, low- and high-speed centrifugation, and buoyant density centrifugation

5 96 FUKUSHIMA ET AL. APPL. ENVIRON. MICROBIOL. TABLE 3. Direct recovery of 12 food-borne pathogens inoculated into 13 food samples Inoculum and concn of cells as determined by viable-cell counting and RTi-qPCR a Food sample S. enterica E. coli O157:H7 Y. enterocolitica P. alcalifaciens C. jejuni C. perfringens Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Beef Bovine liver Pork Chicken Processed cheese Scrambled egg Tofu Chinese noodle Bread Jack horse mackerel Short-neck clam Hamburger steak Whole milk a The values are averages for at least duplicate analyses. RTi-qPCR data were obtained from a standard curve for each food-borne pathogen. VC, viable-cell count; qpcr, RTi-qPCR count. b 0, not detected. were significantly different from the values for other pathogens [t 2.96, t (0.01) t 0 t (0.05)]. BDC. The average rates of recovery of organisms from 13 different food samples by viable-cell counting and RTi-qPCR after BDC were 11.0% and 16.9%, respectively (Table 2). The rates of recovery by viable-cell counting ranged from 0.8% for C. perfringens to 22.7% for E. coli O157, and the rates were significantly different for C. perfringens, B. cereus, and S. aureus and members of other families [t 3.41, t 0 t (0.01)]. The rates of recovery based on RTi-qPCR ranged from 0.4% for Aeromonas hydrophila to 91.2% for E. coli O157:H7, although the target organisms were not detected by RTi-qPCR in the 16 cases shown in Tables 2 and 3. The rates of recovery by RTiqPCR were significantly different for members of the Enterobacteriaceae and members of other families [t 4.75, t 0 t (0.01)]. In experiment 2 (Table 4), all 12 food-borne pathogens added to 25 g of hamburger steak were counted by RTi-qPCR using samples spiked with 10 2 CFU of E. coli O157:H7 and V. parahaemolyticus, 10 3 CFU of S. enterica, C. jejuni, V. cholerae O139, and V. vulnificus, 10 4 CFU of P. alcalifaciens, C. perfringens, and S. aureus, 10 5 CFU of A. hydrophila, and 10 6 CFU of Y. enterocolitica and B. cereus. These pathogens were recovered from all the samples spiked with 10 4 CFU, and seven pathogens (S. enterica, E. coli, C. jejuni, C. perfringens, V. parahaemolyticus, V. cholerae, and A. hydrophila) were recovered from the samples spiked with 10 3 CFU. V. parahaemolyticus was recovered even from the samples spiked with 10 2 CFU. Use of rapid separation and concentration methods for chicken samples. Table 5 shows the results of quantification of S. enterica and C. jejuni in 32 retail chicken samples (16 chopped chicken samples and 16 flat wing samples) using conventional culture methods, as well as the rapid separation and concentration method. Within 3 h after treatment of chicken samples, RTi-qPCR could detect S. enterica serotype Infantis at levels of and CFU/25 g (56 CFU/g and CFU/g) in two chopped chicken samples. On the next day, S. enterica serotype Infantis (15 strains) and one untypeable strain were isolated from 16 (50%) chicken samples (9 chopped chicken samples and 7 flat wing samples) by conven- TABLE 4. Direct recovery of 12 food-borne pathogens inoculated into ready-to-eat hamburger steak as determined by viable-cell counting after BDC and by RTi-qPCR after BDC Test no. Inoculum and concn of cells as determined by viable-cell counting and RTi-qPCR a S. enterica E. coli O157:H7 Y. enterocolitica P. alcalifaciens C. jejuni C. perfringens Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr b a The values are averages for at least duplicate analyses. RTi-qPCR data were obtained from a standard curve for each food-borne pathogen. VC, viable-cell count; qpcr, RTi-qPCR count. b 0, not detected.

6 VOL. 73, 2007 RAPID SEPARATION AND CONCENTRATION OF PATHOGENS 97 TABLE 3 Continued V. parahaemolyticus O3K6 V. cholerae O139 V. vulnificus A. hydrophila S. aureus B. cereus Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr b tional culture methods. Two RTi-qPCR-positive samples were contaminated with and CFU/25 g (48 and 112 CFU/g) as determined by conventional culture methods, but the organisms could not be isolated from the BDC samples because other members of the Enterobacteriaceae were overgrown on MLCB agar. Another 14 chicken samples contaminated with S. enterica at levels of 7.5 to MPN/25 g ( 0.3 to 76 MPN/g) showed negative reactions in the RTiqPCR assay. In the survey for C. jejuni, RTi-qPCR could detect C. jejuni at levels of 45 to 273 CFU/25 g (1.8 to 11 CFU/g) in four samples (one chopped chicken sample and three flat wing samples). After 2 days, C. jejuni was isolated at levels of 40 and 90 MPN/25 g from chopped chicken samples and at a level of 100 CFU/25 g from flat wing samples by plate counting after BDC, but this organism could not be enumerated due to swarming on agar plates prepared from two other PCR-positive flat wing samples. C. jejuni was isolated from 12 (37.5%) chicken samples (5 chopped chicken samples and 7 flat wing samples) by conventional culture methods. RTi-qPCR-positive samples were contaminated with to MPN/25 g (112 to 400 MPN/g) as determined by conventional culture methods, but eight chicken samples (four chopped chicken samples and four flat wing samples) contaminated with C. jejuni at levels of 7.5 to MPN/25 g ( 0.3 to 240 MPN/g) showed negative reactions in the RTi-qPCR assay. Use of rapid separation and concentration in a food-borne poisoning outbreak. In the case of a poisoning outbreak, S. aureus was found at levels of to CFU/25 g in seven food samples by the RTi-qPCR assay after rapid separation and concentration within 3 h (Table 6), but B. cereus and other food-borne pathogens were not detected. As determined by conventional culture methods, seven RTiqPCR-positive samples were contaminated with to CFU/25 g before treatment and with to CFU/25 g after BDC, but a RTi-qPCR-negative sample was contaminated with CFU/25 g before treatment and with CFU/25 g after BDC. S. aureus strains isolated from nine patient and eight food samples were classified into coagulase type I and enterotoxin types B and D. DISCUSSION The separation and concentration of food-borne pathogens directly from food samples without culture enrichment are two TABLE 4 Continued V. parahaemolyticus O3K6 V. cholerae O139 V. vulnificus A. hydrophila S. aureus B. cereus Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr Inoculum VC qpcr

7 98 FUKUSHIMA ET AL. APPL. ENVIRON. MICROBIOL. TABLE 5. Quantification of C. jejuni and S. enterica serotype Infantis in retail chicken samples Cell concn Sample type Sample no. MPN/25 g before treatment C. jejuni S. enterica serotype Infantis CFU/25 g after BDC qpcr count (CFU/25 g) after BDC a MPN/25 g before treatment CFU/25 g after BDC qpcr count (CFU/25 g) after BDC Chopped chicken OG b OG Flat wing c a Real-time PCR values were calculated from a standard curve prepared for cells instead of DNA. b OG, Salmonella strains could not be isolated from the BDC samples because other Enterobacteriaceae were overgrown on MLCB agar. c The S. enterica serotype could not be determined. of the most important aspects of sample preparation. In previous studies, BDC was used for separation of not only live bacteria but also PCR inhibitors from food components (10, 11, 12, 13, 22, 23). Despite only partial recovery of bacteria TABLE 6. Quantification of S. aureus in eight remnants of cooked food samples from a poisoning outbreak Food sample Cell concn (CFU/25 g), as determined by: Selective plate countings a qpcr after Before treatment After BDC BDC b Shiitake Shiitake Negative Fried bean curd Fried bean curd Freeze-dried tofu Chikuwa Butterbur Konjac a Samples were plated on mannitol salt agar and examined to determine the initial cell count. b Real-time PCR values were calculated from a standard curve prepared for cells instead of DNA. (25%) (13), the PCR detection limits for S. flexneri (13), E. coli O157 (12), and Y. enterocolitica (10, 11) with sedimentation BDC were 10 2 to 10 3 CFU/ml in 0.1 to 0.83 ml of cheese, milk, and meat homogenates. Recently, flotation BDC combined with RTi-qPCR showed that cell concentrations of at least Y. enterocolitica CFU/ml of pork juice (22) and as low as C. jejuni CFU/ml of chicken rinse (23) could be positively identified without culture enrichment. However, for future applications in which lower cell concentrations are expected, Wolffs et al. (23) pointed out the possibility that large volumes can be used for flotation and for determining target concentrations, because flotation BDC is performed without concentration; i.e., 1 ml of sample is subjected to flotation, and 1 ml is recovered afterward. In this study, a miniaturized sample preparation method combined with filtration, low- and high-speed centrifugation, and BDC (flotation [22, 23] and sedimentation [10, 11, 12, 13] methods) was developed prior to quantification by viable-cell counting and RTi-qPCR. This method could reduce 250 ml of a 10-fold-diluted 25-g food sample to a 0.1-ml bacterial solution within 2 h, and the target organisms were theoretically concentrated 250-fold in the food samples. Finally, RTi-qPCR

8 VOL. 73, 2007 RAPID SEPARATION AND CONCENTRATION OF PATHOGENS 99 of the DNA samples extracted with 50 l of Instagene matrix from one half of the concentrate was performed within 1 h, and viable-cell counting was done using the other half of the concentrate. In the first step of this method, 10-fold-diluted food homogenates suspended with Tween20-BPW were filtered through Teflon cloth (40 mesh) sealed inside a small plastic bag (9). Then after low-speed centrifugation, more than threefourths of the target bacteria (except C. perfringens and B. cereus) remained in the supernatant fluids of samples. During high-speed centrifugation, about 220 ml of the supernatant fluid containing the target organisms, background flora, and small food particles containing PCR inhibitors was concentrated to 1.5 ml of a bacterial solution which contained onequarter of the initial inoculated levels (except for C. perfringens and B. cereus). These levels are nearly the same as the previously described recovered levels of Listeria monocytogenes, which were 0.1 to 1 log below the initial inoculum levels (21). The concentrates obtained from 25-g food samples by lowand high-speed centrifugation were composed of large volumes of small food particles containing PCR inhibitors which were greater than the volumes in 0.1 to 1 ml of 10-fold-diluted food samples previously used in flotation (22, 23) and sedimentation (10, 11, 12) BDC. The buoyant densities of 12 foodborne pathogens were different for the species, and all buoyant densities were between and g/ml (Table 1). The buoyant densities of almost all food matrices were less than g/ml; the exceptions were the buoyant densities of pork and hamburger steak (1.033 g/ml) and of whole milk (1.049 g/ml). Therefore, 1 ml of g/ml Percoll was used for the low-speed flotation BDC after mixing with 0.5-ml samples. The fractions of small food particles with the lowest buoyant densities were found on top of the density gradient in flotation BDC (22) and were removed after the first centrifugation. The viable cells and food particles with the highest buoyant densities were deposited in the bottom of the tube during flotation BDC. After high-speed sedimentation BDC of this sediment using two layers (densities, and g/ml), the target organisms were separated from the food matrix containing PCR inhibitors by recovering the fractions between and g/ml (1.121 g/ml for S. aureus). It was reported previously that the reduction in PCR inhibition in flotation and sedimentation BDC of small samples is comparable to that of DNA purification with a commercial kit (22). However, DNA purification from large samples in this study was performed carefully using Instagene matrix, because some food particles remained in the fractions that had nearly the same density as the target organisms. Although there was a risk of false-positive PCR results caused by the detection of DNA from dead cells (19), no signals were recovered from the Y. enterocolitica and Campylobacter recovery locations containing dead cells because the dead cells did not reach the target recovery location due to the low flotation speed (22, 23) used in this study. Using RTi-qPCR in this study, 12 food-borne pathogens were detected in 13 different food samples spiked with 10 5 CFU/25 g ( CFU/g) (except for 16 cases of four pathogens shown in Table 2), despite only partial recovery (16.9%) of bacteria, and they were counted the first time in samples spiked with 10 3 to 10 5 CFU/25 g ( to CFU/g) of hamburger steak. Although reports of isolation limits for spiked organisms in food samples with BDC are rare, successful isolation of Y. enterocolitica CFU/ml from meat juice of minced pork spiked with CFU/ml by flotation BDC was reported by Wolffs et al. (22). In this study, 12 food-borne pathogens were isolated from 13 different food samples spiked with 10 5 CFU/25 g ( CFU/g), despite only partial recovery (11.0%) of bacteria, and they were also isolated from the samples of hamburger steak spiked with to CFU/25 g (0.8 to 72 CFU/g). Thus, we demonstrated that BDC combined with flotation and sedimentation is a useful tool for separation of viable cells from a large amount of food matrix. The isolation limits of the pathogens were 100 times lower than the RTi-qPCR detection limits. These results showed that our concentration system was very useful not only for RTi-qPCR detection but also for viable-cell counting of food-borne pathogens in foodstuffs. In recent studies workers detected very low levels (1 to 10 CFU/g) of food-borne pathogens in food samples after enrichment (3, 10, 11, 14, 15, 16, 18, 20). However, the previously reported detection limits of RTi-PCR for L. monocytogenes in meat products, which were analyzed after filtration and Chelex-100-based DNA purification (18), and for C. jejuni in a chicken rinse sample, which was analyzed after flotation BDC (23), were as low as 10 2 and 10 3 CFU/g and 10 3 and 10 4 CFU/g, respectively, because the sample volumes were very small (0.5 to 2 ml of 10-fold-diluted food samples). In the present study, RTi-qPCR analysis of 50- l concentrates from 25-g naturally contaminated chicken samples detected concentrations as low as 10 to 10 2 MPN/g of C. jejuni and S. enterica within 3 h. Although confirmation of the presence of Salmonella directly in chicken in remaining food items using the RTi-PCR assay for an outbreak of acute gastroenteritis among attendees of a church picnic illustrated the feasibility of using RTi-PCR assays during real-world outbreaks, the levels of contaminating organisms in the food samples remained unclear (2). In this study, the RTi-qPCR assay and viable-cell counting after rapid separation and concentration were used for the first time for rapid quantification (within 3 h) of S. aureus in food items suspected to be the source of poisoning. Therefore, our study illustrated the feasibility of using these assays for rapid determination of bacterial food contamination during a real-world outbreak. However, the values obtained by viable-cell counting and the values obtained by RTi-qPCR counting after BDC were sometimes different orders of magnitude. It is possible that the values were affected by factors such as manual handling for recovery during BDC, the characteristics and large amounts of food constituents, the sensitivity of the primers used for RTiqPCR, and the selectivity of the plate agar used for viable-cell counting. In particular, a negative RTi-qPCR may have occurred as a result of the use of concentrates, including large amounts of food constituents, such as cheese (Table 3), which are difficult to eliminate when this protocol is used. The technical pitfalls that may occur as a result of using this method in a routine set-up should be kept in mind. In future studies workers should improve automation of sample recovery to standardize the process and select more suitable primers for food-borne pathogens.

9 100 FUKUSHIMA ET AL. APPL. ENVIRON. MICROBIOL. REFERENCES 1. Benoit, P. W., and D. W. Donahue Methods for rapid separation and concentration of bacteria in food that bypass time-consuming cultural enrichment. J. Food Prot. 66: Daum, L. T., W. J. Barnes, J. C. McAvin, M. S. Neidert, L. A. Cooper, W. B. Huff, L. Gaul, W. S. Riggins, S. Morris, A. Salmen, and K. Lohman Real-time PCR detection of Salmonella in suspect foods from gastroenteritis outbreak in Kerr country, Texas. J. Clin. Microbiol. 40: Ellingson, J. L., J. J. Koziczkowski, J. L. Anderson, S. A. Carlson, and V. K. Sharma Rapid PCR detection of enterohemorrhagic Escherichia coli (EHEC) in bovine food products and feces. Mol. Cell. Probes 19: Fey, A., S. Eichler, S. Flavier, R. Christen, M. G. Hofle, and C. A. Guzman. Establishment of a real-time PCR-based approach for accurate quantification of bacterial RNA targets in water, using Salmonella as a model organism. Appl. Environ. Microbiol. 70: Fukushima, H., and Y. Tsunomori Study of real-time PCR assays for rapid detection of food-borne pathogens. J. Jpn. Assoc. Infect. Dis. 79: Fukushima, H., Y. Tsunomori, and R. Seki Duplex real-time SYBR green PCR assays for detection of 17 species of food- or waterborne pathogens in stools. J. Clin. Microbiol. 41: Ibekwe, A. M., P. M. Watt, C. M. Grieve, V. K. Sharma, and S. R. Lyons Multiplex fluorogenic real-time PCR for detection and quantification of Escherichia coli O157:H7 in dairy wastewater wetlands. Appl. Environ. Microbiol. 68: Jaykus, L.-A Challenges to developing real-time methods to detect pathogens in foods. ASM News 69: Konuma, H., A. Suzuki, and H. Kurata Improved stomacher 400 bag applicable to the spiral plate system for counting bacteria. Appl. Environ. Microbiol. 44: Lambertz, S. T., R. Lindqvist, A. Ballagi-Pordany, and M.-L. Danielsson- Tham A combined culture and PCR method for detection of pathogenic Yersinia enterocolitica in food. Int. J. Food Microbiol. 57: Lantz, P.-G., R. Knutsson, Y. Blixt, W. A. Al-Soud, E. Borch, and P. Radstrom Detection of pathogenic Yersinia enterocolitica in enrichment media and pork by a multiplex PCR: a study of sample preparation and PCR-inhibitory components. Int. J. Food Microbiol. 45: Lindqvist, R Preparation of PCR samples from food by a rapid and simple centrifugation technique evaluated by detection of Escherichia coli O157:H7. Int. J. Food Microbiol. 37: Lindqvist, R., B. Norling, and S. T. Lambertz A rapid sample preparation method for PCR detection of food pathogens based on buoyant density centrifugation. Lett. Appl. Microbiol. 24: Malorny, B., E. Paccassoni, P. Fach, C. Bunge, A. Martin, and R. Helmuth Diagnostic real-time PCR for detection of Salmonella in food. Appl. Environ. Microbiol. 70: Nguyen, L. T., B. E. Gillespie, H. M. Nam, S. E. Murinda, and S. P. Oliver Detection of Escherichia coli O157:H7 and Listeria monocytogenes in beef products by real-time polymerase chain reaction. Foodborne Pathog. Dis. 1: Panicker, G., M. L. Myers, and A. K. Bej Rapid detection of Vibrio vulnificus in shellfish and Gulf of Mexico water by real-time PCR. Appl. Environ. Microbiol. 70: Pertoft, H Fractionation of cells and subcellular particles with Percoll. J. Biochem. Biophys. Methods 44: Rodriguez-Lazaro, D., A. Jofre, T. Aymerich, M. Hugas, and M. Pla Rapid quantitative detection of Listeria monocytogenes in meat products by real-time PCR. Appl. Environ. Microbiol. 70: Scheu, P., K. Berghof, and U. Stahl Detection of pathogenic and spoilage micro-organisms in food with the polymerase chain reaction. Food Microbiol. 15: Sharma, V. K Detection and quantitation of enterohemorrhagic Escherichia coli O157, O111, and O26 in beef and bovine feces by real-time polymerase chain reaction. J. Food Prot. 65: Uyttendale, M., I. Van Hoorde, and J. Debevere The use of immunomagnetic separation (IMS) as a tool in a sample preparation method for direct detection of Listeria monocytogenes in cheese. Int. J. Food Microbiol. 54: Wolffs, P., R. Knutsson, B. Norling, and P. Rådström Rapid quantification of Yersinia enterocolitica in pork samples by a novel sample preparation method, flotation, prior to real-time PCR. J. Clin. Microbiol. 42: Wolffs, P., B. Norling, J. Hoorfar, M. Griffiths, and P. Radstrom Quantification of Campylobacter spp. in chicken rinse samples by using flotation prior to real-time PCR. Appl. Environ. Microbiol. 71: