Introduction

Foodborne diseases present a major global health issue, and food safety is important for both consumers and producers in the food industry (Zhao et al. 2010). Shiga toxin-producing Escherichia coli is a foodborne pathogen that cause a range of symptoms in humans ranging from mild diarrhea to fatal complications, such as hemolytic–uremic syndrome (Wang et al. 2014). Beef and beef products are considered to be the main sources of E. coli infections in humans (Chapman et al. 2001). However, it is difficult to detect foodborne pathogens rapidly by conventional methods because the number of bacteria in food is very low (Dong et al. 2014).

To reduce food poisoning, particularly in the food industry, rapid pathogen detection methods for raw food materials and processed foods are required. These rapid detection methods must be sensitive enough for detection of pathogens at low concentrations that exist in and on foods. Although culture based methods are generally regarded as the gold standard for detecting foodborne pathogens, they are time-consuming (taking 3–4 days including biochemical testing) and labor intensive. To alternatives, nucleic-acid-based methods, such as real-time PCR, are widely used nowadays.

Loop-mediated isothermal amplification (LAMP) is another nucleic acid-based method that is known to be fast, accurate, and easy to perform (Lee et al. 2016). LAMP does not dissociate the DNA, but rather amplifies the DNA in 60 min under isothermal (60–65 °C) conditions using the autocycling strand displacement characteristics of large fragment Bst DNA polymerase (Techathuvanan et al. 2010, Yamazaki et al. 2011). Instead of a PCR thermocycler, DNA can be amplified using a low-cost temperature-holding device, such as a water bath. In addition, LAMP is 10–100 times more sensitive than PCR because of its low effect on inhibitors derived from foods (Techathuvanan et al. 2010; Zhao et al. 2010; Yamazaki et al. 2011). LAMP-amplified DNA can be detected by various techniques, such as electrophoresis, turbidity evaluation, or staining with a fluorescent dye (SYBR Green I) that specifically reacts with dsDNA (Techathuvanan et al. 2010; Yongkiettrakul et al. 2014).

Lateral flow immunoassays (LFAs) have been developed extremely useful for rapid point-of-care testing (POCT) of foodborne pathogens, and they are based on the principles of specific binding between antibodies and bacteria. However, their detection sensitivity is very low and can detect only a high pathogen concentration (105 CFU/mL) (Shan et al. 2015). In the current study, instead of whole bacteria cells, an amplified DNA fragment was used as an epitope. After tagging the DNA amplicon with digoxygenin (DIG) and biotin, it can be detected by an anti-DIG antibody and visualized with a gold particle conjugated to streptavidin (Singh et al. 2015). Thus, a positive amplification result can easily be detected by the naked eye.

In general, for microbiological analysis, a 25-g analytical unit is used and should be pre-enriched at a 1:9 sample-to-broth ratio. Thus, if microorganisms are to be detected from a 25-g food sample, the sample would be diluted in 225 mL of broth, resulting in a potential 10-fold loss in sensitivity because 1-mL aliquots are normally used for detection methods, such as PCR and LAMP. Thus, microbial culture enrichment is routinely conducted to increase the bacterial concentration over the detection limit. This step requires > 10 h, making rapid detection of foodborne bacteria impossible. However, in several detection methods, microbial culture enrichment is now replaced with other concentration methods. Maheux et al. concentrated bacteria using a filter (Maheux et al. 2011). Filter-based methods have the advantage of not only concentrating bacteria but also removing the DNA amplification reaction inhibitors originating from food (D’Urso et al. 2009; Wolffs et al. 2006).

Many inhibitors of PCR and LAMP are present in complex food matrices, and generally, pathogens exist at very low concentrations. Because of these challenges, no reliable biosensors with high selectivity and sensitivity are yet available for rapid pathogen detection in real food systems. Therefore, in this study, we developed a filtration-based LAMP–LFA method for the rapid detection of E. coli O157:H7 present in beef at low numbers without microbial enrichment. Escherichia coli O157:H7 cells were concentrated by filtration instead of microbial enrichment culture, and the target DNA was isothermally amplified by LAMP and visually detected by LFA. This method enabled fast and sensitive detection of E. coli O157:H7 in a real food system.

Materials and methods

Bacterial strain

The bacterial strain used in this experiment was E. coli O157:H7 (ATCC 35150). Bacteria were maintained in 0.5 mL of 50% glycerol and stored at −80 °C. During experiments, bacterial cultures were grown in Tryptic Soy Broth (Difco, Flanklin Lakes, USA) at 37 °C for overnight.

Detection of E. coli O157:H7 from artificially contaminated beef

Beef was obtained from a local market (Seoul, Korea). Samples were artificially contaminated with E. coli O157:H7. The strain was serially diluted 10-fold (105–100 CFU/g) in 0.85% saline solution, and 100 μL of each dilution was inoculated on 25 g of beef and mixed well for uniform distribution. Briefly, 25 g of each beef sample was sliced and placed into a stomacher bag. Each sliced beef was inoculated by adding 100 μL of appropriately diluted inoculum. The inoculum was rubbed into one surface of each inoculated sliced beef by manually massaging (AVERY et al. 1994). For DNA isolation and bacterial quantification, 25 g were taken from the inoculated beef and mixed with 225 mL of 0.85% NaCl saline and homogenized in a Pulsifier (Macrogen, Seoul, Korea) for 1 min. Thereafter, 1 mL of each mixture was added in a 1.5-mL tube and centrifuged at 12,000 rpm for 2 min. DNA was extracted using a TaKaRa DNA extraction kit (Takara Bio Inc., Shiga, Japan), according to the manufacturer’s protocol. To enumerate E. coli O157:H7 cells, serially diluted beef solution was plated on eosin methylene blue plates (EMB agar, Oxoid, Hampshire, UK) (Sagong et al. 2011; Kim et al. 2011). Plates were incubated at 37 °C for 24–48 h, and bacterial cells were counted.

Filter selection

Filtration was performed to concentrate E. coli O157:H7 cells. To select the most suitable filter, the recovery efficiency of the following five filters was tested: 0.45-μm cellulose nitrate filter (Whatman International Ltd., Maidstone, United Kingdom), 0.45-μm cellulose acetate filter (Toyo Roshi Kaisha, Ltd., Japan), 0.45-μm mixed cellulose ester filter (MCE, HYUNDAI Micro Co., Ltd., Korea), 0.45-μm polyethersulfon filter (PES, HYUNDAI Micro Co., Ltd., Korea), and 0.45-μm nylon filter (HYUNDAI Micro Co., Ltd., Korea). Escherichia coli O157:H7 was artificially inoculated on 25 g of beef. After adding 225-mL saline and homogenizing with a Pulsifier, the solution was filtered through the five different filters under a vacuum. After filtration, each filter was disassembled and placed on EMB agar. After incubation at 37 °C for 24–48 h, bacterial counts were determined. The recovery ratio of E. coli O157:H7 was determined by comparison with inoculated microbial loads and recovered counts on EMB agar. All measurements were performed in five replicates.

Escherichia coli O157:H7 recovery from filters

Among the five tested filters, the cellulose nitrate filter showed the highest recovery ratio and was chosen as a suitable filter. After filtering E. coli O157:H7 with a cellulose nitrate filter, three different recovery methods with different times and reps were tested for the recovery of attached or embedded E. coli O157:H7 from the filter. In all cases, the filter was placed in a sterile bag (7.5 × 12.5 cm; Nasco, Fort Atkinson, WI, USA), and 2.25 mL of 0.85% NaCl was added. The filter was then vortexed for 15, 30, or 60 s; stomached (Laboratory Blender Stomacher 400; Seward, MO, USA) for 20, 50, 100, or 200 reps; or sonicated for 15, 30, or 60 s. After the recovery method was performed, 1 mL was taken from 2.25 mL of saline and plated on non-selective media, TSA, and selective media (EMB), respectively. The recovery ratio was determined using the bacterial cell counts on TSA plates and the initial inoculating bacterial count. The recovery percentage was defined as follows:

$$ {\text{Recovery}}\,{\text{rate }}\left( {\text{\%}} \right) = \frac{{{\text{Number}}\,{\text{of}}\,{\text{bacteria}}\,{\text{recovered }}\,{\text{from}}\,{\text{the}}\,{\text{filter}}}}{{{\text{Initial}}\,{\text{inoculated}}\,{\text{bacteria }}\left( {{\text{CFU}} / {\text{mL}}} \right)}} \times 100 {\text{\% }} $$

The injured cell ratio was calculated as the percentage of cells that did not grow on the EMB plate but grew on the TSA plate.

Membrane filtration and isolation of DNA

Escherichia coli O157:H7 was cultured to give an initial bacterial concentration of 1.47 × 109 CFU/mL in TSB overnight and serially diluted with saline to a concentration of 106–102 CFU/mL; 100 μL of this diluted culture was inoculated on 25 g of beef. After mixing well, 225 mL of saline was added, and the beef was homogenized using a Pulsifier and subsequently filtered using a membrane filter (pore size, 0.45 μm) with a 47-mm diameter and vacuum pump (Welch, Ilmenau, Germany). After filtering the 225-mL solution, the filter was disassembled and put into 7.6 × 12.7-cm sterilized polyethylene filled with 2.25 mL of 0.85% NaCl saline. Escherichia coli O157:H7 cells were recovered or detached from the filter by stomaching for 200 reps. DNA was extracted using a TaKaRa MiniBEST Bacterial Genomic DNA Extraction Kit Ver.3.0 (TaKaRa, Dalian, China) from 1 mL of the 2.25-mL homogenate. The extracted DNA was used for LAMP and real-time PCR.

Primer design and conditions for the LAMP reaction

Primers

The Shiga-like toxin 2 (stx2) gene was selected as a target sequence for this method, as described by Zhao et al. (2010). The probe for E. coli O157:H7 was designed using Primer Explorer version 4 (http://primerexplorer.jp/lamp4.0.0/index.html) to target five distinct regions of the stx2 gene (GenBank Accession No. AF162758.1). Primer sequences are listed in Table 1.

Table 1 Primer sequence sets for LAMP and real-time PCR detection of E. coli O157:H7
Full size table

LAMP operating condition

The LAMP reaction was performed in a final volume of 25 µL of the reaction mixture containing 0.8 µM of each of the FIP and BIP inner primers, 1.6 µM of each of the F3 and B3 outer primers, 1.4 mM dNTP mix (ABM Inc., Ontario, Canada), 0.8 M betaine (Sigma-Aldrich, USA), 6 mM MgSO4 (Sigma-Aldrich, USA), 8 U of Bst DNA polymerase large fragment (Middleton, WI, USA), 1 × of the supplied buffer, and the template DNA extracted from inoculated samples. Hot-start LAMP was performed using a SimpliAmp Thermal Cycler (Thermo Fisher Scientific Inc. Grand Island, NY, USA) as follows. Reaction mixtures without Bst polymerase were prepared on ice and then denatured at 94 °C for 2 min. Subsequently, Bst polymerase was added to the reaction mixtures, and LAMP was performed at 58–70 °C for 60 min. LAMP products were analyzed by electrophoresis on a 2% agarose gel. The reaction temperature and time were optimized to 63.5 °C for 45 min for the amplification of E. coli O157:H7 DNA.

LFA for detection of LAMP-amplified product

An oligonucleotide probe was designed to the sequence spanning the FIP and BIP regions and labeled with biotin at the 5′-end (Table 1). Five microliters containing 20 pmol of the labeled DNA probe was hybridized with the LAMP-amplified product at 63 °C for 5 min. An 8-μL sample of LAMP amplicons was transferred to a new tube and mixed with 120 μL of an LFA dilution solution (BoreDa Biotech, Gyeonggi, Korea) and placed on a sample pad of an LFA strip (BoreDa Biotech). LFA strips were immersed in the mixed solution, and DNA amplification was detected as positive (two lines) or negative (one line) after 5 min.

Real-time PCR detection of E. coli O157:H7

Primers and the probe used in real-time PCR were synthesized by IDT (San Diego, CA, USA), and the nucleotide sequences are shown in Table 1. Real-time PCR for stx2 (E. coli O157:H7) was performed to compare its detection limit with the LAMP–LFA method. The reaction mixture for real-time PCR was prepared by adding 10 μL of 2 × TaqMan® universal Master Mix (Applied Biosystems, Foster City, USA), 2 μL of DNA solution, 2 μL of both forward and reverse primers (500 nM), and 1 μL of the probe (250 nM) in a final volume of 20 μL with tertiary distilled water. Real-time PCR was performed on a 96-well microplate using the StepOne Plus real-time PCR system (Applied Biosystems, Foster City, CA, USA). In the case of E. coli O157:H7, the reaction was performed at 50 °C for 2 min and 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Threshold cycle (Ct) values were analyzed using StepOne™ software (Applied Biosystems).

Statistical analysis

In this study, all experiments were performed in triplicate. Significant differences were assessed using analysis of variance at P < 0.05 for all parameters, and the mean comparisons were performed using Scheffe’s test. Statistical analysis was performed by analysis of variance using SPSS (ver. 23.0, IBM Corp., Armonk, NY, USA).

Results

Membrane filter selection and recovery

To select the most suitable filter for the rapid determination of foodborne pathogens, E. coli O157:H7 was diluted to 100 CFU/mL (1–2 CFU/mL) in 225 mL of saline and filtered through five filters. The filters were incubated on EMB plates; after 24 h, characteristic black colonies were enumerated, and the recovery ratio was calculated (Table 2). Uninoculated beef was used as a control but did not form black colonies on EMB agar after incubation. The cellulose nitrate filter showed 96% recovery, and the filtration process took 100 s to complete. In all five repetitive tests, all filters successfully filtered without becoming clogged. The cellulose acetate filter showed 95% recovery but failed in one test as the filter became clogged. The nylon filter showed the lowest recovery ratio. Filtration time through the cellulose nitrate, cellulose acetate and polyethersulfone filters (filtration time of < 1 min) was significantly short (P < 0.05) than that through other filters. However, the polyethersulfone filter was lower in recovered cell (%) than the other two filters. The recovery efficiency of the filter was significantly higher (P < 0.05) in cellulose nitrate and cellulose acetate. The cellulose nitrate filter was similar in recovery efficiency to cellulose nitrate filter, but the production rate of injured cells was high (data not shown). If the percentage of injured cells is high, it may be a limit to detect low concentrations of bacteria. Therefore the cellulose nitrate filter showed the best performance and was selected for further filtration-based microbial concentration methods because it showed higher recovery ratio, speed, and reproducibility than other tested filters.

Table 2 Comparison of filter types for detecting E. coli O157:H7 in beef
Full size table

To detect the E. coli O157:H7 by molecular detection methods, such as PCR or LAMP, DNA must be isolated from E. coli O157:H7. DNA preparation of E. coli O157:H7 in a filter is very difficult, and it is more convenient to detach bacteria from the filter and then isolate the DNA. To identify the most efficient method for detaching E. coli O157:H7 from the cellulose nitrate filter after filtration, the recovery ratio was measured using three methods (vortexing, sonication, and stomaching) at different times (Table 3). The recovery ratio increased with increasing (P < 0.05) treatment time and frequency in all treatments. The percentage of injured cells in all treatments was not significantly different (P > 0.05). Vortexing, sonication, and stomaching methods recovered 41, 85, and 96% of cells from the filter, respectively. Finally, 200 cycles of stomaching were selected as the most efficient recovery method, which took approximately 60 s.

Table 3 Comparison of methods for cell recovery from a cellulose nitrate filter
Full size table

Optimization of temperature and time for LAMP

In LAMP, operating temperature and time are very important. To determine the optimal temperature and time combination for the LAMP assay, DNA was isolated from 106 CFU/mL of E. coli O157:H7, and LAMP was performed at different temperatures and times. LAMP amplicons showed a ladder-like pattern on an agarose gel due to the secondary structures of LAMP amplicons (Fig. 1). To determine the optimum temperature, reactions were conducted at 58–72 °C for 60 min in a thermocycler fixed at one temperature. LAMP amplicons of E. coli O157:H7 were detected at 58–66.9 °C. We confirmed the amplification of LAMP in the range of 58 °C and 66.9 °C. Of the five amplifiable temperatures, 63.5 °C were < 100 bp being the least and > 1500 bp the most. Therefore we used an amplification temperature of 63.5 °C. To optimize the reaction time, LAMP amplicons were obtained at 25, 30, 35, 40, 45, 50, and 60 min. It was thought that LAMP amplicons reached the highest concentration at 45 min (data not shown); therefore, 45 min was selected as the optimal operating time.

Fig. 1
figure 1

LAMP product profiles at different reaction temperatures. Lane M: DNA marker (DL100); Lanes 1–8: LAMP reactions performed for 60 min at 58 °C, 58.9 °C, 60.8 °C, 63.5 °C, 66.9 °C, 69.7 °C, 71.2 °C and 72 °C, respectively

Full size image

Comparison of detection limits in filtered and unfiltered samples

Ten-fold serial dilutions of DNA from E. coli O157:H7-inoculated beef were used as templates for LAMP and real-time PCR to determine the detection sensitivity (Table 4; Fig. 2). In real-time PCR analysis, the sensitivity of unfiltered beef was 103 CFU/g, and filtered beef showed a 37.04 Ct value at 102 CFU/g, thus improving the sensitivity by 10-fold (Table 4). In LAMP, the sensitivity of unfiltered beef was 103 CFU/g and that of filtered beef was 101 CFU/g (Fig. 2). Filtering improves the sensitivity of real-time PCR and LAMP. LAMP–LFA showed 10 times higher sensitivity than real-time PCR.

Table 4 Quantitation of E. coli O157:H7 in artificially contaminated beef by filtration and real-time PCR
Full size table
Fig. 2
figure 2

LAMP sensitivity increased after filtration for detecting E. coli O157:H7 in inoculated beef by electrophoresis (top panel) and LFA (bottom panel). a Unfiltered samples and b filtered samples. A volume of 4 µL of LAMP amplicons from 25 µL reactions was loaded per lane. Lane M: molecular marker; lane N: negative control

Full size image

Discussion

In this study, we successfully established a new filtration-based LAMP–LFA method targeting the stx2 gene of E. coli O157:H7. The new filtration-based assay could process 225-mL buffer containing 25-g sample and could detect a least 101-CFU/g (14–15 CFU/g) bacteria in 2.25-mL aliquot of the 225-mL sample within 3 h.

In artificially inoculated beef, the sensitivity of filtration-based LAMP–LFA was 10 CFU/g (0.8 CFU per reaction), which was 10 times more than that of the filtration-real-time-PCR assay (102 CFU/g; 8 CFU per reaction). In a previous study, the detection limit of a filtration-based LAMP–LFA assay for E. coli O157:H7 was more sensitive than that of an unfiltered LAMP assay using spiked food samples (> 104 CFU mL). Wang et al. (2012) and Zhao et al. (2010) reported that the detection limit of E. coli O157:H7 using LAMP was 10 pg (104 CFU/mL) without using enrichment. The sensitivity of their methods was low to detect foodborne pathogens present in very small amounts (< 103 CFU/mL). In addition, no previous study has been conducted on DNA detection with LAMP using LFA after concentrating the sample by filtration.

The enhanced sensitivity of the methods reported here is believed to be because of the following three reasons. First, LAMP has higher specificity than conventional PCR and rapidly amplifies DNA. LAMP holds great potential as a rapid and sensitive molecular method for detecting DNA from various materials, including foods. Although PCR is inhibited by various food ingredients, such inhibitors do not affect the LAMP assay (Plutzer et al. 2010), which improves the sensitivity of detection. This has already been demonstrated in previous studies, in which the effectiveness of PCR and LAMP was compared using oyster and shrimp samples (Surasilp et al. 2011; Prompamorn et al. 2011). Second, the membrane filtration method physically concentrates E. coli O157:H7. In this study, homogenized, filtered E. coli O157:H7 cells could be efficiently recovered from 2.25-mL aliquot of the 225-mL sample, theoretically resulting in a 100-fold increase in concentration. Several studies have reported the use of a filtration method for similar reasons (Wolffs et al. 2006; Kaevska and Slana 2015). Third, filtration enhances detection sensitivity by removing inhibitors, such as protein and lipid components (Rossen et al. 1992).

LFA can be easily performed by passing on assay buffer through the membrane and can decrease the total time of LAMP assays by eliminating electrophoresis. Furthermore, the LAMP–LFA assay provides more specificity because it involves hybridization with a specific probe to the LAMP amplicons (Prompamorn et al. 2011).

The use of filtration in combination with LAMP–LFA does not require specialized expertise, and the enrichment step increases the assay’s sensitivity. The entire process for detecting E. coli O157:H7 in beef comprises sequential steps of filtration, LAMP, and LFA; the total time for detection can be decreased to ≤ 3 h.

Conclusion

In this study, a new filtration-based LAMP–LFA method was developed to detect the presence of E. coli O157:H7 in beef. The filtration-based pretreatment step effectively reduced the overall analysis time and improved sensitivity by concentrating bacteria up to the detection limit to overcome the need for a microbial enrichment step. The sensitivity of real-time PCR and LAMP of filtered treatment was improved 10–100 fold compared to unfiltered controls. The concentration of 101 CFU/g of E. coli O157:H7 could be detected within 3 h without microbial enrichment. LFA was used to confirm the detection of E. coli O157:H7 with improved accuracy and simplicity. This new filtration-based LAMP–LFA method is potential for POCT because it allows for the detection of E. coli O157:H7 in ≤ 3 h at an isothermal temperature and does not require complex equipment or trained technical personnel. Our study may enable an approach that the food industry can detect before delivering contaminated food.