Viable but Nonculturable Escherichia coli O157:H7 and Salmonella enterica in Fresh Produce: Rapid Determination by Loop-Mediated Isothermal Amplification Coupled with a Propidium Monoazide Treatment

VBNC pathogenic bacteria pose a potential risk to the food industry because they do not multiply on routine microbiological media and thus can evade detection in conventional plating assays. Both E. coli O157:H7 and S. enterica have been reported to enter the VBNC state under a range of environmental stress conditions and to resuscitate under favorable conditions and are a potential cause of human infections. PMA-LAMP methods developed in this study provide a rapid, sensitive, and specific way to determine levels of VBNC E. coli O157:H7 and S. enterica in fresh produce, which potentially decreases the risks related to the consumption of fresh produce contaminated by enteric pathogens in this state. PMA-LAMP can be further applied in the field study to enhance our understanding of the fate of VBNC pathogens in the preharvest and postharvest stages of fresh produce.

ABSTRACT

Escherichia coli O157:H7 and Salmonella enterica are leading causes of foodborne outbreaks linked to fresh produce. Both species can enter the “viable but nonculturable” (VBNC) state that precludes detection using conventional culture-based or molecular methods. In this study, we assessed propidium monoazide-quantitative PCR (PMA-qPCR) assays and novel methods combining PMA and loop-mediated isothermal amplification (LAMP) for the detection and quantification of VBNC E. coli O157:H7 and S. enterica in fresh produce. The performance of PMA-LAMP assays targeting the wzy gene of E. coli O157:H7 and the agfA gene of S. enterica and the performance of PMA-qPCR assays were compared in pure culture and spiked tomato, lettuce, and spinach. No cross-reaction was observed in the specificity tests. The values representing the limit of detection (LOD) seen with PMA-LAMP were 9.0 CFU/reaction for E. coli O157:H7 and 4.6 CFU/reaction for S. enterica in pure culture and were 5.13 × 10³ or 5.13 × 10⁴ CFU/g for VBNC E. coli O157:H7 and 1.05 × 10⁴ or 1.05 × 10⁵ CFU/g for VBNC S. enterica in fresh produce, representing results comparable to those obtained by PMA-qPCR. Standard curves showed correlation coefficients ranging from 0.925 to 0.996, indicating a good quantitative capacity of PMA-LAMP for determining populations of both bacterial species in the VBNC state. The PMA-LAMP assay was completed with considerable economy of time (30 min versus 1 h) and achieved sensitivity and quantitative capacity comparable to those seen with a PMA-qPCR assay. PMA-LAMP is a rapid, sensitive, and robust method for the detection and quantification of VBNC E. coli O157:H7 and S. enterica in fresh produce.

IMPORTANCE VBNC pathogenic bacteria pose a potential risk to the food industry because they do not multiply on routine microbiological media and thus can evade detection in conventional plating assays. Both E. coli O157:H7 and S. enterica have been reported to enter the VBNC state under a range of environmental stress conditions and to resuscitate under favorable conditions and are a potential cause of human infections. PMA-LAMP methods developed in this study provide a rapid, sensitive, and specific way to determine levels of VBNC E. coli O157:H7 and S. enterica in fresh produce, which potentially decreases the risks related to the consumption of fresh produce contaminated by enteric pathogens in this state. PMA-LAMP can be further applied in the field study to enhance our understanding of the fate of VBNC pathogens in the preharvest and postharvest stages of fresh produce.

INTRODUCTION

The “viable but nonculturable” (VBNC) state is a unique bacterial physiological state during which cells lose their capacity to multiply on microbiological media while maintaining metabolic activity (1). VBNC bacteria demonstrate high tolerance of various stresses, including starvation, high or low temperatures, high salinity, and extreme pH (2–6). Entry into the VBNC state has been regarded as a survival strategy in response to adverse environmental conditions. Previous studies have reported that VBNC bacteria are abundant in the natural environment, including aquatic and agricultural ecosystems (7, 8). A total of 67 species of pathogenic bacteria have been shown to enter the VBNC state, including food-transmissible Campylobacter, Listeria, Shigella, Shiga toxin-producing Escherichia coli (STEC), and Salmonella (9). Cumulative evidence indicates that VBNC cells can resuscitate under suitable environmental conditions, regain virulence, and subsequently cause infection. For example, VBNC E. coli O157:H7 was the putative cause of an outbreak associated with the consumption of salted salmon roe (10). Moreover, Listeria monocytogenes cells in the VBNC state were shown to resuscitate in egg yolk, to infect human epithelial cell line HT-29, and to colonize the mouse spleen (11).

E. coli O157:H7 and S. enterica are leading causes of foodborne illnesses, with annual cases totaling 73,000 and 120,000 in the United States, respectively (12, 13). Fresh produce has emerged as one of the major vehicles linked to both E. coli O157:H7 and S. enterica outbreaks (14). A recent E. coli O157:H7 outbreak across 36 states in the United States was associated with romaine lettuce and resulted in a total of 210 illnesses; 96 hospitalizations, including 27 cases of hemolytic uremic syndrome; and 5 deaths (15). In 2015, an outbreak implicating cucumbers contaminated with S. enterica serovar Poona in the United States led to 907 clinically confirmed infections and 6 deaths (16). Previous studies have hinted at the presence of S. enterica and E. coli O157:H7 in the VBNC state in fresh produce (17–19). For example, Dinu and Bach reported that E. coli O157:H7 could enter the VBNC state in the plant phyllosphere (18). Moyne and coauthors showed that E. coli O157:H7 could enter the VBNC state shortly after inoculation onto field lettuce (19). There is mounting evidence that Salmonella can enter the VBNC state on plants when subjected to a variety of stresses, such as limited nutrient availability, low temperature, exposure to UV light, or the presence of chlorine at concentrations between 3 and 100 ppm, which are common conditions during postharvest handling and processing of fresh produce (20). Taken together, the enhanced survival of VBNC cells and difficulties in their detection complicate efforts to lessen the burden of foodborne illnesses associated with E. coli O157:H7 and S. enterica in fresh produce. Clearly, additional practical and effective detection methods are required to assess the risks posed by these pathogens at critical stages along the farm-to-fork continuum.

Estimation of culturable cell counts using a conventional plating assay is normally the first step to quantify VBNC cells in samples containing cells in various physiological states. Differential staining and direct microscopic counts of metabolically active cells by the use of a LIVE/DEAD BacLight bacterial viability kit are then performed to derive estimates of unculturable cells assumed to be in the VBNC state from the differences in cell numbers obtained by the two methods (21–23). However, this method is labor-intensive and time-consuming, and the results have limited accuracy (24). Although PCR-based methods are widely applied in the detection and quantification of S. enterica and E. coli O157:H7 in foods because of their sensitivity and specificity, none can reliably differentiate between culturable, VBNC, and dead cells (25, 26). Modifications that include the use of intercalating dyes for the detection of VBNC cells have been proposed (27). Intercalating dyes applied before PCR analysis can assist in discrimination between live and dead cells by preventing the amplification of dead cell DNA. Propidium monoazide (PMA) has been used for this purpose because it can enter nonviable cells through damaged bacterial cell membranes and intercalate with genomic DNA (gDNA) after photoactivation. Amplification of DNA from dead cells was thereby inhibited during the PCR in a previous study (28), and only counts of viable cells were determined. Although PMA coupled with PCR-based methods can yield reasonably accurate estimates of VBNC cells in a mixed sample, all require access to expensive analytical equipment and are difficult to adapt for use by resource-limited laboratories or for field studies (27).

A nucleic-acid-based amplification method termed loop-mediated isothermal amplification (LAMP) has emerged as a promising alternative for the detection of foodborne bacteria and viruses (29–33). LAMP is an autocycling and strand displacement DNA synthesis method that does not require a sophisticated thermal cycling instrument (34). LAMP utilizes 4 to 6 specially designed primers and a strand-displacing DNA polymerase to amplify the targeted DNA under isothermal (∼65°C) conditions within 1 h (35). Consequently, the reaction can be carried out faster and at a much lower cost than by other PCR methods. No gel electrophoresis is required to interpret the results, because the by-product of isothermal amplification is visible as white precipitate. In addition, this method has proven more sensitive than other PCR-based methods (36). Therefore, comparisons between culturable cell numbers obtained by the conventional cultural assays and viable cell numbers determined using a PMA-LAMP assay should enable the detection of VBNC cells in various sample matrices, such as agri-food commodities.

Detection of bacteria in the VBNC state using LAMP has been reported recently (37–39). Zhong and coauthors developed a PMA-LAMP assay to measure VBNC Staphylococcus aureus levels at a limit of detection (LOD) of 17 CFU/ml in pure culture (39). In another study, PMA treatment combined with LAMP was reported to enable detection of 14 CFU/g of VBNC Vibrio parahaemolyticus in seafood samples (37). To date, PMA-LAMP has not been applied to the detection of VBNC foodborne pathogens in fresh produce. Consequently, we developed a rapid and sensitive PMA-LAMP assay that can be coupled with a conventional cultural assay for the detection and quantification of E. coli O157:H7 and S. enterica in the VBNC state in various produce commodities. The performance of the assay was compared with that of a well-established propidium monoazide-quantitative PCR (PMA-qPCR) assay, and the applicability of the approach was verified through the detection of VBNC bacterial cells in tomato, lettuce, and spinach samples.

RESULTS AND DISCUSSION

Optimization of PMA concentration.

Pretreatment using PMA was integrated with LAMP to eliminate the amplification signals from the dead cells to allow differentiation of VBNC cells. Delivery of an optimal concentration of PMA is critical, as insufficient amount of PMA may not completely block the DNA signal from dead cells, leading to overestimation of viable cell concentrations (40), while excessive concentrations can interfere with DNA amplification of viable cells and result in underestimation (41). In the current study, 10 μM PMA was determined to be optimal for pretreatment for both E. coli O157:H7 and S. enterica, as that concentration completely eliminated signal from up to 10⁶ CFU/ml of dead cells and as no adverse effect was observed upon amplification of viable bacterial cells (Fig. 1). Previous studies reported that the optimal concentration of PMA for determination of the viability of different bacterial species ranged from 2 to 100 μM (42–44). The differences in the sensitivity of cell membranes of different bacteria and in the penetration power of PMA with respect to the corresponding viable and dead cells (45) might result in variations of PMA concentration. Different incubation times, levels of power and types of light, and photoactivation times were also applied in these studies.

FIG 1.

Optimization of working concentration of PMA. Viable and thermally inactivated dead cells were separately treated with different concentrations of PMA (0, 1, 5, 10, 25, 50, AND 100 μmol/liter). (A) Cell concentrations of E. coli O157:H7 were determined by LAMP assay using viable and dead cells (2.3 × 10⁶ CFU/ml) treated with PMA separately. (B) Cell concentrations of S. enterica were determined by LAMP assay using viable and dead cells (3.4 × 10⁶ CFU/ml) treated with PMA separately. Error bars were drawn based on results from three independent replicates.

Specificity of the PMA-LAMP assay.

The specificity of PMA-LAMP was evaluated using a collection of 31 strains of both Gram-negative and Gram-positive bacteria (listed in Table 1). Two non-toxin-encoding genes, wzy and agfA, were selected as the targets in consideration of future studies for assay validation through field studies that would necessitate the use of attenuated bacterial strains (46, 47). The high specificity of the wzy gene for the detection of E. coli O157:H7 was verified by nucleotide BLAST analysis (data not shown) and by the test results of PMA-LAMP (Table 1). Specifically, the threshold time (T_t) values for the detection of different E. coli O157:H7 strains were highly consistent (18.4 ± 0.6 min to 19.2 ± 0.4 min) and confirmed that this highly conserved gene is an appropriate target for the broad detection of different E. coli O157:H7 strains. No positive signal was obtained for any non-E. coli O157:H7 strain. A high level of specificity of the wzy gene was also reported in previous attempts to develop LAMP-based detection of E. coli O157:H7 (48). For S. enterica testing, the T_t values of PMA-LAMP for nine S. enterica strains ranged from 15.3 ± 0.9 min to 17.9 ± 0.8 min, with an average of 16.1 ± 0.9 min. Application of S. enterica-specific PMA-LAMP did not result in initiation of amplification with 22 non-S. enterica strains, for which no T_t value was obtained. In addition, nucleotide BLAST analysis indicated that the agfA gene is highly conserved. Taking the data together, the PMA-LAMP assays targeting wzy and agfA showed 100% specificity for the detection of E. coli O157:H7 and S. enterica, respectively.

TABLE 1.

Bacterial strains used for LAMP specificity tests

Species and serovar	Strain	LAMP resulta
Salmonella enterica serovar
Enteritidis	43353	+
	0EA2669	+
	3512H	+
	ME13	+
	ME14	+
	PT30	+
Heidelberg	1072	+
Typhimurium	SL1344	+
	1228	+
Escherichia coli
O157:H7	EDL 933	+
	ATCC 700728	+
	ATCC 43895	+
	ATCC 43888	+
	ATCC 43889	+
	ATCC 43890	+
	ATCC 35150	+
O8:H16	ATCC 43888	−
O26:H11	M38539	−
O73:H2	T53317	−
O103:H2	W20260	−
O111:Hnon-motile	M21374	−
O121:H19	H32130	−
O118:H16	T53809	−
O165:Hnon-motile	T71841	−
Pseudomonas aeruginosa	ATCC 9721
	PA14
Listeria monocytogenes	15B88
	15B98
Staphylococcus aureus	MRSA-10
Campylobacter jejuni	ATCC 33560
Campylobacter coli	RM1875

^a+, positive LAMP result; −, negative LAMP result.

Sensitivity and quantitative capacity of the PMA-LAMP assay with pure cultures.

The levels of sensitivity and quantification capacity of both the PMA-LAMP and PMA-qPCR assays were evaluated and compared in mixtures of viable and thermally inactivated cells (Fig. 2). Dead cells were included at a concentration of 10⁶ CFU/ml to mimic detection under real-world scenarios wherein cells are expected to be found in a range of metabolic states. The optimized PMA-LAMP assay consistently detected E. coli O157:H7 at levels ranging from 9.0 × 10⁵ to 9.0 × 10¹ CFU/reaction (equivalent to 4.5 × 10⁸ to 4.5 × 10⁴ CFU/ml; samples 1 to 5 in Fig. 2A), with T_t values ranging from 22.0 ± 0.6 to 30.0 ± 1.4 min. The limit of detection (LOD) of the PMA-LAMP assay for E. coli O157:H7 was 9 CFU/reaction (equivalent to 4.5 × 10³ CFU/ml), as two of four replicates at a concentration of 9 CFU/reaction (sample 6 in Fig. 2A) showed an average T_t value of 41.5 min. No amplification was observed when the target concentration was below 9 CFU/reaction (sample 7 in Fig. 2A). The correlation coefficient (R²) for the quantification equation for the E. coli O157:H7-specific PMA-LAMP assay was 0.995, suggesting a good quantitative capacity of this assay (Fig. 2B). The data for the cell concentration of 9 CFU/reaction were excluded, and quantitative capacity ranged from 9.0 × 10⁵ to 9.0 × 10¹ CFU/reaction. The LOD of PMA-LAMP for E. coli O157:H7 was consistent with previous studies where levels of 0.7 to 16 CFU/reaction have been reported (29, 49). For example, Ravan and coworkers demonstrated that the LOD of LAMP for the detection of E. coli O157:H7 was 5 CFU/reaction in pure bacterial culture (50).

FIG 2.

Sensitivity and quantitative capability of PMA-LAMP assays for the detection of viable E. coli O157:H7 (A, B, and C) and S. enterica (D, E, and F) in pure culture. (A) Representative PMA-LAMP amplification graph generated from 10-fold serially diluted viable E. coli O157:H7 cells with counts ranging from 9.0 × 10⁵ to 9.0 × 10⁻¹ CFU/reaction (equivalent to 4.5 × 10⁸ to 4.5 × 10² CFU/ml; samples 1 to 7, respectively) in the background of 4.6 × 10³ CFU/reaction (equivalent to 2.3 × 10⁶ CFU/ml) of dead cells; sample 8 was a negative control with sterile water. (B and C) Standard curves of PMA-LAMP (B) and PMA-qPCR (C) for the detection of 10-fold serially diluted viable E. coli O157:H7 cells. (D) Representative PMA-LAMP amplification graph generated from 10-fold serially diluted viable S. enterica cells with counts ranging from 4.6 × 10⁵ to 4.6 × 10⁻¹ CFU/reaction (equivalent to 2.3 × 10⁸ to 2.3 × 10² CFU/ml; samples 1 to 7, respectively) in the background of 6.8 × 10³ CFU/reaction (equivalent to 3.4 × 10⁶ CFU/ml) of dead cells; sample 8 was a negative control with sterile water. (E and F) Standard curve of PMA-LAMP (E) and PMA-qPCR (F) for the detection of 10-fold serially diluted viable S. enterica cells. Error bars were obtained based on results from four replicates; data for sample 6 were excluded from both standard curves.

In comparison, the LOD of the PMA-qPCR assay for E. coli O157:H7 was also 9 CFU/reaction, with an average cycle threshold (C_T) value of 31.6 cycles (∼1 h). The quantitative range was from 9.0 × 10⁵ to 9.0 CFU/reaction with an R² value of 0.998 (Fig. 2C). Results from similar work indicated that the LAMP and qPCR assays achieved comparable sensitivities (i.e., 1 to 20 CFU/reaction) for the detection of E. coli O157:H7 in pure culture (36), but a shorter detection time was achieved by PMA-LAMP (i.e., 30 min versus 1 h).

The presence of viable S. enterica in pure cultures was consistently detected at concentrations ranging from 4.6 × 10⁵ to 4.6 × 10¹ CFU/reaction using the PMA-LAMP assay (equivalent to 2.3 × 10⁸ to 2.3 × 10⁴ CFU/ml) (samples 1 to 5 in Fig. 2D). The corresponding T_t values ranged from 15.9 ± 0.6 to 23.3 ± 1.0 min. At a concentration of 4.6 CFU/reaction (equivalent to 2.3 × 10³ CFU/ml) (sample 6 in Fig. 2D), two of four replicates showed an average T_t value of 30.1 min, and no amplification was observed at cell concentrations below 4.6 CFU/reaction (sample 7 in Fig. 2D), which implied a LOD of 4.6 CFU/reaction. High quantification capability (R² = 0.963), ranging from 4.6 × 10⁵ to 4.6 × 10¹ CFU/reaction, was obtained for S. enterica (Fig. 2E). The LOD of the PMA-qPCR assay was 4.6 CFU/reaction, and the quantitative range was from 4.6 × 10⁵ to 4.6 CFU/reaction, with C_T values ranging from 15.3 to 32.9 cycles (∼1 h) (Fig. 2F). The LOD (4.6 CFU/reaction) of the PMA-LAMP assay for S. enterica was also comparable to that of PMA-qPCR and was in agreement with another study where 1.8 CFU/reaction viable S. enterica in pure culture was detected by LAMP (51). As VBNC cells are still alive and maintain the integrity of cell structure and function (1), it is appropriate to further apply PMA-LAMP for the detection of E. coli O157:H7 and S. enterica in the VBNC state after their loss of culturability as confirmed by the culture-based methods.

Induction and detection of the VBNC state in E. coli O157:H7 and S. enterica under conditions of osmotic stress.

Both E. coli O157:H7 and S. enterica can enter the VBNC state under conditions of various stresses, including exposure to osmotic stress, starvation, or low temperatures (52, 53). Osmotic pressure is highly relevant to food processing and is exploited for the extension of food shelf life by changing the level of osmotic pressure and removing free water from cell membranes by the use of methods such as salting fish and candying fruit (54). In the current study, we induced entry of E. coli O157:H7 and S. enterica into the VBNC state by using 7% (wt/vol) NaCl solution to mimic the conditions of osmotic stress that occur during food processing (55), which has also been shown to be more efficient than nutrient depletion or exposure to temperature extremes for rapid induction of the VBNC state in a wide range of bacterial species (10, 53, 56). Bacterial culturability and viability were monitored over time using a cultural method and the PMA-LAMP, and the difference between the culturable cell counts and viable cell counts represented the amount of VBNC cells. As shown in Fig. 3A, the gradual reduction in colony counts and sustained quantitative detection of viable cells by PMA-LAMP indicated the transition of viable cells from normal state to the VBNC state upon prolonged exposure to 7% (wt/vol) NaCl. E. coli O157:H7 was no longer culturable after 6 days (<0.3 CFU/ml), while the viable cell concentration determined by PMA-LAMP was 10⁷ CFU/ml, indicating that these cells had been induced to completely enter into the VBNC state. S. enterica colony counts also fell immediately, and no colonies were seen on day 15 (Fig. 3B), at which time 10⁶ CFU/ml of viable cells was detected by the PMA-LAMP assay, Hence, approximately 1% of the S. enterica and 10% of E. coli O157:H7 cell populations were induced to enter the VBNC state after exposure to 7% (wt/vol) NaCl for 6 and 15 days, respectively. A longer duration was required for induction of the VBNC state in S. enterica, indicating that that species may be more resistant to osmotic stress than E. coli O157:H7. Exposure to high NaCl concentrations has been used in the previous studies to induce E. coli and S. enterica to enter the VBNC state. For example, E. coli O157:H7 was shown to enter the VBNC state in salmon roe containing 13% NaCl (10). It has been suggested that the differences in induction speeds and final VBNC cell concentrations may represent the influence of the sigma factor RpoS (55, 57), a central regulator of bacterial responses to environmental stresses (58).

FIG 3.

Survival curves of E. coli O157:H7 (A) and S. enterica (B) incubated in 7% (wt/vol) NaCl solution. Open circles (∘) represent the culturable cell counts of E. coli O157:H7 determined by the plating assay. Closed circles (•) represent the viable cell counts of E. coli O157:H7 determined by the PMA-LAMP assay. Open squares (☐) represent the culturable cell counts of S. enterica determined by the plating assay. Closed squares (■) represent the viable cell counts of S. enterica determined by the PMA-LAMP assay. Error bars were calculated based on results from three replicates. The difference between the viable cell counts and culturable cell counts represents the number of VBNC cells.

Fluorescent microscopy coupled with a LIVE/DEAD BacLight bacterial viability kit was used to confirm the presence of bacteria in the VBNC state. As shown in Fig. 4A and C, the majority of E. coli O157:H7 and S. enterica cells in the exponential phase were stained by SYTO 9 (shown in fluorescent green), which is indicative of bacterial cell membrane integrity. After 6 days of exposure to 7% (wt/vol) NaCl, no culturable E. coli O157:H7 cells were present as determined by the plating assay whereas a proportion of cells still appeared green (Fig. 4B), indicating that they remained viable but in the VBNC state. S. enterica was no longer culturable after 15 days of incubation in 7% (wt/vol) NaCl solution, but the presence of viable cells (stained in green in Fig. 4D) was also indicative of a transition to the VBNC state. It should also be noted that VBNC cells of both E. coli O157:H7 and S. enterica cells exposed to 7% (wt/vol) NaCl were smaller and assumed coccoid rather than the rod shapes evident in exponential-phase cells (Fig. 4B and D). Alterations in cellular forms can be attributed to changes in cell membrane composition. Peptidoglycan, a major constituent of the cell membrane, plays an important role in cell division and in growth-phase shift. It was previously shown that peptidoglycan demonstrated a high degree of cross-linking when E. coli cells entered the VBNC state and became coccoid (59).

FIG 4.

Micrographs of bacterial cells stained by the use of a LIVE/DEAD BacLight bacterial viability kit under conditions of fluorescence microscopy. (A) E. coli O157:H7 in the exponential phase. (B) E. coli O157:H7 cells incubated in 7% (wt/vol) NaCl solution for 6 days. (C) S. enterica in the exponential phase. (D) S. enterica cells incubated in 7% (wt/vol) NaCl solution for 15 days. Cells that appear green were viable, while red cells were dead. In the experiments represented by panels B and D, no colonies were detected by the plating assay; thus, the cells stained in green were considered to be in the VBNC state (n = 3).

Quantification of VBNC E. coli O157:H7 and S. enterica in fresh produce with the PMA-LAMP assay.

The performance of the PMA-LAMP assay for the detection and quantification of VBNC cells was also assessed in samples of fresh produce inoculated with E. coli O157:H7 and S. enterica, in which the VBNC state was induced by exposure to 7% (wt/vol) NaCl for 6 days and 15 days, respectively. Lettuce, tomato, and spinach samples spiked with VBNC E. coli O157:H7 and S. enterica at concentrations ranging between 10³ and 10⁸ CFU/g were analyzed using the PMA-LAMP and PMA-qPCR assays. Table 2 summarizes the sensitivity and quantitative capability data for both approaches. The LOD of VBNC E. coli O157:H7 analyzed by PMA-LAMP was 5.13 × 10³ CFU/g in romaine lettuce and was 5.13 × 10⁴ CFU/g in tomato and spinach. In comparison, analysis by PMA-qPCR yielded an LOD of 5.13 × 10⁴ CFU/g for all types of fresh produce. The standard curves for PMA-LAMP and PMA-qPCR assays for E. coli O157:H7 in spiked fresh produce are shown in Fig. S1 in the supplemental material. Consistent quantification of VBNC E. coli O157:H7 cells at levels ranging between 5.13 × 10⁴ or 5.13 × 10⁵ and 5.13 × 10⁸ CFU/g was achieved in lettuce, tomato, and spinach by PMA-LAMP (R² = 0.960 to 0.996) and at levels ranging between 5.13 × 10⁴ and 5.13 × 10⁸ CFU/g by PMA-qPCR (R² = 0.995 to 0.999). The LOD of PMA-LAMP for VBNC S. enterica ranged from 1.05 × 10⁴ CFU/g in tomato samples to 1.05 × 10⁵ CFU/g in lettuce and spinach samples (Table 2). PMA-qPCR was able to detect VBNC S. enterica at concentrations of no less than 10⁴ CFU/g in all fresh produce. Concentrations ranging from 1.05 × 10⁵ to 1.05 × 10⁸ CFU/g were quantified by PMA-LAMP (R² = 0.925 to 0.962) and by PMA-qPCR (R² = 0.992 and 0.999) (Fig. S2). Overall, the sensitivity of PMA-LAMP was slightly lower than that of the PMA-qPCR assays for the detection of VBNC S. enterica in fresh produce, while the quantitative capacities were comparable.

TABLE 2.

Sensitivity and quantitative capacity of PMA-LAMP and PMA-qPCR assays for the detection of VBNC cells in spiked fresh produce^a

Sample type	Method	LOD (CFU/g)b	Quantification equation	Linear R2
E. coli O157:H7-spiked lettuce	PMA-LAMP	5.13 × 103	y = −1.20x + 26.35	0.966
	PMA-qPCR	5.13 × 104	y = −3.54x + 50.90	0.999
E. coli O157:H7-spiked tomato	PMA-LAMP	5.13 × 104	y = −1.97x + 37.88	0.960
	PMA-qPCR	5.13 × 104	y = −3.25x + 50.13	0.995
E. coli O157:H7-spiked spinach	PMA-LAMP	5.13 × 104	y = −3.63x + 51.77	0.996
	PMA-qPCR	5.13 × 104	y = −3.55x + 51.28	0.997
S. enterica-spiked lettuce	PMA-LAMP	1.05 × 105	y = −2.93x + 38.03	0.925
	PMA-qPCR	1.05 × 104	y = −3.30x + 50.15	0.996
S. enterica-spiked tomato	PMA-LAMP	1.05 × 104	y = −4.09x + 52.84	0.962
	PMA-qPCR	1.05 × 104	y = −3.31x + 48.64	0.992
S. enterica-spiked spinach	PMA-LAMP	1.05 × 105	y = −2.80x + 37.63	0.961
	PMA-qPCR	1.05 × 104	y = −3.68x + 49.64	0.999

^aQuantitative equation and R² data were calculated based on sensitivity testing of VBNC cells over a range from 10⁴ or 10⁵ to 10⁸ CFU/g in spiked produce samples. “y” represents the average T_t or C_T value, and “x” represents the log CFU per gram of VBNC bacterial cells.

^bFor the E. coli O157:H7-spiked lettuce samples, two of three replicates were positive for 5.13 × 10³ CFU/g. For the S. enterica-spiked tomato samples, one of three replicates was positive for 1.05 × 10⁴ CFU/g.

These results were in agreement with those of previous studies that employed PMA-LAMP assay to detect Salmonella and E. coli O157:H7 in various fresh produce. For example, Wang and coauthors reported that PMA-LAMP could detect 10³ to 10⁴ CFU/g of E. coli in spiked lettuce and spinach (36, 48). In another study, the LOD of PMA-LAMP for Salmonella was 6.1 × 10³ or 6.1 × 10⁴ CFU/g in cantaloupe, spinach, and tomato (60). However, most of those studies targeted determinations of levels of viable cells instead of levels of cells in the VBNC state. The current study utilized PMA-LAMP combined with a plating assay to determine the VBNC pathogens in fresh produce, and the sensitivity was comparable to that achieved by the assays developed for viable cells. Slight (10-fold) variations in the sensitivity of LAMP-based detection of enteric bacterial species in different foods have been ascribed to differences in the inherent properties of food matrices, such as pH, minerals, phenolic compounds, and composition of the background microbiota (60), which could serve as potential inhibitors to interfere with nucleic acid amplification during the analysis of fresh produce (48, 51). For example, low (i.e., 1.1 to 2.9 CFU/25 g) concentrations of Salmonella were detected by LAMP after enrichment of spiked cantaloupe and lettuce samples but not sprouted vegetables, which are known to harbor an abundant and complex microbiota (51). The effect of different fresh produce varieties on the achievable LOD emphasizes the need to evaluate PMA-LAMP-based methods in a commodity-specific manner (61). A similar quantification result was reported by Chen and coworkers, who found that both PMA-LAMP and PMA-qPCR could quantify 10⁴ to 10⁷ CFU/g of viable Salmonella in spiked produce (60). Overall, the developed PMA-LAMP assay is as effective as the well-established PMA-qPCR assay in the detection of VBNC cells in food samples. Previous studies demonstrated that LAMP was even more robust than PCR-based methods regarding tolerance of inhibitors in some clinical and biological samples (61, 62).

In the previous studies, bacteria were usually inoculated in the sample homogenates (36), whereas surface inoculation was used in the current study to simulate the naturally contaminated produce seen under real-world conditions. In addition, genomic DNA was extracted from the targeted bacteria by the use of a simple boiling method, and no purification was required. With loop primers included, the LAMP assay required only 20 to 30 min at 65°C to generate positive results, which represented a significant reduction in detection time compared with either the conventional direct viable count (DVC) method or qPCR assay. Taking the results together, PMA-LAMP enables reliable, robust, and rapid determination of VBNC E. coli O157:H7 and S. enterica counts in spiked fresh produce without complicated sample preparation procedures. The developed assay should be further validated and applied to examine the contamination of VBNC pathogens in real fresh produce.

Conclusion.

We developed a rapid, specific, and sensitive PMA-LAMP assay which, coupled with a plating assay, quantified E. coli O157:H7 and S. enterica cells in the VBNC state in pure culture and in samples of fresh produce. Significant economies of time (i.e., 30 min versus 1 h) were possible in comparison with PMA-qPCR while delivering comparable sensitivity and quantitative capacity. Furthermore, LAMP is an isothermal assay that can be conducted in a simple water bath, making it applicable for use by resource-limited laboratories or for field deployment. Our study validated the performance of the PMA-LAMP assay for the detection of VBNC enteric bacterial pathogens in fresh produce. Given ongoing concerns about potential contamination with E. coli O157:H7 and S. enterica, the proposed approach can provide previously unavailable information to supplement ongoing efforts to enhance the safety of fresh produce.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

E. coli O157:H7 EDL933 and Salmonella enterica serovar Enteritidis 43353 were used as type strains for the development of E. coli O157:H7-specific and S. enterica-specific PMA-LAMP assays, respectively. The specificity testing of PMA-LAMP was conducted using various Gram-positive and Gram-negative bacteria (Table 1). S. enterica, E. coli, Pseudomonas aeruginosa, L. monocytogenes, and Staphylococcus aureus bacterial strains were grown on tryptic soy agar (TSA) (BD Diagnostic Systems, Sparks, MD, USA) at 37°C under aerobic conditions for 16 h. Campylobacter jejuni and C. coli strains was grown at 37°C on Mueller-Hinton agar (BD Diagnostic Systems, Sparks, MD, USA) supplemented with 5% sheep blood under microaerophilic conditions (85% N₂, 10% CO₂, and 5% O₂) for 48 h. All strains were preserved in 20% (vol/vol) glycerol at −80°C.

PMA treatment and DNA extraction.

To optimize the concentration of PMA treatments to differentiate between viable and dead E. coli O157:H7 and S. enterica, viable and dead cells (10⁶ CFU/ml) were separately treated with PMA at a concentration of 0, 1, 5, 10, 25, 50, or 100 μmol/liter before DNA extraction. Viable cells were collected by diluting the exponential-phase culture, and dead cells were prepared by heating viable cells at 100°C for 15 min in a heating block (VWR International, Radnor, PA, USA). PMA solution (Biotium Inc., Hayward, CA, USA) was added to 100 μl of the tested E. coli O157:H7 and S. enterica cells to reach the aforementioned specific concentrations. The mixture was incubated in the dark with constant agitation at 100 rpm for 10 min before being exposed for 15 min to a 300-W halogen light (GE Lighting, General Electric Co., Cleveland, OH, USA) (120 V) that was maintained at a distance of 20 cm from the sample. Following three washes with sterile distilled water, the mixture was boiled at 95°C for 10 min to release genomic DNA (gDNA). An aliquot of 2 μl of DNA template was used for LAMP assay. The optimal working concentration of PMA was determined as the concentration that could fully inhibit DNA amplification of dead cells without interfering with the viability of the remaining cells. PMA treatment at an optimal concentration was used prior to LAMP and qPCR in the following experiments.

LAMP primers and reaction.

O-antigen polymerase synthesis gene wzy and fimbrin-encoding gene agfA were selected as the specific targets for the detection of E. coli O157:H7 and S. enterica, respectively. LAMP primers were designed using the PrimerExplorer program (online version4; Fujitsu Limited, Japan). Each set of LAMP primers contained five to six primers, including two outer primers (F3 and B3), two inner primers (FIP and BIP), and one or two loop primers (LF and LB) targeting specific locations of the corresponding gene (35). The sequences of the primers are provided in Table S1 in the supplemental material.

The LAMP reaction was conducted according to protocols described by the manufacturer (New England Biolabs, Ipswich, MA, USA) with a few modifications. The reaction mixture used for the detection of E. coli O157:H7 was set as 20 μl in total, consisting of 2 μl of 1×Thermopol reaction buffer (New England Biolabs, Ipswich, MA, USA), 1.2 μl of 8 mM MgSO₄, 12 μl of a 1.2 mM concentration (each) of deoxynucleoside triphosphate (dNTP), 1.8 μl of 1.6 μM forward inner primer/backward inner primer (FIP/BIP), 1.6 μM LF/LB, 0.8 μM F3/B3, 1 μl of 0.4 U/μl Bst 2.0 DNA polymerase (New England Biolabs, Ipswich, MA, USA), and 2 μl of DNA template. Reaction mixture (20 μl) for the detection of S. enterica consisted of 2 μl of 1×Thermopol reaction buffer, 1.2 μl of 8 mM MgSO₄, 12 μl of 1.2 mM each dNTP, 1.8 μl of 1.6 μM FIP/BIP, 0.8 μM LF/LB and 0.2 μM F3/B3, 1 μl of 0.4 U/μl Bst 2.0 DNA polymerase, and 2 μl of DNA template. A negative control (2 μl of sterile water) was included in each run of LAMP reaction mixtures. The LAMP reaction was carried out at 65°C for 30 to 60 min and terminated after being maintained at 80°C for 5 min in an LA-320C Loopamp real-time turbidimeter (Eiken Chemical Co., Ltd., Tokyo, Japan). This reaction can produce magnesium pyrophosphate as a white precipitate by-product, which changes the turbidity of the LAMP solution. Turbidity changes were continuously monitored every 6 s, and the T_t value (in minutes) was determined from the time when the increase in turbidity exceeded a threshold value of 0.1.

qPCR assay.

A qPCR assay was also developed for comparison to validate the performance of LAMP assay. The qPCR primers used for targeting the wzy gene for E. coli O157:H7 detection and the agfA gene for S. enterica detection were designed using the Primer-BLAST tool (National Center for Biotechnology Information, USA). The primer sequences are shown in Table S1. The reaction volume was set at 20 μl, and the reaction mixture consisted 10 μl of 1× SensiFAST SYBR Lo-ROX mix (Bioline USA Ltd., MA, USA), 0.16 μl of a 0.4 μM concentration of each reverse and forward primer, 7.84 μl of sterile distilled water, and 2 μl of DNA template. The reaction was carried out in a CFX96 real-time PCR thermocycler (Bio-Rad, Hercules, CA, USA) and included an initial denaturation step at 95°C for 10 min, followed by denaturation at 95°C for 15 s and annealing and extension at 60°C for 1 min for 39 cycles. The melting curve experiment was conducted over a temperature range of 65°C to 95°C with an increase of 0.5°C per 0.05 s. The C_T value (in cycles) was determined when the fluorescence reading exceeded the threshold set by the instrument. Sterile water was used as the negative control in each run of the experiment.

Specificity and sensitivity of PMA-LAMP.

The specificity test was conducted using 31 different bacterial species (listed in Table 1). An overnight bacterial culture of each strain adjusted to an optical density at 600 nm (OD₆₀₀) of 1 (∼10⁹ CFU/ml) was subjected to PMA treatment and DNA extraction as described above. An aliquot (2 μl) of the gDNA was used as the template for LAMP amplification of E. coli O157:H7 and S. enterica. Specificity tests were performed in at least triplicate.

The sensitivity of the PMA-LAMP assay was evaluated in 10-fold serially diluted viable E. coli O157:H7 and S. enterica cells at levels ranging from 10⁸ to 10² CFU/ml. Thermally inactivated dead cells (10⁶ CFU/ml) were added to the reaction mixture to mimic real-world detection against a background of bacterial cells in different states. Serially diluted viable cell and dead cell suspensions were mixed and treated with the optimal concentration of PMA. An aliquot (2 μl) of the template DNA prepared as described above was used in the LAMP assay. Four replicates were performed. The LOD (CFU/reaction in pure culture) was the lowest number of bacterial cells that could be detected by using PMA-LAMP.

Induction of VBNC E. coli O157:H7 and S. enterica and confirmation of VBNC using fluorescence microscopy.

E. coli O157:H7 and S. enterica strains were grown on TSA at 37°C for 16 h. Cells harvested by centrifugation were washed three times with phosphate-buffered saline (PBS). The final cell pellets were suspended in a 7% (wt/vol) NaCl solution to reach a final concentration of 10⁸ CFU/ml and were incubated at 37°C with constant agitation at 180 rpm. An aliquot was collected each day for the measurement of culturable cell counts by plating on TSA and for analysis by PMA-LAMP. VBNC cell counts were determined by subtracting the culturable cell counts from viable cell counts determined by PMA-LAMP. On the basis of the formation of 1 colony on triplicate TSA plates (each containing 1 ml of sample), the LOD of the plating assay was 0.3 CFU/ml. When culturable cell counts decreased to an undetectable level (<0.3 CFU/ml), the remaining viable cells detected by PMA-LAMP were considered to be in the VBNC state. The presence of VBNC cells was further confirmed by fluorescence microscopy following staining using a LIVE/DEAD BacLight bacterial viability kit (Molecular Probes, Eugene, OR, USA), according to protocols provided by the manufacturer. This kit contains two fluorescent dyes, including SYTO 9 and propidium iodine (PI). SYTO 9 stains all viable cells with both intact and compromised membranes, whereas PI stains only dead cells with a damaged membrane. Briefly, 1 ml of bacterial culture was incubated with 3 μl of a mixture of the dyes (1:1 [vol/vol]) in the dark for 30 min. One microliter of the stained bacterial suspension was added to a slide for examination by the use of a fluorescence microscope (IXplore Standard; Olympus, Shinjuku, Tokyo, Japan). Images were collected using an Axiocam camera (Carl Zeiss) at 488 nm for green fluorescent protein (GFP) signal and 543 nm for red fluorescent protein (RFP) signal. Viable cells and dead cells were stained in green and red, respectively.

LAMP assay for the detection of VBNC cells in fresh produce.

Fresh produce (i.e., romaine lettuce, spinach, and cherry tomato) were purchased from local supermarkets in Vancouver, and no culturable E. coli O157:H7 or S. enterica was detected prior to the analysis. Romaine lettuce and spinach were sorted and cut into portions of 10 g each. Cherry tomatoes weighing around 10 g were selected. VBNC cells induced in 7% (wt/vol) NaCl as described above were used to inoculate the samples when no culturable cells were determined using the plating assay (completely induced to the VBNC state). VBNC cells were spiked into fresh produce using a spot inoculation method (63). Briefly, 1 ml of VBNC bacterial cell culture was separately spot inoculated onto the surface of both romaine lettuce and spinach samples to a concentration ranging from 10⁸ to 10³ VBNC cells/g produce. For tomato samples with a smooth surface, the inoculum (100 μl) was applied in small droplets to enable rapid drying and even distribution. Uninoculated samples were included as negative controls. All of the samples were air-dried in a biosafety cabinet for 2 h (48, 60).

Inoculated samples (10 g) were placed in stomacher bags with 90 ml of buffered peptone water (VWR International, Radnor, PA, USA) and agitated on a rotary shaker operating at 100 rpm for 30 min at 22°C to recover bacteria. After recovery, 1 ml of sample wash was spun in a centrifuge at 1,000 × g for 5 min to remove plant and soil particles, followed by centrifugation at 5,000 × g for 10 min to collect bacterial cell pellets. Pellets were resuspended in 100 μl of PrepMan Ultra sample preparation-reagent (Applied Biosystems, Foster City, CA, USA) and used to perform the PMA treatment and LAMP/qPCR as described above. Experiments were conducted in triplicate. The LOD (in CFU per gram in fresh produce sample) was defined as the lowest number of bacterial cells that could be detected using PMA-LAMP and PMA-qPCR assays.

Data analysis.

Both means and standard deviations of T_t and C_T values for LAMP and qPCR were calculated using Excel (Microsoft, Seattle, WA, USA). Standard curves for LAMP and qPCR assays were obtained by plotting the C_T (cycle) or T_t (minute) values against the concentrations of VBNC cells (CFU count/reaction of pure culture or CFU count per gram of fresh produce) of E. coli O157:H7 and S. enterica. Analyses of variance (ANOVA) were performed to test statistical significance using SPSS Statistics (IBM, Armonk, NY, USA) (P < 0.05).