With the increased consumptions of fresh produce, foodborne illnesses related to contaminated produce become a huge food safety concern to public. According to the Center for Disease Control and Prevention (CDC), contaminated produce causes 46% of foodborne illness and 23% of foodborne illness-related deaths  . Salmonellosis is one of the major food-illnesses in the outbreaks of produce. From 1973 to 2018, Salmonella sp. was associated with outbreaks in alfalfa sprouts, melons, apple/orange juices, leafy greens, tomatoes, cucumbers, precut celery and mixed fruits    . On April 30th-July 2nd, 2018, CDC reported multistate outbreaks of Salmonella Adelaide infections linked to pre-cut melon supplied by the Caito Foods, LLC with 77 people infected and 36 people hospitalized  . From harvesting in farms to the dining table, food safety of fresh produce needs to be inspected and monitored, before it can reach the retail stores and ultimately, consumption by individuals.
The current Bacteriological Analytical Manual 8th edition (BAM) from U.S. The Food and Drug Administration (FDA)  for the detection of Salmonella in produce requires several steps before providing results. The FDA standard testing procedures for identification of Salmonella in foods are classified by food types. The standard steps for detection of Salmonella in leafy greens samples include: sample preparation procedures, 1st pre-enrichment step with lactose broth for 24 hours, 2nd pre-enrichment step with Rappaport-Vassiliadis (RV) medium and Tetrathionate (TT) broth for 24 hours, and isolation of Salmonella from selective media such as Xylose lysine desoxycholate (XLD) agar or Lysine iron agar (LIA) for another 24 hours (Figure 1(a)). After the 72 hours of Salmonella identification steps, it takes up to another 3 days to perform other serotyping tests to confirm the Salmonella enterica serotypes in the contaminated samples  . According to the descripted procedures, this standard method is very laborious and time consuming.
Since 2001, FDA started to include rapid methods, such as some antibody based methods or DNA based real-time polymer-chain reaction (RT-PCR) assays for quick detections of foodborne pathogens in BAM  . These methods can serve as alternative procedures to detect Salmonella in the pre-enrichment samples with good sensitivities and fast screening results. However, some of
Figure 1. (a) The illustration of FDA BAM procedures (8th edition) for Salmonella detection in food and the applicability of phage ME biosensors into FDA methods; (b) Prototype of detection device for phage ME biosensors with adjustable working stage, DC magnetic fields, and a network analyzer for data analysis.
those methods also encounter problems in certain produce types. For example, chlorophyll from damaged leaves in the leafy samples will be released in the pre-enrichment steps and inhibit the PCR reaction. This usually results in the decreasing of sensitivity and failure to detect low counts of Salmonella DNA  in leafy samples. As for tomatoes, endogenous polyphenol and a low pH range (pH 4.0 - 4.4) are factors that may affect and interfere with antibody methods and PCR detections, especially when smaller Salmonella populations are present in the samples   . While following FDA guidelines for detecting Salmonella in problematic produce types (such as tomatoes and leafy greens), one only would have choices of new techniques from the laborious BAM pre-enrichment steps or rapid PCR methods which may sacrifice the detection sensitivities. In this case, other rapid, more sensitive, and reliable methods are urgently needed for Salmonella detection on those problematic produce samples  .
Using bacteriophage, instead of antibodies or aptamers, as bio-recognition element on the sensor is a recent developed technique    . The magnetoelastic (ME) biosensors consist of a freestanding striped-shaped ME resonator coated with engineered phage oligonucleotide probes that specifically binds with the pathogens of interest  . When a time-varying magnetic field is applied (Figure 1(b)), the ME biosensors can be placed into mechanical resonance by magnetostriction. Upon the attachment of the target pathogen toward on the phage immobilized sensor, a resonant frequency shift of the biosensor is recorded through a network analyzer and gives a positive detection signal.
The phage ME biosensors are successfully shown to be rapidly detected Salmonella on produce surfaces with Salmonella detection limit at 500 cfu/mm2 on the surfaces of tomatoes, apples, shell eggs, watermelons, and spinach leaves    . These biosensors can also detect the bacteria in the liquid format    and on raw chicken meat  . With the new designed wireless platform, it speeds up the biosensor detection time to less than 2 minutes  . Not to mention, the cost of the phage ME sensor was as low as $0.001 per sensor  . Combining the rapid detection time with the high sensitivity of detecting low Salmonella cells using bacteriophage probes, the ME biosensors is truly applicable to detect Salmonella in FDA-BAM pre-enrichment process and shorten the detection time. In this study, we evaluated the use of the wireless phage ME biosensors by studying the Salmonella detection relative effectiveness in four different media as an outline of 1st pre-enrichment step in FDA’s Salmonella identification procedure. Four media used in this study were lactose broth, modified peptone water (MPW), universal pre-enrichment (UP) broth, and Lennox broth (LB). Lactose broth, MPW, and UP broth are the recommended pre-enrichment media for FDA-BAM and USDA-FSIS protocols   , while LB broth is a common medium used in the most labs for growing Enterobacteriaceae. The shortest incubation time for ME biosensor detection was also investigated. This study is an extension of our previous paper of Wang et al., 2017  . In this paper, two problematic produce, tomatoes and spinach were studies; microbiological data was shown to explain details of suitable pre-enrichment broth selection in the exploratory of four types of media in early detection of Salmonella.
2. Materials and Methods
2.1. Preparation of Salmonella Typhimurium Solution
Salmonella enterica Typhimurium (ATCC 13311) was used in this study. Salmonella Typhimurium was grown from a single colony in Lennox Broth (LB broth) overnight in a shaking incubator at 37˚C, at a speed of 200 rpm. Overnight bacterial cultures were centrifuged at 5500 rpm for 10 min at 4˚C and re-suspended in PBS twice. The bacterial populations were then adjusted to an OD of 600nm, which equates to 1.0 in PBS. Salmonella suspensions were then further diluted to 103 cfu/mL.
2.2. Spiking of Salmonella on Spinach and Tomatoes and Pre-Enrichment in Different Broths
Packages of triple washed spinach leaves were purchased from local supermarkets in Auburn, AL. The leaves, with a total weight of 125 grams, were collected and rinsed with filtered water in a sterilized beaker. Under a biosafety cabinet, the leaves were sprayed with 70% EtOH and air dried under the cabinet. Each batch of spinach leaves (now 25 grams/ batch) was sterilely picked up, placed inside a sterile plastic tote bag, sprayed with a fresh Salmonella Typhimurium suspension (total 100 cfu) with a sterile adjustable sprayer (Spray Anywhere, Fisher Scientific, Pittsburgh, PA.) The bag of leafy greens (25 grams) with the final Salmonella suspension of 4 cfu/g was then mixed manually for 10 min to be homogenized in the biosafety cabinet.
Three different pre-enrichment broths were used in spinach study. They are lactose broth, LB, and MPW. Each broth (50 ml), was added into the bag and mixed well. The sterile bag was folded loosely to ensure oxygen aeration for Salmonella growth. The pre-enrichment solution was held in room temperature for 1 hour to stabilize the pH and then incubated at 37˚C for 5, 7, and 22 hours. At each incubation time, the pre-enrichment solution was centrifuged at 10,000 rpm for 10 minutes twice. The centrifuged pellet was re-suspended to a 1.0 mL final volume using filter sterilized water.
Boxes of Campari tomatoes were also purchased from local supermarkets in Auburn, AL. Two Campari tomatoes were collected as a group. Each group of tomatoes was then weighted, rinsed, and cleaned with 70% EtOH as previously described. Tomatoes were placed in a sterile plastic bag and sprayed with Salmonella Typhimurium suspension at (4 cfu/g) using a sterile adjustable Sprayer. After the spiking of Salmonella on tomatoes, the three pre-enrichment broths were used. They are UP broth, LB broth, and MPW broth. Each broth, at a ratio of 1:2 (tomatoes: broth, w/v), was added into the bag with contaminated tomatoes and mixed well. The sterile bag was folded loosely to ensure oxygen aeration for Salmonella growth, and the rest of pre-enrichment steps were performed the same as in the previous paragraph.
2.3. Preparation of Phage ME Biosensors
Magnetoelastic (ME) biosensors were fabricated by a commercially available Metaglas 2826MB ribbon (Metaglas, Inc). Further detailed descriptions of sensor fabrication, dicing, and final treatment processes are included in Horikawa et al., 2015  . The ME sensors used in this study had a final layer coated with Au (gold), and were cut into a strip shape at the size of 1 mm × 0.2 mm × 0.028 mm. Phages are bound to the gold-coated sensor layer due to physical adsorption. Phage E2 was an fd-tet filamentous phage and was selected through Phage Display method. This phage was used as bio-recognition elements in this study for specific binding to Salmonella Typhimurium in produce samples. The detail bio-panning procedures, sensitivity tests, and specificity of this phage were described in Sorokulova at al., 2005  . Phage sensors were prepared by coating the ME sensor with 1 × 1011 virons of phage solution in TBS for one hour at room temperature on a rotator. The phage coated sensor was then washed three times with TBS. Bovine Serum Albumin (BSA) at 0.1% was used as a blocking reagent for preventing non-specific binding. The phage sensor was put in 0.1% BSA solution for 1 hour at room temperature on a rotator. Sensors coated with 0.1% BSA (without phages) served as negative controls in the frequency measurement experiments.
2.4. Frequency Measurement by Phage ME Biosensors and Statistics
The final 1.0 mL solution from pre-enriched spinach leaves or tomato samples (Section 2.2) was further concentrated down to 330 µL in filtered water by centrifugation. After the phage ME biosensor was incubated with Salmonella suspensions for one hour at room temperatures, each sensor was washed with 1× TBS three times and taken out for frequency measurements and data analysis. The procedures for frequency measurements were the same as described in Wang et al., 2017  . Each batch of inoculated produce was tested by 8 sensors and the experiment was repeated twice. Three non-phage sensors with BSA only served as controls in each group.
Overall, data collected from frequency measurements was analyzed by an on-tailed unpaired student’s t-test with p < 0.05 and α = 0.05 (as described in Wang et al., 2017  ). The degree of dissimilarity between control and measurement sensors was calculated and the shortest incubation time was decided for sensor detection. Confidence level of difference equated to (1 − p value) × 100%.
2.5. Salmonella Numbers on the Sensors and Salmonella Capture Rates in Broth
XLD agar were used in this section of experiments. Frist, the 10 µL of the final 1.0 mL bacterial suspension concentrated from each test group (various media types and incubation times) was diluted and plated on XLD agar. The resulted plates were incubated at 37˚C for 16 - 18 h for total Salmonella counts in pre-enrichment (as Total Salmonella Counts). The rest of final 1mL bacterial suspension was then incubated with a phage coated sensor in a 1.5 mL micro-tube for 1 hour at room temperature with gentle rocking. The sensor was washed three times by 1 × TBS. Salmonella cells on the sensor were eluted by 0.1 M Glycine buffer (pH 2.2) for 10 min and neutralized by adding 1M Tris-HCl (pH 9.1). The neutralized bacterial suspension was diluted with PBS and plated on XLD agar plates for recorded as Salmonella number on sensor. The Salmonella capture rate was calculated by Salmonella capture rate in broth = (Salmonella number on one sensor/Total Salmonella Counts in broth) × 100%.
Each test group had duplicate plates. Each experimental parameter (various broth types and incubation times) was tested by four sensors and the whole experiment was repeated twice.
3. Results and Discussions
According to FDA BAM “Chapter 5-Salmonella”  , different pre-enrichment media should be used for different food types. Lactose broth is the recommended broth to use in the first pre-enrichment step of Salmonella detection leafy samples. However, in the study of Wang et al., 2015  , it was found that some other broths appeared to be more effective in PCR assays than the current BAM suggested broths. It may be because the BAM suggested broths was evaluated to have the maximum results in the culture-dependent methods. Broths recommended by BAM may not provide the best results in other type of assays, such as DNA-based methods or phage-based biosensor detection platforms. Incubation techniques meant to raise the number of target microorganisms above detection thresholds were also addressed in the paper. This is applicable especially in food samples contaminated with low Salmonella counts, which are needed to raise target bacterial numbers for detection. Therefore, there is a need to study a suitable enrichment broth to be used on each detection platform and the minimum incubation time for the detection. The aim of this study was to find a better broth for phage ME biosensors platform to be used in the pre-enrichment condition with the shortest incubation time to detect low Salmonella spiked spinach leaves or tomato samples.
3.1. Frequency Shift Measurements in Artificially Spiked Spinach and Tomatoes Samples
Table 1 shows the data of the frequency shift of biosensors for Salmonella detection on spiked spinach leaves in MPW, Lac broth, and LB broth for 7 and 22 hours pre-enrichment times. In this study, a one-tailed student t test with p value smaller than 0.05 (α = 0.05) was used. The confidence level of differences (CLD) of 95% or more was needed to achieve the significant Salmonella positive signal from the differences of frequency shifts between measurements and control sensors. According to Table 1, after 7 hours of incubation, the value of CLD
Table 1. Frequency measurements of phage ME biosensor platform for Salmonella detection in artificially spiked spinach leaves in MPW, Lac, and LB broths for 7 hours, and 22 hours.
Note: a-frequency shifts (kHz) of measurement sensors; b-frequency shifts (kHz) of control sensors; c-confident level of difference (%).
of both MPW and Lac broths did not reach or exceed 95% and was lower than the CLD of 96.76% in LB broth. This data showed that LB broth is the only broth among the three to effectively yield Salmonella positive signals in artificially spiked spinach at a 7 hours incubation period. After 22 hours of incubation, all three broths produced significant Salmonella positive signals in frequency shifts with high CLD values exceed 95% (Table 1).
As for tomatoes samples, the detail study of resonant frequency changes in three broths (MPW, UP, and LB) after 5, 7, and 22 hours of pre-enrichment are presented in Table 2. By analyzing the data with a CLD above 95%, tomato samples in LB broth demonstrated significant differences in frequency shift of measurement sensors in compared to controls as early as 5 hours of incubation time. However, tomato samples in MPW and UP after 5 hours and 7 hours pre-enrichment did not show any significant frequency differences between measurements and control sensors. After an incubation time of 22 hours, significant frequency differences were observed in all three broths.
In order for easy understanding, CLD values, measured by frequency shift (Table 1 and Table 2) and passed 95% significant differences between the measurement and control sensors in various broths, are expressed as Salmonella positive signals. Otherwise, the CLD values lower than 95% were categorized as Salmonella negative signals. Table 3 and Table 4 are the summary tables of Salmonella detection signals of artificially spiked spinach leaves tomatoes in four different broths for 7 and 22 hours of incubation using phage ME biosensor platform. This summary data clearly demonstrated that LB is the best pre-enrichment medium to be use for early detection of 5-7 hours incubation for detection low Salmonella contaminations in both tomato and spinach samples.
Table 2. Frequency measurements of phage ME biosensor platform for Salmonella detection in artificially spiked tomatoes in MPW, UP, and LB broths for 5 hours, 7 hours, and 22 hours of pre-enrichment.
Note: a-frequency shifts (kHz) of measurement sensors; b-frequency shifts (kHz) of control sensors; c-confident level of difference (%).
Table 3. Salmonella detection signals of artificially spiked spinach leaves in MPW, Lac, and LB broths for 7 and 22 hours of incubation using phage ME biosensor platform.
Note: Negative Salmonella detection signal (-) represented CLD is under 95%; positive signal (+) represented CLD equals or larger than 95%. Eight sensors were used for each test condition and the experiment was repeated twice.
Table 4. Salmonella detection signals of artificially spiked tomatoes in MPW, UP, and LB broth for 5, 7, and 22 hours incubation using phage ME biosensor platform.
1Note: Negative Salmonella detection signal (−) represented CLD is under 95%; positive signal (+) represented CLD equals or larger than 95%. Eight sensors were used for each test condition and the experiment was repeated twice.
3.2. Salmonella Numbers on the Phage ME Biosensor in Contaminated Spinach and Tomato Samples
In order to understand the detail performances of phage biosensors in each broth, the number of Salmonella on the sensor in all three broths for 7 and 22 hour incubation times were evaluated. The data are shown in Figure 2 for spinach samples and Figure 3 for tomato samples. According to the report of Li et al., 2010  , the Salmonella detection limit of phage ME biosensor was 500 cfu/mm2 as determined by the direct detection of Salmonella on the tomatoes surfaces. Their study of phage biosensor direct detection of Salmonella was conducted by air-dried Salmonella cells on the surface of tomatoes. Therefore, the detection unit was expressed as cfu/mm2. In this study, the same Salmonella amount was used for set up the detection threshold of Salmonella. Since this study was performed in the liquid detection format, the unit was expressed as cfu/sensor. After a 7 hour incubation period, the phage biosensor was able to capture 937 cfu/sensor of Salmonella with spinach samples in LB, which passed the detection threshold of Salmonella counts (present as a red line in Figure 2). In MPW and Lac broth enrichment, Salmonella numbers did not exceed 500 cfu/sensor. Therefore, these two broths failed to pass the detection threshold and also didn’t show Salmonella positive signals in frequency tests. At 22 hours incubation, all three broths were able to demonstrate a high Salmonella number of 104 - 105 cfu/sensor (Figure 2), which triggered the Salmonella positive signal in biosensor frequency detections with high CLD values. With the early detection of 7 hours incubation time in the 1st pre-enrichment step of FDA procedures, LB appeared to be the best among the three media for spinach samples to use in Salmonella detection by ME biosensors.
Figure 2. Salmonella numbers on the sensor in three broths after 7 h and 22 h incubation of artificially spiked spinach leaves at low Salmonella population (4 cfu/ gram). The red line represents the Salmonella detection threshold (500 cfu/sensor) of the phage ME biosensor detection platform.
Figure 3. Salmonella numbers on the sensor in MPW, UP, and LB broth after 5 h, 7 h, and 22 h incubation of artificially spiked spinach leaves at low Salmonella population (4 cfu/ gram). The red line represents the Salmonella detection threshold (500 cfu/sensor) of the phage ME biosensor detection platform.
The microbiological analysis of contaminated tomato samples in MPW, UP, and LB broths after 5, 7, and 22 hours of pre-enrichment time are shown in Figure 3. According to Figure 3, LB broth was the only broth in the three, which is able to capture enough Salmonella cells (719 cfu/sensor) and passed the biosensor detection threshold in the fifth hour of the early incubation stage. MPW and UP did not show enough Salmonella numbers on the sensor until 22 hours of incubation time. This data may indicate that MPW and UP didn’t promote Salmonella growth in tomato samples after short hours of incubation. As the same concept mentioned previously in the study of Wang et al, in 2015  , MPW and UP may not be the suitable broth for quick incubation of phage ME biosensors for detecting Salmonella in tomatoes.
3.3. Salmonella Capture Rate Study
When cross referencing the microbiological data of Salmonella numbers on the sensors and the data of frequency shifts, it yields interesting findings. In the LB broth with both 7 and 22 hours of incubation time, the CLD values in tomatoes (Table 3) were higher than the values in spinach samples (Table 1). However, the Salmonella number on the sensors was lower in tomatoes (Figure 3) than the number in spinach (Figure 2). The frequency measurements and microbiological data of LB in spinach and tomatoes samples seemed to contradictory to each other. Therefore, it is curious to know whether the affinity actions of the phages to Salmonella was still performed normally in the same broths, but different produce types. In the report of Qiang et al., in 2017  , proteins and components in different blocking buffers would inhibit the affinity of phage probes to its target and promote the non-specific binding in ELISA assays. To answer this question, Salmonella capture rates of the phage biosensors were studies. The Salmonella capture rate was calculated by the percentages of Salmonella numbers captured by phages on the sensor and divided by the total Salmonella growth in each test broth with 7 hours of incubation period. The data truly revealed the fruit types played as a factor to affect the phage affinity to capture target pathogens in LB. Figure 4 shows that in 7 hours of pre-enrichment of LB, 9.606% of Salmonella was captured by phages on the biosensors in tomatoes samples, while the capture rate was only 8.57% in spinach. Apparently, substances in spinach samples (such as chlorophyll or flavonoids) may slightly inhibit phage affinity to Salmonella. More studies are needed to seek a conclusive statement about this finding. Besides, in this set of data, LB again demonstrated the highest capture rates among all broths, while MPW showed the lowest Salmonella capture ability for phage biosensors to detect Salmonella in two tested produce samples.
Figure 4. Salmonella captured percentages on the sensor in MPW, Lac, UP, and LB broth after 7 h incubation of artificially spiked tomatoes and spinach leaves at low Salmonella population (4 cfu/ gram).
Overall, the phage ME biosensors were able to detect Salmonella in the LB broth within 5 hours of incubation in tomatoes samples and within 7 hours in low Salmonella spiked spinach samples. MPW may not be a suitable short time pre-enrichment medium for detecting Salmonella in both produce samples using phage ME biosensor. The comparison of the phage ME biosensors to FDA BAM culture method and qRT-PCR method are summarized in Table 5. The detection limit of Salmonella by phage ME sensor method may not be as low as two other methods, but the pre-enrichment time is tremendously decreased. This may be compensated by extending the pre-enrichment time less than one hour to reach the same detection limit. The cost of the ME sensor method is also much lower than the two other methods. Our data also demonstrated that it is possible to implant phage ME biosensors into FDA BAM methods for screening and shortening the Salmonella pre-enrichment times and then perform the BAM culture or PCR methods.
Phage ME biosensors have been demonstrated as a powerful and rapid Salmonella detection platform in contaminated fresh produce and liquid systems. In this study, the phage ME biosensor showed to have a great potential application as an early detection method in FDA BAM pre-enrichment procedures for Salmonella detection in the problematic produce. For Salmonella detection in spinach leaves and whole tomatoes, the proposed phage biosensor platform was able to reduce the detection time from 72 hours to 5 - 7 hours. FDA recommended pre-enrichment broths for detecting Salmonella in spinach and tomato samples didn’t demonstrate the maximum results in phage ME biosensor detection, as compared to the LB broth. By using LB broth as an alternative pre-enrichment medium along with the phage ME biosensor method, the detection time can be reduced to as short as 5 hours in tomatoes samples and 7 hours in spinach samples.
Table 5. Comparisons of phage-Me biosensors with LB to two FDA-BAM methods for Salmonella detection in produce.
This work was supported by the Specially Crop Research Initiative Competitive Grant No. USDA-2011-51181-30642A and Auburn University Detection and Food Safety Center (AUDFS). We would like to thank Eric Ching-Feng Au and Harrison Michael for their assistance in this study.