Advertisement

Analytical and Bioanalytical Chemistry

, Volume 410, Issue 22, pp 5455–5464 | Cite as

Label-free screening of foodborne Salmonella using surface plasmon resonance imaging

  • Jing Chen
  • Bosoon Park
Research Paper
Part of the following topical collections:
  1. Food Safety Analysis

Abstract

It is estimated that 95% of the foodborne infections are caused by 15 major pathogens. Therefore, rapid and effective multiplex screening techniques for these pathogens with improved efficiencies could benefit public health at lower costs. Surface plasmon resonance imaging (SPRi) provides a label-free, multiplex analytical platform for pathogen screening. In this study, we have developed a singleplex immunoassay for Salmonella to evaluate the potential of SPRi in pathogen detection. Anti-Salmonella and control ligands were arrayed onto the SPRi sensor chip in a microarray format. The influences of ligand immobilization pH and concentration were optimized, and a pause flow protocol was adopted to improve assay rapidity and sensitivity. The method shows good specificity against 6 non-Salmonella species and was able to detect 5 of 6 Salmonella serotypes, including 3 serotypes most frequently associated with outbreaks. Limits of detection were found to be 2.1 × 106 CFU/mL in phosphate-buffered saline and 7.6 × 106 CFU/mL in the presence of chicken rinse matrix with 8.9 × 107 CFU/mL of indigenous microflora. The condition of antibody array regeneration was optimized for sequential sample injections. Finally, the SPRi immunoassay was used to detect Salmonella directly from artificially spiked chicken carcass rinse samples. As low as 6.8 CFU/mL of Salmonella could be detected after overnight enrichment in buffered peptone water, demonstrating the potential in streamlined pathogen screening with minimal sample preparation and without detection labels.

Graphical abstract

Keywords

Surface plasmon resonance imaging Salmonella Foodborne pathogen Label-free detection Chicken rinsate Food safety 

Introduction

It is estimated by the Centers for Disease Control and Prevention (CDC) that approximately 48 million people in the USA become ill annually from a foodborne pathogen infection, resulting in 128,000 hospitalizations and 3000 deaths. Over 95% of infections are caused by only 15 pathogens, including Salmonella, Campylobacter, Listeria, and Shiga toxin-producing Escherichia coli (STEC) [1, 2]. Salmonella alone causes more a million cases each year, 200,000 of which can be attributed to poultry and poultry products [3]. The average national cost of foodborne illness was estimated between $55.5 billion and $93.2 billion [4]. To reduce the healthcare and economical burdens associated with foodborne infections, real-time and deployable microbial detection and source identification becomes crucial. Conventional microbiological methods usually require sample pre-enrichment, selective enrichment, and confirmation, and rapid methods such as polymerase chain reaction (PCR) and enzyme-linked immunosorbent assay (ELISA) have been aimed at eliminating or shortening such lengthy steps. Still, inhibitors from complicated food matrices and the need for DNA extraction remain major drawbacks of PCR based methods, and ELISA requires manual application of reagents in multiple steps, with often limited dynamic range. Thus, continued efforts in developing faster, more reliable, and more cost-effective methods are desired to complement the drawbacks of existing techniques.

On the other hand, multiple pathogens are frequently found to co-exist in the same food sample [5, 6]. In current food safety monitoring systems, pathogens are detected using independent protocols, each taking several days to complete [7]. This approach can be inefficient and costly for investigating more than one pathogen. Therefore, a higher multiplex capacity with attainable high-throughput and process automation are desired for future screening methods [8]. Current multiplex rapid assays mainly involve multiplex PCR and microarray. Most of these assays use fluorescence detection, whose multiplexing capacity is limited to 3–4 genomic fragments by the availability of fluorophores that can be separated in the output spectrum [9]. Simultaneous detection of multiple hazardous agents remains a challenge in food pathogen detection.

Surface plasmon resonance (SPR) is a label-free optical technique based on the SPR phenomenon occurring at the interface of metallic thin films and a dielectric (e.g., buffer solutions). Biomolecular events at the interface change the refractive index of the surrounding medium and result in a shift in the SPR angle and intensity changes in reflected light [10]. SPR is a versatile technique capable of monitoring a wide array of interactions, including nucleic acid hybridization, antibody-epitope binding, protein-carbohydrate, protein-protein, and protein-DNA interactions. Its applications have been reported in clinical diagnosis, environmental monitoring, and food safety detection [11]. In the realm of microbial food safety, SPR assays were developed for pathogen cells [12, 13], toxins [14], as well as pathogen DNA and PCR amplicons [15, 16], using immobilized antibody, aptamer, phage, and oligonucleotide as recognition elements [12, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26]. Compared to molecular methods and traditional immunoassays, SPR offers a less complicated process suitable for on-site detection. Avoidance of detection labels further improves target binding and reduces cost.

Despite its many advantages, SPR has little applications in real-world pathogen detection especially in foods, due to the lack of efficiency in traditional single channel SPR setup. Multi-channel SPR instruments which allow for monitoring of several interactions at a time have been reported [26]. More recently, SPR imaging (SPRi) emerged as a new platform for multiplex detection and high-throughput ligand screening. In SPRi, multiple ligands may be immobilized onto the same surface in a microarray format. A CCD detector captures two-dimensional intensity contrast images of the refractive index distribution on the surface; thus, hundreds of interactions can be monitored simultaneously. Flexible detection schemes (hybridization of genomic or fragmented DNAs or affinity-based interactions with whole cells or toxins and virulence factors) allow for simultaneous screening of various types of foodborne hazards (such as coexisting pathogen cells and secreted toxins, as well as the genetic markers), which may be impossible using purely molecular or immunological methods such as PCR or ELISA. Since its advent, the SPRi technique has been reported in analysis of pathogen DNA [27, 28] and pathogenic cells in buffers [29, 30]. A recent study also reported detection of Shiga toxin-producing E. coli in ground beef [31]. However, in most whole cell immunoassays each sample was typically incubated in the SPRi flow cell for prolonged periods of time (5–16 h), which compromised the rapidity of the tests.

The first step towards a multi-target SPRi detection is to test the feasibility of using SPRi as an analytical technique for pathogen cell detection in food. Therefore, the objective of this study was to develop a singleplex assay for foodborne Salmonella. In this paper, the effects of ligand immobilization conditions on the SPRi sensor chip and SPRi flow conditions were studied, the specificity and limits of detection (LODs) were assessed, and the method was used for direct detection in chicken rinse samples with minimal sample preparation.

Materials and methods

Materials

SPRi Biochips™ were purchased from Horiba Scientific (Edison, NJ, USA). Mercaptoundecanoic acid (MUA), 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC), N-hydroxysuccinimide (NHS), skim milk powder, and Trizma® hydrochloride were obtained from Sigma-Aldrich (St. Louis, MO, USA). Buffered peptone water (BPW), phosphate-buffered saline (10×), anti-Salmonella antibody (PA1-85849), and mouse IgG control were purchased from ThermoFisher Scientific (Waltham, MA; USA). Anti-E. coli O/K (ab31499) was obtained from Abcam (Cambridge, MA, USA). Tryptic soy broth (TSB), tryptic soy agar (TSA), and brilliant green sulfa (BGS) agar were from BD (Franklin Lakes, NJ, USA). Raw chicken breast was obtained from a local supermarket.

Bacterial strains

Salmonella serotypes S. Enteritidis, S. Kentucky, S. Infantis, S. Javiana, S. Heidelberg, S. Typhimurium, E. coli, Enterococcus faecalis, Listeria monocytogenes, Listeria innocua, Citrobacter koseri, and Klebsiella oxytoca were chicken carcass rinsate isolates and have been stored at − 80 °C long term and 4 °C short term. Before each experiment, a loopful of bacterial colonies maintained on TSA slants were inoculated into 10 mL of TSB medium and incubated at 37 °C overnight. The bacterial culture was harvested, rinsed once with PBS, and serially diluted in PBS or chicken rinse overnight culture (see below) for SPRi injections. Cells were enumerated by plating onto TSA plates.

Ligand immobilization

The antibody and control ligands were immobilized onto the sensor surface through conventional EDC/NHS mediated cross-linking [32]. Briefly, new SPRi Biochips™ consisting of a 50-nm gold layer on high refractive index prisms were soaked in a 2-mM MUA ethanol solution overnight at room temperature (RT) to form a self-assembled monolayer of MUA on gold, followed by rinsing with ethanol and deionized (DI) water. The chip was then soaked in EDC/NHS solution (300 mM/300 mM in pH 4.6 sodium acetate buffer) for 20 min at RT to activate the carboxyl groups. After rinsing with water and drying with nitrogen, antibodies and IgG control at 10–1000 μg/mL in spotting buffer (sodium acetate pH 4.6, 5.6 or PBS pH 7.4 containing 30% of glycerol) were spotted onto the sensor chip using a XactII™ compact microarray spotter (LabNEXT, West New York, NJ) with humidity controlled at > 75%. The spotting volume was 5 nL/spot, and a maximum of 64 ligands could be spotted within the sensing area at a spotting diameter of ~ 500 μm. After spotting, the sensor chip was incubated at RT for 3 h in a Petri dish containing a dampen paper towel. Subsequently, excess ligands were rinsed off with copious amount of DI water and dried with nitrogen for continued modification in the SPRi instrument.

Surface plasmon resonance imaging

The antibody-modified sensor chip was immediately loaded into the SPRi system equipped with a 810-nm LED laser, a CCD camera, and a 11–12-μL flow cell accessed via a 6-way valve (Openplex, Horiba Scientific, Edison, NJ). After each ligand spot was defined, 200 μL of 5% skim milk solution in PBS was injected at a flow rate of 10 μL/min to block the sensor surface against non-specific binding. Then 20 mM PBS and Tris-Cl buffer (pH 7.4) were injected at 100 and 20 μL/min, respectively, to remove excess skim milk proteins and deactivate the unreacted carboxyl groups. Finally, the reflectivity variations across the sensor surface were calibrated by injecting 200 μL of 20 mM PBS buffer at 50 μL/min.

At the beginning of kinetic measurements, the running buffer (PBS) was injected 3–4 times to stabilize the signal. The influence of injection flow rate on SPR signal intensity was tested by injecting samples at 5, 10, 25, 50, and 100 μL/min. Alternatively, pause flow injections were carried out by injecting samples at 50 or 100 μL/min until half of the injected volume passed the flow cell, and then pausing the flow for > 25 min before resuming the flow at initial rates. Real-time SPR signal (percent reflectivity variations relative to injection starting point), flow cell images, and difference images (i.e., real-time flow cell images subtracted by reference flow cell images at the beginning of each injection) were acquired using the SPRi View software (Horiba).

Regeneration and reuse of the sensor chip

Between injections, the bound target cells and non-specific binding were removed by injecting regeneration solutions at high flow rate (100 μL/min). The effectiveness of regeneration was assessed using 10–100 mM NaOH (pH 12–13), glycine-HCl (pH 1.5–2.5), or high salt solutions (1 M NaCl, 2–4 M MgCl2). After each use, the organic layer on the sensor chip was removed by soaking in a stabilized Piranha solution (Nano-Strip®, Cyantek, Fremont, CA) at 50 °C for 2 h, ultra-sonication in DI water for 10 min, and exposure to air plasma (45 W) (Plasma Etch, Carson City, NV) for 5 min. The functionalized sensor chips could be stored in PBS at 4 °C for up to 1 week before being used in SPRi.

Detection of Salmonella in chicken rinse

Chicken rinse samples were collected using a modified procedure based on the Microbiology Laboratory Guidebook MLG 4.09 [7]. Briefly, 25 ± 2.5 g of chicken breast was placed in a sterile bag and mixed with 225 mL of BPW and hand massaged for 1–2 min until clumps were dispersed. The liquid portion of the sample was transferred to a sterile container for subsequent use. In the meantime, the samples were confirmed to be free of Salmonella by plating on BGS agar. To establish the calibration curve of Salmonella in chicken rinse matrix, the collected rinse liquid in BPW was incubated at 37 °C for 16 h to yield an overnight culture of the indigenous microflora in the initial rinse matrix. Different concentrations (0, 5.14 × 104, 5.14 × 105, 5.14 × 106, 2.57 × 107, 3.85 × 107, 5.14 × 107, and 5.14 × 108 CFU/mL) of S. Typhimurium were spiked into this culture before SPRi measurements. To mimic using the SPRi assay in real-world detection, artificially contaminated chicken rinse samples were tested. In brief, 10 mL of chicken rinse samples in BPW were inoculated with 6.8–6.8 × 107 CFU/mL of S. Typhimurium and incubated at 37 °C statically for 16 h. Chicken rinse overnight culture containing the indigenous microflora but without inoculation of Salmonella was used as a blank control. Two hundred microliters of spiked or inoculated sample were directly injected into the SPRi system without further treatment.

Data analysis

The SPRi sensorgrams were analyzed using the SPRi Analysis software (Horiba, Edison, NJ) and Microsoft Excel (Microsoft, Seattle, WA). In this process, signal from the ligand spot replicates were first averaged and then superposed at the beginning of each injection. Finally, signal from the antibody spots was subtracted by that from the IgG control spots at equivalent concentrations. Reflectivity variations (ΔR) at the end of each injection were reported as the SPR response. Unless otherwise mentioned, the sensorgrams are the difference curves between the antibody and control IgG curves. Limit of detection (LOD) is defined as the concentration of Salmonella at which the SPR signal equals to the LOD threshold (i.e., 3 times standard deviation of blank measurements). Lowest calibrated level (LCL) is the lowest tested concentration of Salmonella which yielded a response higher than the LOD threshold. Sensorgrams were visualized using Origin 2015 (OriginLab, Northampton, MA), and the SPRi difference images were subject to the autocorrect once function using Microsoft Office Picture Manager (Microsoft, Seattle, WA).

Results and discussion

Influence of antibody immobilization pH and concentration

Antibodies immobilization was carried out at three pH values (4.6, 5.6, and 7.4) and varied concentrations (10–1000 μg/mL). According to the flow cell images after antibody immobilization (see Electronic Supplementary Material (ESM) Fig. S1a inset (1)), as pH decreased from 7.4 to 4.6, the surface density of immobilized antibody increased, shown as increased contrast between the spots and the surrounding areas. This is consistent with the well-established association between the EDC-mediated coupling efficiency and ligand immobilization pH [33]. Moreover, exposure to low pH buffers during immobilization (3 h) did not significantly impact the antibody activity, as the SPRi intensity and contrast of the difference image after Salmonella injection were both highest when the ligands were immobilized at pH 4.6 (ESM Fig. S1a, Fig. S1a inset (2)).

Similarly, a positive association was found between anti-Salmonella concentration during immobilization and the resultant surface density (ESM, Fig. S1b inset (1)) and SPRi signal intensity (ESM, Fig. S1b and Fig. S1b inset (2)). Below 250 μg/mL, no positive signal could be obtained after injecting 108 CFU/mL of Salmonella, and the highest intensity was found at 1000 μg/mL. When antibody concentration was above 1000 μg/mL, the spots tended to cause significant smears in adjacent areas when excess ligands were rinsed off after conjugation, due to continued reaction between high concentrations of proteins with activated carboxyl functionalized surfaces. Therefore, to avoid cross-contamination of antibodies on the ligand array, a maximum ligand concentration of 1000 μg/mL was immobilized, and all subsequent experiments were carried out with anti-Salmonella and controls immobilized at 1000 μg/mL and pH 4.6.

Detection of Salmonella in PBS buffer

Specificity

Specificity of the SPRi assay was evaluated using six Salmonella serotypes (S. Enteritidis, S. Heidelberg, S. Infantis, S. Javiana, S. Kentucky, and S. Typhimurium) and six non-Salmonella species (E. coli, E. faecalis, C. koseri, K. oxytoca, L. innocua, and L. monocytogenes) at ~ 1.5 × 108 CFU/mL. Figure 1 shows the SPRi sensorgrams (Fig. 1b) and difference images (Fig. 1a) after each injection. Final SPRi signals (Fig. 1c) suggest good specificity of the assay against non-Salmonella species. It was noticed that at a similar concentration (~ 1.5 × 108 CFU/mL), L. monocytogenes tended to induce a much higher reflectivity compared to other species, resulting in a negative response after the IgG control signal was subtracted from the anti-Salmonella response. This was possibly caused by the capacity of L. monocytogenes to rapidly and persistently adhere to contact surfaces [34, 35], which led to prominent non-specific adsorption onto the sensor surface. On the other hand, preferential affinity to Salmonella species provided the anti-Salmonella antibody with greater degree of immunity against non-specific adsorption of other organisms, and in the presence of a high background inducing sample, such as L. monocytogenes, the anti-Salmonella spots appeared less bright in the difference images than the surrounding. Except S. Infantis, all Salmonella serotypes yielded positive results. Since the immunogen of the polyclonal antibody was a mixture of S. Enteritidis, S. Heidelberg, and S. Typhimurium, it is not surprising that the SPRi signals for these three Salmonella serotypes were generally stronger than other serotypes, and some of the other Salmonella strains reacted weakly to the antibody.
Fig. 1

Assay specificity. a Difference images. b Average SPR sensorgrams (n = 3). c Final SPR intensity obtained after injections of E. coli (E. c), E. faecalis (E. f), C. koseri (C. k), K. oxytoca (K. o), L. innocua (L. i), L. monocytogenes (L. m), S. Enteritidis (S. E), S. Heidelberg (S. H), S. Infantis (S. I), S. Javiana (S. J), S. Kentucky (S. K), and S. Typhimurium (S. T). Anti-Salmonella polyclonal antibody was immobilized at 1000 μg/mL. All samples were injected at ~ 1.5 × 108 CFU/mL in PBS. ΔR: real-time reflectivity variations as a function of time after each injection. Final ΔR: reflectivity variation at the end of each injection (t = 16 min)

Influence of flow rate on assay sensitivity

Our preliminary results suggest that the strong binding between Salmonella and the antibody led to absence of an apparent dissociation phase hours after sample injection had completed (data not shown). Thus, each injection could be terminated arbitrarily taken into consideration the total assay time and desired sensitivity. In this study, the analysis cut-off time for each injection was set to twice the time for the injected volume to pass the flow cell plus the inactive time for injected samples to reach the flow cell, or t analysis = t inactive + 2 × t sample. This allowed loosely adsorbed materials to be removed from the sensor surface by the running buffer before specific signals could be obtained. The influence of flow rate on SPR was assessed by injecting samples at different constant flow rates. As shown in Fig. 2a, the final SPR signal more than tripled as the flow rate reduced from 100 to 5 μL/min. Meanwhile, the LCL decreased from 108 to 106 CFU/mL (Table 1), indicating an inverse correlation between flow rate and detection sensitivity. However, since both t inactive and t sample were determined by flow rate, slow flows led to much longer overall assay time despite lower detection limits. The overall analysis time at 5 μL/min was 90 min, including 10 min of inactive time and 45 min of post-injection rinsing time, while a total of only 5 min was needed for injections conducted at 100 μL/min (Table 1). Clearly, a balance between assay rapidity and detection limit must be achieved for optimal detection efficiency.
Fig. 2

a Final SPR intensity obtained using various flow rates. b Average SPR sensorgrams of 108 and 109 CFU/mL of S. Typhimurium using a constant flow rate of 100 μL/min (n = 3). Anti-Salmonella polyclonal antibody was immobilized at 250–1000 μg/mL. ΔR: real-time reflectivity variations as a function of time after each injection. Final ΔR: reflectivity variation at the end of each injection (t = 4.5, 9, 18, 45, and 90 min for 100, 50, 25, 10, and 5 μL/min flow rates, respectively)

Table 1

Inactive time (t inactive), sample interaction time (t sample), and total analysis time (t analysis) of continuous flow SPRi assays and corresponding lowest calibration levels (LCLs) at different flow rates

Flow rate (μL/min)

t inactive (min)

t sample (min)

t analysis (min)

LCL (CFU/mL)

5

10

40

90

106

10

5

20

45

106

25

2

8

18

107

50

1

4

9

108

100

0.5

2

4.5

108

On the other hand, it is noticed that SPR signal continued to increase after sample injections had completed (Fig. 2b). This effect was more pronounced under fast flow conditions, in which the continued signal increase did not end before the pre-determined analysis cut-off time. Regardless of the flow rate, 15–20 min was needed for the interaction between Salmonella and its antibody to reach equilibrium. This could be attributed to slower diffusion of large Salmonella cells to the sensor surface compared to smaller targets such as proteins and small molecules, which resulted in an apparent lag between sample and response. In addition, due to cell’s ability to re-distribute its surface antigens across the outer membrane [36], more surface antigens might have migrated to the sensor interface after initial cell attachment, leading to new binding events. Thus, the movement of antigens on cell surfaces might be another source of continued increase in SPR response. The affinity behaviors of cellular targets are distinctive from small, monovalent molecular targets, which also manifested as “grainier” ligand spots even in high signal-to-noise ratio difference images, as compared to molecular binding such as DNA hybridization. Therefore, fast flows (analysis time < 5 min) should be avoided in cell target analyses to minimize measurement error and to minimize false-negative results in low concentration samples.

To address the unnecessarily long inactive time at slow flow rates and lack of sensitivity at fast flow rates, a pause flow protocol was adopted. At first, the sample was injected at fast flow rates (50 or 100 μL/min). The flow was paused when half of the injected volume reached the interaction cell. After a pre-determined static time (up to 25 min), the flow was resumed at the initial rate. SPRi signal was recorded during the pause period and after the entire sample volume had passed through the flow cell. This way, sample entry and exit time was minimized, and the bacterial cells were given sufficient time to interact with the antibodies before being flushed out of the system. Indeed, SPRi signal intensity significantly increased compared to continuous flow protocols within similar time (30 min in Fig. 3a). Figure 3b shows the calibration curve of S. Typhimurium detection in PBS using final SPRi intensities at t = 28 min, and the LOD was 2.1 × 106 CFU/mL. In fact, ~ 20 min of total assay time was sufficient using the pause flow protocol, since no further reduction in the LODs was observed with prolonged interaction (Table 2). However, the LODs tended to increase when the assay time was shortened to 15 min or less, which was expected considering previously observed signal lag at fast flow rates. Thus, pause flow protocols were employed in all subsequent tests.
Fig. 3

a SPR sensorgrams of 109 CFU/mL S. Typhimurium using the stop flow and continuous flow protocols. In the stop flow protocol, flow rate was sequentially set to 3 min at 50 μL/min 25 min at 0 μL/min, and 3 min at 50 μL/min. b Calibration curve of S. Typhimurium using a stop flow protocol (1.5 min at 100 μL/min, 25 min at 0 μL/min, 1.5 min at 100 μL/min). Signal (Final ΔR) was collected at t = 28 min. Anti-Salmonella polyclonal antibody was immobilized at 1000 μg/mL

Table 2

Limits of detection (LODs) obtained at different analysis times under pause flow conditions

Analysis time (min)

5

10

15

20

25

28

LOD (× 106 CFU/mL)

9.2

8.6

4.2

2.3

1.9

2.1

Detection of Salmonella in chicken rinse

LODs in chicken rinse overnight culture

To assess detection sensitivity in the presence of matrix components and overwhelming background microflora, chicken rinse overnight culture in BPW was spiked with different levels of Salmonella and was directly injected into the SPRi system. Due to the difference in refractive index between the running buffer (PBS) and the sample fluid, which mainly consisted of BPW and chicken meat juices, a dramatic increase in reflectivity was observed once the sample was injected. This increase was best reflected by the raw reflectivity changes before IgG correction (Fig. 4a). In the difference image (Fig. 4a inset), the binding signal from the anti-Salmonella spots was masked by a bright background. The signals of positive and negative responses were best distinguished by the naked eye upon the completion of sample injection, when the refractive index returned to baseline level (PBS), non-specific materials and organisms were rinsed off the view, and the specific signals from S. Typhimurium were highlighted (Fig. 4b inset). Nevertheless, real-time binding signal could be accurately determined based upon corresponding sensorgrams of each ligand family. Figure 4b shows the IgG-corrected sensorgrams, which demonstrate a steady increase of reflectivity after the flow was stopped and until equilibrium was reached. When the flow was resumed and sample matrix was removed, a second significant increase from the anti-Salmonella spots was observed due to a greater decrease of the IgG control signal (the target signal was a differential signal between the raw reflectivity at the anti-Salmonella spots and the IgG control spots).
Fig. 4

a Raw SPR sensorgrams of a Salmonella injection in chicken rinse matrix. Inset: SPRi difference image at t = 15 min (S: anti-Salmonella spots; E: anti-E. coli spots; I: IgG control spots). b Processed SPR sensorgrams of the same injection. The raw sensorgrams at antibody spots are subtracted by the raw sensorgram at the IgG spots. Inset: SPRi difference image at t = 28 min. Samples were injected from top of the antibody array and exited from the bottom of the array

The chicken rinse samples were confirmed to be free of Salmonella by plating on BGS agar. After overnight enrichment at 37 °C, the indigenous microflora in the chicken rinse sample yielded a total aerobic count of 8.9 × 107 CFU/mL. Then varied concentrations of S. Typhimurium were spiked into this chicken rinse overnight culture to obtain serial dilutions of Salmonella (final Salmonella concentrations = 0, 5.14 × 104, 5.14 × 105, 5.14 × 106, 2.57 × 107, 3.85 × 107, 5.14 × 107, and 5.14 × 108 CFU/mL) in the presence of matrix residues and an overwhelming population of competitive bacteria. In the chicken rinse matrix, non-specific adsorption of residual protein and fatty tissues as well as a large amount of background flora were expected to decrease the binding efficiency between Salmonella and the anti-Salmonella antibody. More significantly, such non-specifically adsorbed materials could lead to increased background signal from the IgG control and result in reduced intensity once anti-Salmonella signal was subtracted by the IgG control. This translates into a higher detection limit and faster signal saturation at high target concentrations. Indeed, the SPRi difference images indicate pronounced non-specific adsorption from the matrix (Fig. 5b). At low target concentrations, signal from the control ligands was equivalent to or higher than that from the anti-Salmonella spots. Even at high target concentrations, the signal from the control ligand spots was ineligible despite stronger signal from anti-Salmonella. However, the matrix effects did not cause a major issue for direct detection from chicken rinse, as the LOD of spiked Salmonella was 7.6 × 106 CFU/mL, which was 3.6 times higher than that in PBS (2.1 × 106 CFU/mL). The linear range was more than one order of magnitude narrower in chicken rinse (7.6 × 106–5.1 × 107 CFU/mL) (Fig. 5a) than in PBS buffer (2.1 × 106–4.1 × 108 CFU/mL), suggesting a more significant impact of the matrix on the dynamic range and potential in quantitative measurements.
Fig. 5

a Calibration curve of S. Typhimurium in chicken rinse overnight culture (n = 9). 0, 5.14 × 104, 5.14 × 105, 5.14 × 106, 2.57 × 107, 3.85 × 107, 5.14 × 107, and 5.14 × 108 CFU/mL of S. Typhimurium were spiked into the chicken rinse overnight culture. Negative results (S. Typhimurium concentrations ≤ 5.14 × 105 CFU/mL) were omitted from the calibration curve. b Representative difference images of samples spiked with 0, 5.14 × 105–5.14 × 108 CFU/mL of Salmonella. The top left image illustrates the spot arrangement: S: anti-Salmonella spots; E: anti-E. coli spots; I: IgG control spots. Samples were injected from the top to the bottom of the ligand array

Artificially contaminated samples

Detection of S. Typhimurium from chicken was simulated by spiking various concentrations (0, 6.8, 68, 6.8 × 102, 6.8 × 103, 6.8 × 104, 6.8 × 105, 6.8 × 106, and 6.8 × 107 CFU/mL) of the pathogen into freshly collected chicken rinse. Due to the high LODs established in previous experiments, an overnight culture step was necessary for low-level detection. After enrichment in BPW at 37 for 16 h, 200 μL of the samples were directly fed into the SPRi system for detection. According the SPRi reflectivity (Fig. 6a) and difference images (Fig. 6b), Salmonella grew to detectable concentrations after enrichment in the matrix regardless of the initial spiking level, and as low as 6.8 CFU/mL of the pathogen could be detected. Comparing the signal intensity of the samples co-enriched with the chicken rinse matrix to the calibration curve in chicken rinse (Fig. 5a), the concentrations of Salmonella in these artificially contaminated samples were estimated at approximately 107 CFU/mL after enrichment. In contrast, cells grown in uninhibited environments such as pure growth media can typically reach more than 108 CFU/mL. Since the final concentrations after enrichment were a result of incubation time, Salmonella growth kinetics in the matrix [37], and potential competition from background flora, the SPRi results suggest that the chicken rinse matrix and/or indigenous microflora did pose some, but minor inhibition to Salmonella growth. Nevertheless, growth of Salmonella which were initially spiked at different levels all plateaued at the stationary phase after overnight enrichment, resulting in little difference in observed SPRi intensity among these samples.
Fig. 6

a Final average SPR intensities of chicken rinse samples artificially spiked with 6.8–6.8 × 107 CFU/mL of S. Typhimurium. Signals were averaged from 9 identical ligand spots in the array. Samples were enriched at 37 °C overnight before injection. b Corresponding SPRi difference images. Spot arrangement was identical to Fig. 5b. Samples were injected from the top to the bottom of the ligand array

A previous study has demonstrated successful microbial culture directly on SPRi sensor chip using electrochemically grafted antibodies [29], which was capable of detecting 2.8 ± 19.6 CFU/mL of S. Typhimurium within ~ 5 h of enrichment. In pathogen screening, however, we find it more practical to enrich multiple samples simultaneously outside the SPR instrument before signal measurements. This way, the SPRi instrument could be dedicated to the detection task while enrichment tasks could be completed in higher throughput using conventional means and the overall efficiency of batch sample testing could be improved. On the other hand, a higher rate of antibody degradation at culture temperatures could be avoided to enable each sensor chip to process more sample injections before being recycled.

As shown in the chicken rinse studies, the SPRi method is remarkably tolerant of highly complex matrices compared to classical rapid detection techniques such as PCR and ELISA as no enzymes are involved in signal generation. This means minimal sample preparation is needed before detection, which saves time and labor. On the other hand, as a real-time flow-based technique, it provides a more robust and accurate readout than ELISA, which requires a plate reader to record the signal within a certain time window. In contrast, once programmed, the SPRi reflectivity variation information can be automatically recorded and stored for subsequent data review, much like chromatography based detection.

Regeneration and reuse of the sensor chip

Bound Salmonella cells and non-specifically adsorbed materials were removed and the ligands were regenerated after each injection. To disrupt the antibody-antigen binding, acidic (glycine-HCl pH 1.5–2.5), alkali (NaOH pH 10–13), and high ionic strength (1 M NaCl, 4 M MgCl2) solutions were flushed through the flow cell at 100 μL/min for 2 min. The most effective regeneration solution was found to be NaOH (pH 13), followed by glycine-HCl (pH 1.5), and NaCl (1 M) (ESM Fig. S2). Moreover, ligand regeneration exhibited a two-phase exponential decay function with respect to consecutive regeneration cycles (ESM Fig. S3), in which the first regeneration cycle moved the most bound materials, and removal rate gradually decreased in additional cycles. The impact of regeneration on the immobilized ligands was observed in two aspects: ligand loss and inactivation. After each regeneration cycle, the baseline decreased by 0.09% at 1000 μg/mL ligand spots and 0.02% at 250 μg/mL spots. On the other hand, the specific SPRi response typically decreased by 3–5%/cycle compared to identical injections prior to regeneration (ESM Fig. S4). Still, a sufficiently large number of samples could be analyzed on each sensor chip (e.g., at an inactivation rate of 5%, ~ 45 samples could be analyzed before 90% of the antibodies are inactivated). The regeneration induced signal loss could also be taken into account when semi-quantification is needed. In the future, the regeneration conditions could be further fine-tuned to achieve effective removal of the target while maintaining maximal antibody activity.

Conclusions

In this study, we have developed a SPRi-based assay for label-free rapid detection of foodborne Salmonella. The antibodies were immobilized onto the sensor chip via EDC mediated coupling, in which low pH (4.6) and high antibody concentration (up to 1000 μg/mL) were found to improve cross-linking efficiency, thereby enhancing resultant SPRi signal. A pause flow protocol involving sample injection at high flow rates (50–100 μL/min) and a short paused flow period for target-ligand binding was adopted to address the low sensitivity, slow equilibrium, or long assay time under continuous flow conditions. As a result, a balance was achieved between the duration and sensitivity of the assay. The method shows good specificity towards targeted Salmonella serotypes (S. Enteritidis, S. Heidelberg, and S. Typhimurium, which are most frequently found in foodborne outbreaks), with LODs of 2.1 × 106 in PBS and 7.6 × 106 CFU/mL in the presence of chicken rinse matrix and enriched indigenous microflora. Since the food matrix and competitive microflora had little impact on detection limits, no sample preparation was needed to remove inhibition and the samples could be directly injected into the SPRi system for analysis. Artificially spiked Salmonella were detected from chicken rinse samples at as low as 6.8 CFU/mL after overnight enrichment.

At the current stage, the SPRi assay is limited by two factors. First, though the assay proves successful in qualitative detection, its capability in quantitative analyses is limited by a relatively narrow linear dynamic range (approximately one order of magnitude). With continued improvement in optics and further method development, however, widened dynamic ranges are expected in the future. Second, due to the relatively high LOD at the current stage, an overnight enrichment step is still required for low-level detection. In order for SPRi to be employed for versatile multiplex pathogen screening, potential growth inhibition during non-selective culture must be considered. Alternatively, SPRi detection may be coupled with non-culture-based sample preparation techniques for targeted enrichment, which remains a future research direction.

Notes

Acknowledgements

The authors thank Dr. Marinella Sandros, Dr. Fatima Hibti, and Dr. Chiraz Frydman for their helpful discussions and Dr. Nasreen Bano for the maintenance and preparation of bacterial culture.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflicts of interest.

Supplementary material

216_2017_810_MOESM1_ESM.pdf (485 kb)
ESM 1 (PDF 484 kb)

References

  1. 1.
    Scallan E, Griffin PM, Angulo FJ, Tauxe RV, Hoekstra RM. Foodborne illness acquired in the United States-unspecified agents. Emerg Infect Dis. 2011;17:16–22.CrossRefPubMedPubMedCentralGoogle Scholar
  2. 2.
    Scallan E, Hoekstra RM, Angulo FJ, Tauxe RV, Widdowson MA, Roy SL, et al. Foodborne illness acquired in the United States-major pathogens. Emerg Infect Dis. 2011;17:7–15.CrossRefPubMedPubMedCentralGoogle Scholar
  3. 3.
    Hoffmann S, Maculloch B, Batz M. Economic burden of major foodborne illnesses acquired in the United States. USDA, ERS. 2015. https://www.ers.usda.gov/webdocs/publications/43984/52807_eib140.pdf?v=42136. Accessed 20 Sep 2017.
  4. 4.
    Scharff RL. State estimates for the annual cost of foodborne illness. J Food Protect. 2015;78:1064–71.CrossRefGoogle Scholar
  5. 5.
    Suo B, He Y, Paoli G, Gehring A, S-I T, Shi X. Development of an oligonucleotide-based microarray to detect multiple foodborne pathogens. Mol Cell Probes. 2010;24:77–86.CrossRefPubMedGoogle Scholar
  6. 6.
    Wang RF, Cao WW, Cerniglia CE. A universal protocol for PCR detection of 13 species of foodborne pathogens in foods. J Appl Microbiol. 1997;83:727–36.CrossRefPubMedGoogle Scholar
  7. 7.
  8. 8.
    Gehring AG, Tu S-I. High-throughput biosensors for multiplexed food-borne pathogen detection. Annu Rev Anal Chem. 2011;4:151–72.CrossRefGoogle Scholar
  9. 9.
    Elnifro EM, Ashshi AM, Cooper RJ, Klapper PE. Multiplex PCR: optimization and application in diagnostic virology. Clin Microbiol Rev. 2000;13:559.CrossRefPubMedPubMedCentralGoogle Scholar
  10. 10.
    Dudak FC, Boyacı İH. Rapid and label-free bacteria detection by surface plasmon resonance (SPR) biosensors. Biotechnol J. 2009;4:1003–11.CrossRefPubMedGoogle Scholar
  11. 11.
    Shankaran DR, Gobi KV, Miura N. Recent advancements in surface plasmon resonance immunosensors for detection of small molecules of biomedical, food and environmental interest. Sensor Actuat B: Chem. 2007;121:158–77.CrossRefGoogle Scholar
  12. 12.
    Fratamico PM, Strobaugh TP, Medina MB, Gehring AG. Detection of Escherichia coli O157 : H7 using a surface plasmon resonance biosensor. Biotechnol Tech. 1998;12:571–6.CrossRefGoogle Scholar
  13. 13.
    Mazumdar SD, Hartmann M, Kämpfer P, Keusgen M. Rapid method for detection of Salmonella in milk by surface plasmon resonance (SPR). Biosens Bioelectron. 2007;22:2040–6.CrossRefPubMedGoogle Scholar
  14. 14.
    Naimushin AN, Soelberg SD, Nguyen DK, Dunlap L, Bartholomew D, Elkind J, et al. Detection of Staphylococcus aureus enterotoxin B at femtomolar levels with a miniature integrated two-channel surface plasmon resonance (SPR) sensor. Biosens Bioelectron. 2002;17:573–84.CrossRefPubMedGoogle Scholar
  15. 15.
    Zhang D, Yan Y, Li Q, Yu T, Cheng W, Wang L, et al. Label-free and high-sensitive detection of Salmonella using a surface plasmon resonance DNA-based biosensor. J Biotechnol. 2012;160:123–8.CrossRefPubMedGoogle Scholar
  16. 16.
    Kai E, Ikebukuro K, Hoshina S, Watanabe H, Karube I. Detection of PCR products of Escherichia coli O157:H7 in human stool samples using surface plasmon resonance (SPR). FEMS Immunol Med Mic. 2000;29:283–8.CrossRefGoogle Scholar
  17. 17.
    Subramanian A, Irudayaraj J, Ryan T. A mixed self-assembled monolayer-based surface plasmon immunosensor for detection of E. coli O157: H7. Biosens Bioelectron. 2006;21:998–1006.CrossRefPubMedGoogle Scholar
  18. 18.
    Waswa J, Irudayaraj J, DebRoy C. Direct detection of E. coli O157: H7 in selected food systems by a surface plasmon resonance biosensor. Lwt-Food Sci Technol. 2007;40:187–92.CrossRefGoogle Scholar
  19. 19.
    Waswa JW, Debroy C, Irudayaraj J. Rapid detection of Salmonella enteritidis and Escherichia coli using surface plasmon resonance biosensor. J Food Process Eng. 2006;29:373–85.CrossRefGoogle Scholar
  20. 20.
    BK O, Kim YK, Park KW, Lee WH, Choi JW. Surface plasmon resonance immunosensor for the detection of Salmonella typhimurium. Biosens Bioelectron. 2004;19:1497–504.CrossRefGoogle Scholar
  21. 21.
    Oh BK, Lee W, Chun BS, Bae YM, Lee WH, Choi JW. Surface plasmon resonance immunosensor for the detection of Yersinia enterocolitica. Colloid Surface A. 2005;257-58:369–74.CrossRefGoogle Scholar
  22. 22.
    Oh BK, Lee W, Kim YK, Lee WH, Choi JW. Surface plasmon resonance immunosensor using self-assembled protein G for the detection of Salmonella paratyphi. J Biotechnol. 2004;111:1–8.CrossRefPubMedGoogle Scholar
  23. 23.
    Dudak FC, Boyacı İH. Development of an immunosensor based on surface plasmon resonance for enumeration of Escherichia coli in water samples. Food Res Int. 2007;40:803–7.CrossRefGoogle Scholar
  24. 24.
    Lan Y-B, Wang S-Z, Yin Y-G, Hoffmann WC, Zheng X-Z. Using a surface plasmon resonance biosensor for rapid detection of Salmonella Typhimurium in chicken carcass. J Bionic Eng. 2008;5:239–46.CrossRefGoogle Scholar
  25. 25.
    Vaisocherová-Lísalová H, Víšová I, Ermini ML, Špringer T, Song XC, Mrázek J, et al. Low-fouling surface plasmon resonance biosensor for multi-step detection of foodborne bacterial pathogens in complex food samples. Biosens Bioelectron. 2016;80:84–90.CrossRefPubMedGoogle Scholar
  26. 26.
    Taylor AD, Ladd J, Yu Q, Chen S, Homola J, Jiang S. Quantitative and simultaneous detection of four foodborne bacterial pathogens with a multi-channel SPR sensor. Biosens Bioelectron. 2006;22:752–8.CrossRefPubMedGoogle Scholar
  27. 27.
    Foudeh AM, Trigui H, Mendis N, Faucher SP, Veres T, Tabrizian M. Rapid and specific SPRi detection of L. pneumophila in complex environmental water samples. Anal Bioanal Chem. 2015;407:5541–5.CrossRefPubMedGoogle Scholar
  28. 28.
    Aura AM, D'Agata R, Spoto G. Ultrasensitive detection of Staphylococcus aureus and Listeria monocytogenes genomic DNA by nanoparticle-enhanced surface plasmon resonance imaging. ChemistrySelect. 2017;2:7024–30.CrossRefGoogle Scholar
  29. 29.
    Bouguelia S, Roupioz Y, Slimani S, Mondani L, Casabona MG, Durmort C, et al. On-chip microbial culture for the specific detection of very low levels of bacteria. Lab Chip. 2013;13:4024–32.CrossRefPubMedGoogle Scholar
  30. 30.
    Yamasaki T, Miyake S, Nakano S, Morimura H, Hirakawa Y, Nagao M, et al. Development of a surface plasmon resonance-based immunosensor for detection of 10 major O-antigens on Shiga toxin-producing Escherichia coli, with a gel displacement technique to remove bound bacteria. Anal Chem. 2016;88:6711–7.CrossRefPubMedGoogle Scholar
  31. 31.
    Mondani L, Delannoy S, Mathey R, Piat F, Mercey T, Slimani S, et al. Fast detection of both O157 and non-O157 Shiga-toxin producing Escherichia coli by real-time optical immunoassay. Lett Appl Microbiol. 2016;62:39–46.CrossRefPubMedGoogle Scholar
  32. 32.
    Grabarek Z, Gergely J. Zero-length crosslinking procedure with the use of active esters. Anal Biochem. 1990;185:131–5.CrossRefPubMedGoogle Scholar
  33. 33.
    Anonymous. Crosslinking technical handbook. Thermo Fisher Scientific. https://tools.thermofisher.com/content/sfs/brochures/1602163-Crosslinking-Reagents-Handbook.pdf. Accessed 14 Sep 2017.
  34. 34.
    de Oliveira MMM, Brugnera DF, Alves E, Piccoli RH. Biofilm formation by Listeria monocytogenes on stainless steel surface and biotransfer potential. Braz J Microbiol. 2010;41:97–106.CrossRefPubMedPubMedCentralGoogle Scholar
  35. 35.
    Mafu AA, Roy D, Goulet J, Magny P. Attachment of Listeria monocytogenes to stainless steel, glass, polypropylene, and rubber surfaces after short contact times. J Food Protect. 1990;53:742–6.CrossRefGoogle Scholar
  36. 36.
    Bretscher MS. Recap on cell migration. Traffic. 2008;9:198–9.CrossRefPubMedGoogle Scholar
  37. 37.
    Maier RM. Bacterial growth. In: Pepper IL, Gerba CP, Gentry TJ, editors. Environmental microbiology. 3rd ed. San Diego: Academic Press; 2015. p. 37–56.CrossRefGoogle Scholar

Copyright information

© This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2017

Authors and Affiliations

  1. 1.United States Department of Agriculture, Agricultural Research ServiceU.S. National Poultry Research CenterAthensUSA

Personalised recommendations