Rapid and simple on-site salmonella detection in food via direct sample loading using a lipopolysaccharide-imprinted polymer

Abstract

Salmonella is a major foodborne pathogen that causes salmonellosis, which is characterized by symptoms such as diarrhea, fever, and abdominal cramps. Existing methods for detecting Salmonella, such as culture plating, ELISA, and PCR, are accurate but time-consuming and unsuitable for on-site applications. In this study, we developed a rapid and sensitive electrochemical sensor using a molecularly imprinted polymer (MIP) to detect Salmonella typhimurium (S. typhimurium) by targeting lipopolysaccharides (LPS). Polydopamine (PDA) was used as the polymer matrix because of its cost-efficiency and functional versatility. The sensor demonstrated high sensitivity and selectivity, with a detection limit of 10 CFU/mL and a linear response over the 10²–10⁸ CFU/mL range. The specificity of the sensor was validated against other gram-positive and gram-negative bacteria and showed no significant cross-reactivity. Furthermore, the sensor performed effectively in real food samples, including tap water, milk, and pork, without complex preprocessing. These results highlight the potential of the LPS-imprinted MIP sensor for practical on-site detection of S. typhimurium, improving food safety monitoring and preventing outbreaks in food-handling environments.

Introduction

According to the World Health Organization, 58 food safety incidents involving Salmonella have recently been reported across member states and territories [1]. Salmonella is a major foodborne pathogen found in various food products, including poultry, eggs, meat, dairy products, and water [2]. It causes salmonellosis, which is characterized by symptoms such as diarrhea, fever, and abdominal cramps, and can lead to severe complications, particularly in vulnerable populations such as children, the elderly, and immunocompromised individuals [3]. Among the serotypes, Salmonella typhimurium (S. typhimurium) is the most prevalent, accounting for a large proportion of infections, which can result in symptoms such as diarrhea, abdominal pain, and fever, with severe cases often leading to a high mortality rate [4]. Therefore, food safety organizations such as the Food and Drug Administration (FDA) and the European Food Safety Authority (EFSA) regularly inspect food processing facilities and restaurants for S. typhimurium to prevent contamination and ensure food safety [5].

Current detection methods of S. typhimurium include culture plating, enzyme-linked immunosorbent assay (ELISA), and polymerase chain reaction (PCR) [6]. However, these methods are typically time-consuming and require laboratory facilities, rendering them unsuitable for on-site testing [7]. Therefore, rapid and sensitive detection methods are crucial for the early identification of S. typhimurium contamination, particularly in food processing or handling environments where timely results are critical to prevent outbreaks [8,9,10,11]. Molecularly imprinted polymers (MIPs) are synthetic materials designed to selectively bind specific target molecules, mimicking the lock-and-key mechanisms of biological receptors [12, 13]. MIPs exhibit high selectivity, cost effectiveness, and stability, making them promising candidates for advanced sensing systems. In pathogen detection, MIPs are primarily designed to imprint whole bacteria [12, 14]. However, targeting entire bacteria poses challenges owing to their complex surface chemistry, toxicity, structural variability, and susceptibility to deformation, which can affect the uniformity and reliability of the resulting MIPs [15]. To overcome these limitations, lipopolysaccharide (LPS) imprinting, which targets major components of the outer membrane of gram-negative bacteria, are used. The structure of LPS is highly variable among bacterial species, particularly in the O-polysaccharide chain, which is specific to individual species and plays a crucial role in serotype differentiation [16]. This structural diversity arises from differences in sugar composition, sequence, glycosidic linkages, and substitutions, leading to an extensive range of O-chain conformations. Additionally, variations in the Core oligosaccharide and Lipid A regions further contribute to species-specific LPS profiles [17]. These distinct structural features enhance the selectivity of the imprinting process, enabling the precise molecular recognition of the target bacterial species and minimizing cross-reactivity with non-target bacteria. Compared to whole-bacteria imprinting, LPS imprinting offers improved thermal and pH stability, facilitating efficient template imprinting and removal, while eliminating the risks associated with handling live bacteria. In addition, LPS imprinting enhances the reproducibility of the fabrication process, increasing its potential for practical applications.

In this study, we attempted to imprint LPS of S. typhimurium using polydopamine (PDA) for the sensor. PDA, which is a polymer, was chosen because of its cost efficiency and high functionalization capabilities [18]. The O-antigen of LPS, which contains D-rhamnose residues, introduces cis-diol groups that strongly interact with boronic acid-functionalized MIPs [19]. Thus, S. typhimurium LPS-imprinted MIPs demonstrate high sensitivity and selectivity for foodborne pathogen detection using electrochemical methods. Notably, the sensor exhibited excellent detection performance for food samples, highlighting its potential for on-site applications with direct sample loading and minimal preprocessing requirements.

Materials and methods

Reagents and apparatus

The reagents were purchased from the following suppliers. Screen-printed electrodes (SPE, 110SWCNT, Metrohm DropSens, Oviedo, Spain). LPS from salmonella enterica typhimurium, ammonia solution (28–30%), dopamine hydrochloride, ethanol (99.9%), 4-formylphenylboronic acid, sodium cyanoborohydride, acetic acid (glacial, 99.9%), acetonitrile (99.9%), potassium ferricyanide, and potassium ferrocyanide were obtained from Sigma-Aldrich Corp. (St. Louis, MO, USA). Phosphate buffer (0.2 M, pH 8.5) and Tris buffer (10 mM, pH 8.5) were purchased from Biosesang (South Korea). The morphology of the MIP was measured using an S-4800 field-emission scanning electron microscope (FE-SEM; Hitachi, Japan) and a JEM-2100 F analytical scanning transmission electron microscope (STEM; JEOL Co., Japan). Electrochemical measurements were performed using a CHI627b analyzer (CH Instruments Inc., Austin, TX, USA). Size and composition analyses were performed using an ELSZneo dynamic light scattering (DLS) analyzer (Otsuka Electronics Co., Ltd., Japan) and a Nicolet iS50 Fourier-transform infrared spectrometer (FT-IR; Thermo Fisher Scientific Inc., USA). NanoDrop One/One^C (Thermo Fisher Scientific Inc., USA) was used to measure the optical density of bacterial solutions at a wavelength of 600 nm (OD600).

Synthesis of polydopamine-based MIP

To prepare the PDA core, an ammonia solution (28–30%), deionized water, and ethanol were mixed at a volume ratio of 1:30:13 and stirred for 30 min [20]. Subsequently, dopamine hydrochloride was added to the mixture and stirred gently at room temperature for 24 h. The resulting particles were then washed three times with deionized water and dried at 70 °C. A mixture of polydopamine particles (20 mg), 4-formylphenylboronic acid (FPBA, 20 mg), and sodium cyanoborohydride (20 mg) was dispersed in 4 ml of ethanol. The resulting dispersion was sonicated for 10 min and stirred for 24 h. The boronic acid-functionalized polydopamine particles were washed three times with ethanol. The collected particles were then stirred gently with LPS in phosphate buffer at a weight ratio of 1:0.04 for 1 h to allow binding interactions. Unbound LPS was washed off with phosphate buffer. LPS-bound nanoparticles were imprinted onto dopamine hydrochloride in Tris buffer (2.8 mL) for 45 min. To remove the LPS template, the collected particles were washed with an eluting solvent (5% acetic acid solution containing 30% acetonitrile) to yield the LPS-imprinted MIP. For non-imprinted polymer (NIP) preparation, the same procedure was followed, excluding the addition of LPS as template.

Preparation of MIP sensor and electrochemical measurements

To fabricate MIP-doped electrodes, MIP dispersed in deionized water was drop-cast onto a carbon working electrode. The electrode was dried at 60 °C, and the same procedure was repeated three times. Any excess unbound MIP on the electrode was rinsed several times with deionized water, resulting in the formation of an MIP-coated electrode. The prepared electrode was incubated in an analyte-containing solution for the optimized incubation time prior to measurement (Figure S1). All electrochemical measurements were performed in phosphate buffered saline (PBS) (pH 7.4) containing 5 mmol/L [Fe(CN)₆]^3−/4−. Differential pulse voltammetry (DPV) measurement was performed in the range of -0.3–0.4 V with a pulse amplitude of 0.05 V and a pulse width of 0.05 s. Cyclic voltametric (CV) measurement was recorded by cycling potential from − 0.3 to 0.4 V at a scan rate 50 mV s^− 1.

Bacterial strain and preparation

All biological resources were obtained from the Korean Collection for Type Cultures (KCTC). Staphylococcus aureus (KCTC3881) and Staphylococcus epidermidis (KCTC3958) were cultured on tryptic soy agar plates, whereas Salmonella typhimurium (KCTC2055), Escherichia coli (KCTC2571), and Klebsiella pneumoniae (KCTC12385) were cultured on nutrient agar plates. The cells were collected from the plates and suspended in PBS, followed by centrifugation at 5000 rpm for 10 min to remove impurities. The pellets were subsequently resuspended in PBS and diluted to the desired concentration.

Food matrices preparation

Tap water, milk, and pork were used as the test matrices to evaluate the performance of the sensor in real food samples [21]. Food samples were inoculated with S. typhimuruim at the desired concentrations before the experiment. For solid-state samples, such as pork, the inoculated samples were mixed with water and allowed to settle. The supernatant obtained was used for sensor measurements. Liquid-state samples were loaded directly onto the sensor without pretreatment.

Results and discussion

Figure 1 shows a schematic of the highly sensitive and selective sensor for S. typhimurium, which was developed by imprinting LPS onto a PDA surface. Polydopamine is an excellent template for MIP synthesis because of its versatile functional groups (catechol and amine), which enable strong interactions, tunable thickness depending on pH, and strong adhesion to various surfaces [22]. In particular, the thickness of the PDA coatings can be easily tuned based on the pH conditions during deposition [23]. This allows for precise control of the template layer thickness and optimization of the cavity structure in the resulting MIP. Nanospherical MIPs with large surface areas have been used to increase the number of binding sites [24]. Sodium cyanoborohydride was used to stabilize the binding of FPBA to the PDA core, thereby enabling the formation of boronic acid on the surface for cis-diol interaction with LPS [19, 25]. Salmonella is controlled in food trucks and restaurants to ensure food safety through hygiene training, safe ingredient sourcing, proper cooking and storage temperatures, prevention of cross-contamination, regular hand washing, thorough sanitation, and routine inspection [26]. Furthermore, onsite sensor technology for rapid and easy Salmonella detection is essential for enhancing food safety and preventing contamination.

Fig. 1

Schematic of MIP preparation and application for on-site pathogen detection in food samples

Characterization of MIP

Figure 2A shows the structure of the LPS-imprinted MIP. An LPS-recognizing MIP was produced by introducing 4-formylphenylboronic acid (FPBA) onto the surface of the PDA nanoparticles to bind with the cis-diol structure of LPS. A nanoscale PDA shell was subsequently formed over the core particle and LPS was eluted to create binding sites on the MIP. Figure 2B shows transmission electron microscopy (TEM) images and energy-dispersive X-ray spectroscopy (EDS) mapping of the PDA core + FPBA and final MIP (PDA core + FPBA + PDA shell). After adding FPBA to the PDA core, the atomic compositions of B, C, N, and O were 10.3, 76.9, 4.1, and 8.8%, respectively. Following the formation of the PDA shell, the surface atomic compositions changed to 8.6%, 79.6%, 4.96%, and 6.79%, respectively. These changes indicate that the increased boron atom composition confirmed the successful formation of boronic acid on the PDA core surface, whereas the enrichment of C, N, and O validated the subsequent formation of the PDA shell [22, 27]. In addition, the changes in the functional groups throughout the fabrication process, from the PDA core to PDA-FPBA to the final MIP, were confirmed using FT-IR (Fig. 2C). Characteristic absorption bands of PDA in the PDA core and MIP were identified, including 3368 cm⁻¹ for N-H stretching and 1609 cm⁻¹ for C = C stretching [28]. For PDA-FPBA, a peak corresponding to B-O-B stretching appeared at 1020 cm⁻¹, whereas the intensity of the N-H and C = C stretching bands decreased, confirming the binding of FPBA to the PDA core [19]. XPS analysis was performed to examine the essential elements present on the surface. The presence of C, O, N, and B in the PDA core, PDA-FPBA, and MIP was confirmed. The deconvoluted C 1s, O 1s, and N 1s spectra are shown in Figs. 2E and S2. In the PDA-FPBA sample, the decrease in surface PDA compared to the core sample was accompanied by a reduction in the C = O, O = C, C = C, and C = N peaks. Consistent with the FT-IR results, the modification of the PDA core with FPBA resulted in the appearance of a B 1s peak. In the MIP sample, the C = O, O = C, C = C, and C = N peaks were partially recovered, most likely owing to the presence of the polydopamine shell [29, 30]. Therefore, the results of the FTIR and XPS analyses supported the uniform incorporation of boronic acid and dopamine into the MIP.

For the imprinting time optimization, the thickness of shell prepared with 2-hour and 1-hour, 45 min, and 30 min imprinting times were measured with DLS (Figure S3). It was determined that the MIP with 45 min imprinting time allows efficient LPS imprinting without hindering subsequent LPS rebinding, exhibiting the best rebinding performance. As shown in Fig. 2D, the average MIP particle size increased by approximately 50 nm from the PDA core size of 295.5 nm to 358.2 nm, suggesting that an imprinting time of one hour was optimal through the optimization process.

Fig. 2

MIP fabrication and characterization: (A) internal structure of MIP nanoparticle, (B) TEM images and EDS mapping of FPBA + MIP core and MIP, (C) FT-IR spectra of MIP, PDA-FPBA, and PDA core, (D) XPS and high-resolution XPS spectra of N 1s region of MIP, and (E) DLS result and histogram of PDA core and MIP

LPS detection performance of MIP sensor

The sensing platform was fabricated using a working electrode made of single-walled carbon nanotubes (CNTs) and a counter electrode made of carbon both of which exhibit high electrical conductivity in various electrolytes [31, 32]. Ag/AgCl was used as the reference electrode. The working electrode was further modified with the final MIP using drop-casting to evaluate its capture performance (Fig. 3A). The SEM image in Fig. 3B shows that the nanosphere-shaped MIP was uniformly attached to the CNT electrode.

To evaluate the sensing functionality of Salmonella, the MIP-coated electrode was incubated with the LPS of S. typhimurium in PBS for 20 min, followed by washing with PBS to remove the unbound LPS, as shown in Fig. 3A. The incubation time was optimized by measuring DPV at 10-minute intervals, and pH 7.4 PBS was used to prevent structural changes in LPS (Figure S1). The binding of Salmonella LPS to SL-MIP was confirmed through cyclic voltammetry by comparing the bare electrode (CTL), MIP-coated electrode (SL-MIP), and LPS-incubated MIP electrode (LPS + SL-MIP). As shown in Fig. 3C, the oxidation and reduction peaks were observed at 0.185 and 0.057 V, respectively. Following MIP modification, a slight decrease in peak current and an increase in ΔEpa were observed, indicating a minor increase in resistance caused by the PDA layer. Notably, the addition of S. typhimurium LPS (1 mg/mL) to the SL-MIP electrode considerably reduced the oxidation and reduction current peaks, indicating increased resistance owing to LPS binding to SL-MIP.

Furthermore, we conducted differential pulse voltammetry (DPV), which has higher sensitivity and discrimination of analytes than similar oxidation potentials [33]. As shown in Fig. 3D, the DPV peak current at potentials of 0.11–0.12 V decreases with increasing LPS concentration, indicating the binding of LPS to the SL-MIP electrode. This concentration-dependent response exhibited a remarkable linear relationship between the logarithmic concentration of LPS (0.1–100 µg) and the DPV peak current, with an R² value exceeding 0.999 (Fig. 3E). The linear regression results for Fig. 3E showed a p-value of 1.39$\:\times\:$10⁻⁷, indicating a highly significant correlation. When comparing the responses of NIP and MIP to 100 µg of LPS, MIP exhibited a current change of 37.6 µA, whereas NIP exhibited a significantly lower and inconsistent response of 5.5 µA (Fig. 3F). The p-value for the difference between MIP and NIP was 0.043, confirming a statistically significant difference (p < 0.05). These findings demonstrated that the imprinting cavities on the LPS-mediated MIP surface were highly specific for Salmonella LPS. To further evaluate the reproducibility and repeatability of the prepared MIP sensor, identical electrodes were fabricated with MIPs produced on different days and used to detect LPS at a 1.0 mg/mL concentration (Figure S4). The peak current values remained consistent across the different electrodes, yielding a relative standard deviation (RSD) of 1.36%, demonstrating high reproducibility. Meanwhile, the DPV measured from MIPs produced on the same day exhibited minimal variation, a RSD of 1.98%, confirming repeatability.

Fig. 3

LPS-detection performance of the SL-MIP: (A) experimental procedure for LPS and pathogen detection, (B) SEM image of bare electrode and SL-MIP, scale bar$\:=1\:\mu\:m$, (C) CV curves of bare electrode, MIP, and LPS-bound MIP at 50mVs⁻¹, (D) DPV curves for the different concentrations of LPS from 0.1 to 100 $\:\mu\:$g, (E) linear calibration curve of the SL-MIP for different LPS concentrations, and (F) current changes of MIP and NIP for 100 $\:\mu\:$g LPS

Salmonella detection performance of MIP sensor

Figures 4A and S5 show an SEM image of MIP incubated with S. typhimurium for 30 min, followed by washing with PBS. Compared to LPS, the surface of the whole cell exhibits a more complex structure, which may explain the need for a longer incubation time to ensure sufficient binding strength (Figure S1). The areas marked with red dashed lines indicate the locations of the bacterial cells or their remnants surrounded by MIP. When comparing the quantitative detection capability for whole S. typhimurium bacteria, increasing cell concentration resulted in a gradual decrease in peak current at a potential of 0.11–0.12 V, similar to the response of MIP to LPS (Fig. 4B). The coefficient of determination (R²) was 0.997, indicating a strong linear correlation between the logarithmic concentration of S. typhimurium and the peak current within the 10²–10⁸ CFU/mL range, with a detection limit of 10 CFU/mL (Figs. 4C and S6). The linear regression results for Fig. 4C showed a p-value of 4.42 $\:\times\:$ 10⁻⁴, confirming a statistically significant correlation. For the NIP, no significant difference was observed between the current responses to PBS and S. typhimurium at 10⁸ CFU/mL (Fig. 4D). The p-value for the difference between MIP and NIP was 0.0077, indicating a highly significant difference (p < 0.01). To evaluate the selectivity, the response of the MIP sensor was tested against various bacterial strains commonly found in food samples. First, we compared S. typhimurium (a gram-negative bacterium) with S. aureus and S. epidermidis (gram-positive bacteria), which lack LPS on their cell surfaces. As shown in Fig. 4E, the MIP template generated by the S. typhimurium LPS exhibited no response to S. aureus or S. epidermidis (10⁸ CFU/mL). We investigated whether the MIP sensor could discriminate between gram-negative bacteria, considering the presence of the LPS on the cell surface. Notably, the MIP sensor responded only to S. typhimurium and not to other gram-negative bacteria, such as E. coli and K. pneumonia (Fig. 4F). This selectivity can be attributed to the species-specific variations in the O-polysaccharide chain, as well as structural differences in the core and lipid A regions, which influence binding interactions with the imprinted cavities. Therefore, these findings suggest that the imprinted sites of MIP are highly specific to S. typhimurium, highlighting the ability of the sensor to differentiate between LPS from various bacterial species. Furthermore, the reproducibility and repeatability of the sensor were also evaluated for S. typhimurium at 10⁷ CFU/mL (Figure S7). The peak current measured using MIPs produced on the same day demonstrated a repeatability RSD of 3.31%. When MIPs produced on different days were tested, the reproducibility RSD was 2.14%, indicating high consistency across different production batches. Additionally, Cohen’s kappa analysis was performed to compare the developed sensor with the traditional culture method [34]. The Cohen’s kappa index at 10² CFU/mL was 0.57, indicating moderate agreement, which improved to substantial (10³ CFU/mL) or almost perfect (10⁴ CFU/mL) agreement at higher inoculum concentrations, demonstrating the sensor’s high correlation with conventional detection methods.

Fig. 4

Detection performance of the SL-MIP for S. typhimurium: (A) SEM image of S. typhimurium cell traces surrounded by MIP nanoparticles, (B) DPV curves for different concentrations of S. typhimurium (from 10⁰ to 10⁸ CFU/mL), (C) linear calibration curve of the SL-MIP for different concentrations of S. typhimurium, (D) current changes of MIP and NIP for 10⁸ CFU mL^− 1S. typhimurium, (E) DPV curves for gram-positive bacteria (S. aureus and S. epidermidis) and S. typhimurium at 10⁸ CFU mL^− 1, (F) DPV curves for gram-negative bacteria (E. coli and K. pneumoniae) and S. typhimurium at 10⁸ CFU mL^− 1

Salmonella detection performance of MIP sensor in food matrices

Salmonella is controlled in food trucks and restaurants to ensure food safety through hygiene training, safe ingredient sourcing, proper cooking and storage temperatures, prevention of cross-contamination, regular hand washing, thorough sanitation, and routine inspection [1]. Therefore, we investigated whether our MIP sensor could rapidly detect S. typhimurium in real food samples for the on-site detection of foodborne pathogens. Figure S8 shows the DPV response to various concentrations of S. typhimurium in food samples, including tap water, milk, and pork, whereas Fig. 5B illustrates the peak current of the DPV corresponding to these concentrations [35, 36]. The concentrations were selected based on levels known to cause diseases in humans. Notably, the MIP sensor exhibited a concentration-dependent response to the samples, even without pre-treatment. A baseline shift was observed owing to interfering with molecules in each sample, which could alter the conductivity of the sensor. Nevertheless, the degree of the decrease in the response current remained consistent across S. typhimurium concentrations. Linear regression curves with identical slopes were obtained for all the food samples, with corresponding R² values of 0.80, 0.90, and 0.87. To highlight the performance of the prepared MIP sensor, a comparison of the results with recently reported S. typhimurium sensors within the last three years is provided in Table 1.

Fig. 5

Performance evaluation of SL-MIP for the detection of S. typhimurium in food samples: (A) on-site detection of S. typhimurium in food trucks and restaurants, (B) peak current of DPV curves for varying S. typhimurium concentrations in tap water, milk, and pork

Table 1 Comparison of Salmonella typhimurium Senor performance

Conclusion

The detection of Salmonella is essential because it is a major cause of foodborne illnesses, leading to symptoms such as diarrhea, fever, and abdominal pain [41]. It poses critical risks to vulnerable populations such as children, the elderly, and immunocompromised individuals [42, 43]. Early detection facilitates in preventing large-scale outbreaks and contamination. Compliance with strict food safety regulations requires monitoring of Salmonella. In this study, a MIP specific to Salmonella LPS was developed for the detection of S. typhimurium, a common foodborne pathogen. Compared with whole-bacterial imprinting, LPS imprinting offers several advantages. It provides higher specificity by reducing cross-reactivity with non-target bacteria. This process is simpler and safer and minimizes biosafety concerns. LPS templates exhibit high stability and reproducibility. In addition, LPS-imprinted MIPs are easier to handle and more versatile than other detection platforms. The developed sensor exhibited high analytical performance for both S. typhimurium LPS and whole bacterial cells. The detection limit was determined to be 10 CFU/mL, with a linear relationship between the DPV peak current and S. typhimurium concentrations in the 10²–10⁸ CFU/mL range. However, the assay time was approximately 30 min, as it required sufficient time for bacteria cells to bind to MIP. To address this limitation, replacing the PDA core with a magnetic core, such as iron oxide, could enable rapid magnetic separation. Nevertheless, the MIP successfully captured S. typhimurium in various food samples. This approach exhibits high potential for on-site detection of foodborne pathogens and offers highly selective sensitive, and a simple sample-loading process.