Abstract
This study developed a rapid detection method for Salmonella based on real-time recombinase polymerase amplification (real-time RPA). The method exhibited excellent specificity and could amplify target genes within 20 min at 39°C. It achieved a Limit of Detection (LOD50) of 47 CFU/mL. To evaluate detection performance, artificially contaminated food samples—including egg products, chocolate products, meat products, grain-based products, and soy products—were tested. Prior to real-time RPA detection, the samples underwent an enrichment step by shaking incubation at 36°C for 6 h. The real-time RPA method demonstrated consistent and robust performance across diverse food matrices, with relative LOD (RLOD) values below 2.5, satisfying the validation criteria outlined in GUOBIAO 4789.45 (GB 4789.45). A chi-square test conducted on bulk pork samples further confirmed no significant difference between the real-time RPA method and the GB 4789.4 standard method (p > 0.05). These findings highlight the potential of real-time RPA as a reliable and efficient alternative to GB 4789.4 for detecting Salmonella, enhancing food safety monitoring practices.
Introduction
Salmonella is one of the most prevalent foodborne pathogens globally, with nearly all warm-blooded and many cold-blooded animals serving as natural reservoirs (Yang et al., 2010). These facultatively anaerobic, Gram-negative bacilli belong to the family Enterobacteriaceae and are commonly found in a wide range of food products, including pork, eggs, poultry, seafood, unpasteurized dairy products, and vegetables (Gu et al., 2018; Jackson et al., 2013). Salmonella infections pose a significant public health burden in both developing and industrialized countries, with an estimated 93.8 million cases of nontyphoidal gastroenteritis and approximately 155,000 associated deaths reported annually (Ao et al., 2015; Kirk et al., 2015; Majowicz et al., 2010). Many infections are zoonotic in origin, as the intestinal tracts of domestic and wild animals serve as reservoirs for this pathogen (Horton et al., 2013). Transmission may occur through multiple routes, including cross-contamination with fecal matter from infected animals, fecal-oral transmission between individuals, or exposure to contaminated food or environmental sources (Heymans et al., 2018). Given the diversity of transmission pathways, strict hygiene measures throughout the food production and supply chain are critical for preventing contamination (Huis In ‘t Veld et al., 1994; White et al., 1997).
In China, the detection of Salmonella is governed by the National Food Safety Standard 4789.4 (GB 4789.4), which outlines a culture-based workflow involving nonselective pre-enrichment, selective enrichment, plating on selective agar, biochemical identification, and serotyping according to the Kauffmann–White classification scheme (Diep et al., 2019). Although slide agglutination serotyping is considered the “gold standard,” it is time-consuming, labor-intensive, requires numerous antisera, and demands a high level of technical expertise. Moreover, ambiguous results may occur in some cases (Wattiau et al., 2011). Molecular methods have significantly shortened the time required for pathogen detection—often delivering results in under an hour—but their reliance on expensive instrumentation limits their implementation in field or resource-constrained environments (Quintela et al., 2022; Sousa et al., 2024).
Recombinase polymerase amplification (RPA), first introduced by Piepenburg et al. in 2006, is a novel isothermal amplification method that allows target DNA to be amplified rapidly (within 5–30 min) using recombinase enzymes, strand-displacing polymerases, single-stranded DNA-binding proteins, and sequence-specific primers (Lobato and O’Sullivan, 2018; Piepenburg et al., 2006). Unlike conventional polymerase chain reaction, RPA operates at a constant temperature and does not require a thermal cycler, offering operational simplicity and strong resistance to inhibitors in complex food matrices—making it highly suitable for foodborne pathogen detection (Kersting et al., 2014).
In addition to its rapid amplification, RPA supports multiple endpoint detection formats. Among these, real-time fluorescence detection is particularly advantageous for its sensitivity, ability to monitor amplification kinetics in real time, and potential for semi-quantitative analysis (Leta et al., 2024). Other formats, such as lateral flow strips (LFS), colorimetric, and turbidity-based assays, are beneficial in low-resource settings due to their ease of use and minimal equipment requirements (Li et al., 2019). The choice of detection strategy depends on the intended application and the balance among sensitivity, speed, and accessibility. Real-time fluorescence monitoring provides higher sensitivity and reduces contamination risk due to its sealed system format, whereas LFS methods are more intuitive but susceptible to cross-contamination from aerosolized amplification products (Lobato and O’Sullivan, 2018; Munawar, 2022; Tan et al., 2022).
In this study, we developed a laboratory-based real-time RPA method for the rapid detection of Salmonella in food. Although this method still requires fluorescence detection instrumentation, it minimizes the risk of contamination and achieves high sensitivity, providing a promising tool for efficient screening under controlled laboratory conditions.
Materials and Methods
Bacterial strains
All standard strains were sourced from accredited microbial culture collections and preserved at –80°C in brain heart infusion broth supplemented with 20% (v/v) glycerol. Additional strains used in the inclusivity and exclusivity tests were isolated from food samples during routine surveillance, in accordance with the procedures outlined in the GB 4789.4. All isolates were confirmed using conventional microbiological and biochemical assays following the standard. A full list of strains is provided in Table 1.
The Information of Bacterial Strains
DNA extraction from bacterial cultures
The real-time RPA assay demonstrated tolerance to crude DNA, allowing for simplified extraction. For bacterial cultures, 1 mL of overnight culture was centrifuged at 12,000 × g for 3 min. The pellet was resuspended in 100 μL of lysis buffer (Yi Zhi Technology, China) and incubated at 100°C for 10 min. The resulting lysate was used directly as a DNA template without further purification. RPA reactions were initiated immediately after template preparation to ensure optimal amplification performance.
DNA extraction from enriched food samples
Food sample preparation followed GB 4789.4 with modifications to the pre-enrichment step. Specifically, 25 g (mL) of each sample was homogenized with 225 mL buffered peptone water using a stomacher for 1∼2 min. Instead of static incubation (8∼18 h at 36°C), samples were incubated with shaking at 36°C and 180 rpm for 6 h. Afterward, 1 mL of the enriched culture was centrifuged at 1000 × g for 3 min, and the supernatant was used for DNA extraction using the same lysis protocol as for bacterial cultures.
Design of RPA primers and probes targeting inv A
The inv A gene was selected as the target due to its high conservation across Salmonella species. Ten sets of primers and exo probes were designed using Primer Express software (Applied Biosystems, USA) according to TwistDx guidelines. Primers were 30∼35 nucleotides long, with 40∼60% GC content. All oligonucleotides were synthesized by Sangon Biotech (Shanghai, China).
Real-time RPA assay
RPA reactions were carried out using the TwistAmp™ exo kit (TwistDx, UK) in a 50 μL volume. Each reaction contained 29.4 μL reaction buffer, 8.5 μL ddH2O, 4 μL of primers (10 μM each), 0.6 μL probe (10 μM), and 1 μL DNA template (1∼10 ng). The reaction was initiated by adding 2.5 μL of 280 mM MgOAc, followed by incubation at 37°C for 20 min. Fluorescent signals were recorded in real time.
Inclusivity and exclusivity evaluation
Inclusivity and exclusivity tests were performed according to the National Food Safety Standard—General Principles for the Method Validation of Microbiological Examination (GB 4789.45). Thirty non-Salmonella strains (two per species) and multiple Salmonella serovars (including isolates from food and standard strains) were tested. Details are shown in Table 1.
Limit of detection assay
A standard Salmonella strain was diluted 10-fold in saline from a McFarland 1 suspension. Colony counts were conducted to determine Colony-Forming Unit (CFU). DNA was extracted from each dilution and subjected to real-time RPA to determine the preliminary LOD. To calculate LOD50, DNA at half the preliminary LOD was tested in 24 replicates, following GB 4789.45. The LOD50 was calculated using the following formula (Regulation NHCotPsRoCaSAfM, 2023).
Assessment of the detection consistency between real-time RPA assay and GB 4789.4 in food samples
To further evaluate the practical performance of the real-time RPA assay in food samples, a range of food matrices were selected based on the criteria outlined in “National Food Safety Standard-Limit of Pathogenic Microorganisms in Prepackaged Food” (GB 29921) (Regulation NHCotPsRoCaSAfM, 2021) and Monitoring of Salmonella was conducted on egg products (egg yolk liquid, salted duck eggs, frozen egg liquid); chocolate products (chocolate, chocolate with cocoa substitute); meat products (chicken, sausage, ham); cereal products (cake, biscuits, puffed snacks); and soy products (tofu, dried tofu, natto). Artificially contaminated samples were prepared by spiking Salmonella into sterilized, homogenized food sample solutions at concentrations of 1∼10 CFU/mL. The detection performance of the real-time RPA assay was assessed by comparison with the GB 4789.4 method, with both methods applied to 20 parallel tests for each sample. The relative limit of detection (RLOD) was calculated using formula (2) to evaluate the performance of the real-time RPA assay developed in this study according to GB 4789.45 (Regulation NHCotPsRoCaSAfM, 2023).
Detection of naturally occurring contaminants in pork samples
A total of 90 bulk pork samples were collected from the Beijing market for the detection of Salmonella using both real-time RPA and GB 4789.4.
Statistical analysis
Chi-square (χ2) tests were performed to assess the agreement between the RPA method and GB 4789.4. The formula used was:
Results
Design and initial screening of the real-time RPA primers and probe for Salmonella
Ten sets of primers and exo probes targeting conserved regions of the inv A (GenBank Accession No. NP_461817.1) were designed using Primer Express software. Each candidate set was screened for amplification efficiency, specificity, and inclusivity using representative Salmonella strains and nontarget bacteria. One primer–probe set exhibited superior specificity and broad inclusivity across all tested strains and was selected for subsequent real-time RPA assay development.
The selected set targets a conserved region within the Salmonella enterica genome, corresponding to positions 994,818∼995,045 (GenBank Accession No. CP026569.1), and yields an amplicon of 227 bp. The primer–probe configuration includes a 5′ biotin-labeled forward primer, an unmodified reverse primer, and an exo probe modified with a 5′ Fluorescein amidite (5′ FAM) fluorophore, an internal Tetrahydrofuran（THF） site, a 3′ Black Hole Quencher (3′ BHQ1) quencher, and a 3′ C3 spacer to prevent polymerase extension. Detailed information on the selected sequences and modifications is provided in Table 2.
Sequences of the Selected Primers and Probe for RPA Detection of Salmonella
RPA, recombinase polymerase amplification.
Specificity analysis of the real-time RPA assay
To evaluate the specificity of the primers and probe for the real-time RPA detection of Salmonella, exclusivity and inclusivity assay were conducted. Exclusivity tests were performed using DNA from 30 nontarget bacterial strains (two clone strains of each species), including closely related Salmonella species and other common foodborne pathogens. As shown in Figure 1, positive fluorescence signals were obtained exclusively with Salmonella genomic DNA, while no signals were detected from nontarget bacterial strains.
Exclusivity testing of the real-time RPA assay for Salmonella detection. A total of 30 non-Salmonella strains (two strains per species) were tested. No amplification was observed, demonstrating high specificity of the assay. RPA, recombinase polymerase amplification.
Inclusivity tests used genomic DNA from 30 Salmonella strains, including 8 standard strains and 22 isolates from food samples. As shown in Figure 2, all tested strains produced positive signals, indicating complete inclusivity of the primers and probe for Salmonella. These results confirm the high specificity of the established real-time RPA detection system, with no cross-reactivity observed with nontarget bacterial strains.
Inclusivity testing of the real-time RPA assay for Salmonella detection. Eight different Salmonella serotypes and 22 food-derived Salmonella isolates were tested. All samples yielded positive amplification signals, confirming broad inclusivity of the assay. RPA, recombinase polymerase amplification.
The detection of LOD50 assay
The preliminary LOD of the developed real-time RPA assay was determined using serial 10∼fold dilutions of a standard Salmonella strain (ATCC 17802). The results, shown in Figure 3, the assay detected Salmonella cells at concentrations as low as 200 CFU/mL. To further assess the LOD50, 24 parallel tests were conducted using DNA templates at a bacterial concentration of 100 CFU/mL. Positive signals were observed in 22 out of 24 tests, yielding an LOD50 value of 47 CFU/mL, calculated using formula 1.
Sensitivity testing of real-time RPA assay for Salmonella detection. Preliminary sensitivity testing of the real-time RPA assay for Salmonella detection. Serial 10-fold dilutions of Salmonella cultures were tested to identify the lowest concentration that consistently produced positive results. RPA, recombinase polymerase amplification.
Assessment of the detection consistency between the real-time RPA assay and GB 4789.4 in food samples
To assess the practical performance of the real-time RPA assay in food samples, artificially contaminated samples, including egg products (egg yolk liquid, salted duck eggs, frozen egg liquid), chocolate products (chocolate, chocolate with cocoa substitute), meat products (chicken, sausage, ham), cereal products (cake, biscuits, puffed snacks), and soy products (tofu, dried tofu, natto) were tested using both real-time RPA and the GB 4789.4 method. To enhance sensitivity, a 6-h enrichment step prior to real-time RPA was implemented. The results are summarized in Table 3. RLOD values varied slightly among different food matrices, with egg yolk liquid and chicken showing the highest RLOD values of 2.21, while salted duck eggs, frozen egg liquid, ham, biscuits, tofu, and natto exhibited the lowest RLOD value of 1. Importantly, across all tested food matrices, the RLOD values remained below 2.5, meeting the validation criteria specified in GB 4789.45. This indicates that the developed real-time RPA assay is comparable with GB 4789.4 in terms of detection consistency.
Assessment of the Detection Consistency Between the Real-Time RPA Assay and GB 4789.4 in Food Samples
RLOD, relative limits of detection; RPA, recombinase polymerase amplification.
Detection of naturally occurring contaminants in pork samples
To further assess the practical applicability of the real-time RPA method, 90 bulk pork samples were collected from the Beijing market and tested for naturally occurring Salmonella contamination. The real-time RPA method detected 23 positive samples, corresponding to a positivity rate of 25.6%, while the GB 4789.4 method detected 24 positive samples, yielding a positivity rate of 26.7% (Table 4). The results of the two methods were statistically compared using a chi-square test. As shown in Table 4, the Pearson chi-square test yielded a p-value of 0.865 (p > 0.05), indicating no significant difference between the two methods. These findings confirm that the real-time RPA assay is a reliable alternative to the GB 4789.4 for detecting Salmonella in food samples.
Detection and Statistical Analysis of Salmonella in Pork Samples Using Real-Time RPA and GB 4789.4
The chi-square test was performed to assess the agreement between the two methods.
Zero cells (0.0%) had an expected count <5. The minimum expected count was 23.5.
Fisher’s exact test result: p = 1.000 (two-sided).
All 90 samples were tested in parallel by both methods.
RPA, recombinase polymerase amplification.Choose a building block.
Discussion
Traditional methods for detecting foodborne pathogens, including Salmonella, are often time-consuming, labor-intensive, and reliant on sophisticated laboratory infrastructure. These limitations hinder their practical use in rapid food safety assessments. In contrast, real-time RPA offers distinct advantages such as rapid detection (often within 20 min), operation at constant low temperatures, minimal equipment requirements, and high tolerance to food matrix inhibitors (Chen et al., 2023; Hu et al., 2019; Li et al., 2019; Liu et al., 2017; Zhao et al., 2021). These characteristics make RPA particularly suitable for on-site testing and field deployment (Liu et al., 2024).
Numerous studies have explored the application of RPA for Salmonella detection. For instance, Choi et al. (2016) developed a centrifugal microdevice using direct RPA for detecting foodborne pathogens in milk, achieving multiplex detection within 30 min at a sensitivity of 4 cells per 3.2 µL (Choi et al., 2016). Liu et al. (2017) combined RPA with LFS technology and reported a detection limit of 1.05 CFU/mL in milk and chicken samples after 2-4 h of enrichment (Liu et al., 2017). Li et al. (2019) employed the fim Y gene as a target and achieved 12 CFU/mL sensitivity in pure cultures and 129 CFU/mL in spiked food matrices (Li et al., 2019). Later, Li et al. (2021) used a photochemical colorimetric readout targeting inv A and achieved a postenrichment sensitivity of 3 CFU/mL in milk (Li et al., 2021). Chen et al. (2023) introduced PES membrane-based microfluidic platforms for detecting Salmonella in various matrices such as lettuce and meat, while Liu et al. (2024) developed a phage-based magnetic capture coupled with real-time RPA to enhance sensitivity in milk samples (Chen et al., 2023; Liu et al., 2024).
While these studies demonstrate the technical feasibility of RPA for Salmonella detection, many lack comprehensive validation under regulatory frameworks, limiting their application in routine food safety monitoring. In this study, we addressed this gap by conducting extensive validation of a real-time RPA method according to GB 4789.45, using the culture-based GB 4789.4 method as a reference. The validation parameters included specificity, sensitivity, repeatability, and matrix robustness, evaluated using both artificially contaminated and naturally contaminated samples from diverse food categories.
Our findings demonstrate that the developed RPA method is both accurate and reliable across multiple food matrices. The RLOD values for all tested samples were within 2.5-fold of the GB 4789.4 reference method, meeting the criteria set by GB 4789.45. Furthermore, chi-square analysis showed no statistically significant differences between the two methods in field samples (p > 0.05), supporting the method’s suitability for regulatory screening.
Despite these advantages, several discrepancies were observed. These can be attributed to fundamental differences in detection principles: RPA identifies DNA from both live and dead cells, whereas culture-based methods detect only viable organisms. The shortened 6 h enrichment used in this study may also preferentially amplify fast-growing Salmonella strains, potentially underrepresenting injured or slow-recovering cells. In addition variations in DNA extraction efficiency and matrix components could introduce further bias.
Several limitations should be noted:
Lack of certified reference materials (CRM): No CRMs or internal amplification controls were used, limiting the comparability of results across laboratories. Absence of limit of blank (LoB) determination: LoB was not evaluated, which is important for establishing analytical thresholds in quantitative assays. Lack of comparison with international standards: This study validated the method against GB 4789.4, but did not benchmark it against ISO 6579-1:2017, which could limit its international harmonization and adoption.
Despite the promising performance of real-time RPA in pathogen detection, its global regulatory acceptance remains limited. Currently, the method has not been formally adopted by the Codex Alimentarius Commission as a standardized detection technique. This poses challenges for its application in international trade and regulatory settings. Key barriers include the lack of harmonized validation protocols, insufficient multicenter collaborative data, and limited precedent in cross-border regulatory recognition. To overcome these challenges, future efforts should prioritize the establishment of internationally accepted validation frameworks, conduct interlaboratory studies across a wide range of food matrices, and promote engagement with international standard-setting bodies. Addressing these issues will be critical for facilitating the broader implementation of RPA-based methods in global food safety monitoring systems.
Looking forward, the versatility of the RPA platform presents opportunities for broader applications. Future efforts could incorporate multiplexed RPA systems, visual readouts (e.g., RPA-LFS), or integration with portable microfluidic devices to enable high-throughput and multipathogen screening in the field. These enhancements would further strengthen the utility of RPA in food safety surveillance, outbreak response, and point-of-need diagnostics.
Conclusion
In this study, a rapid detection method for Salmonella based on real-time RPA was successfully developed, achieving detection within 20 min with an LOD50 of 47 CFU/mL. By incorporating a 6 h enrichment step at 36°C with shaking, the method demonstrated reliable performance across various food matrices, including egg products, chocolate products, meat products, grain-based products, and soy products. Notably, the RLOD values for all tested samples were no more than 2.5-fold higher than those observed with the GB 4789.4, meeting the sensitivity requirements outlined in GB 4789.45. Additionally, chi-square analysis revealed no significant difference (p > 0.05) between the real-time RPA and GB 4789.4 when applied to retail bulk pork samples, further supporting the method’s accuracy and consistency.
Authors’ Contributions
L.L.: Designed the primers and probes. L.S.: Writing, reviewing, editing, and supervision. Q.W.: Designed the primers and probes. Y.S.: Collected and preserved strains. Y.Z.: Investigation and data curation. Q.L.: Investigation and data curation. Y.C.: Investigation and data curation. C.X.: Investigation and data curation. D.W.: Investigation and data curation. Z.H.: Investigation and data curation. J.J.: Conceptualization, methodology, research, and project administration.