The bacterial strains used in this study were from the culture collection at Defence Institute of Biodefence Technologies, Mysore, India and are listed in Table 1. All bacterial strains (16 Salmonella and 18 non-Salmonella) used in the inclusivity and exclusivity studies were confirmed biochemically using automated microbial identification system BD Phoenix^™ M50 (BD Diagnostic Systems, Sparks, MD, United States) before use in the assay. The bacterial strains were maintained as glycerol stocks and cultured overnight in Brain Heart Infusion broth (Himedia, India) with shaking (160 rpm) at 37°C.

Viable plate counts of Salmonella from overnight culture or artificially spiked food enrichments were obtained by serially diluting the cultures in 1X Phosphate Buffered Saline (PBS) and plating in Salmonella-Shigella (SS) agar (Himedia, India). Following overnight incubation for 12–16 h at 37°C, Salmonella appeared as colourless colonies with black centres on SS agar plates.

Thermal lysates were prepared from the bacterial cultures or food enrichments for use in real-time PCR assay. Briefly, 1 mL of overnight culture (12–18 h) or food enrichments were pelleted and resuspended in 300 μL of sterile distilled water, boiled at 100°C for 15 min to lyse the cells, and centrifuged at 10,000 rpm for 5 min to remove cell debris. The supernatant (thermal lysate) was transferred to a new Eppendorf tube. The thermal lysates thus prepared were stored at −20°C and used as template DNA (1 μL) in the real-time PCR assay as described in Section 2.3. The same method for thermal lysate preparation was followed for enriched food samples in Sections 2.7 and 2.9 as well.

Artificial contamination experiments were performed with both processed (retort processed, ready-to-eat (RTE) chicken pulao and chicken biriyani) and raw (frozen chicken and egg) foods. RTE chicken pulao and chicken biriyani were acquired from the production facility at Defence Institute of Biodefence Technologies, Mysore, India. Frozen chicken and raw eggs were obtained from local supermarkets in Mysore, India.

A comparative study was conducted to evaluate the efficacy of the invJ real-time PCR method against the conventional ISO 6579-1:2017 method for detection of Salmonella contamination in natural (raw and processed) food samples. The study encompassed 52 food samples and included raw chicken (n = 8), frozen chicken (n = 2), raw meat (beef, pork and mutton) (n = 17), raw egg (n = 5), raw milk (n = 10), processed food (n = 6) and fresh fruits and vegetables comprising carrots, tomatoes, radish, and lettuce (n = 4). All food samples, except the processed ones, were sourced from local supermarkets in Mysore and stored at 4°C until testing. The processed food samples were retort processed and were sourced directly from the DIBT facility.

Complete coding sequence of invJ gene was retrieved from the genome of Salmonella enterica subsp. enterica serovar Typhimurium str. LT2 (Genbank accession ID NC_003197.2:3034342–3035253). Using the above sequence as query, BLAST search was performed and consequently, a total of 122 nucleotide sequences of Salmonella invJ gene from various serovars were downloaded from NCBI BLAST¹ results. The nucleotide sequences were aligned using the multiple sequence alignment program CLUSTAL-W2 using default parameters and the alignments were examined further by alignment explorer (MEGA-11) before phylogenetic analysis. The alignment file was saved with an extension “.mas” for the phylogeny construction. The phylogenetic tree was constructed with MEGA-11 using neighbor-joining method (Saitou and Nei, 1987) with 1,000 bootstrap replicates. The evolutionary distances were computed using the maximum composite likelihood method (Tamura et al., 2004) and are in the units of the number of base substitutions per site. The number of synonymous substitutions per synonymous site (dS) and the number of non-synonymous substitutions per non-synonymous site (dN) averaging over all sequence pairs was calculated with Nei–Gojobori model (Nei and Gojobori, 1986) using MEGA-11.

The multiple sequence alignment was used to design invJ oligonucleotide primers with degenerate nucleotide positions and TaqMan^® MGB (minor groove binder) probe for use in the Salmonella real-time PCR assay. Nucleotide variations in the forward and the reverse primer binding regions among the 122 invJ gene sequences from diverse serovars were accounted for, using degenerate bases such as R and Y. Sequences of the primers and probes used for the real-time PCR assay are provided in Table 2. To assess primer specificity, a virtual PCR was conducted using a set of 690 nucleotide sequences derived from BLAST results of the invJ gene. This dataset included representative sequences from Salmonella bongori (17 strains), S. enterica subsp. salamae (40), S. enterica subsp. arizonae (14), S. enterica subsp. diarizonae (9), and S. enterica subsp. houtenae (11), using the tool available at https://www.bioinformatics.org/sms2/pcr_products.html. The High Purity Salt Free^®-purified Primers (Eurofins Genomics, India) and HPLC purified TaqMan^® MGB probe (431064, Applied Biosystems^™, United States) with FAM fluorophore and NFQ-MGB quencher were custom synthesized for the assay.

The real-time PCR assay was optimized using Salmonella enterica Typhimurium ATCC 14028. The final 10 μL reaction volume comprised of: 1X TaqMan^® Mastermix (Premix Eq Taq RR390A, Takara, Japan), 0.5 μM Forward Primer, 0.5 μM Reverse Primer, 0.5 μM TaqMan^® MGB Probe and 1 μL of DNA template. Reaction conditions were optimized as: denaturation at 95°C for 30 s, followed by 40 cycles of denaturation at 95°C for 3 s and 57°C for 35 s annealing/extension. Sterile distilled water was used as no-template control for the experiments. The real-time PCR was performed in QuantStudio^™ 3 Real-Time PCR System (Applied Biosystems, United States). All analyses were performed in QuantStudio^™ Design and Analysis Software v1.4.3.

Specificity of the real-time assay was evaluated with the inclusivity and exclusivity test strains provided in Table 1. Thermal lysate prepared from the all the bacterial strains, as described in Section 2.1, was used as DNA template to determine the specificity of the assay. The assay was carried out as per the optimized protocol in Section 2.4 in duplicates. Sterile distilled water was used as no-template negative control.

Amplification efficiency and limit of detection (LOD) of the real-time PCR assay were determined using 10-fold serial dilutions of Salmonella enterica ser. Typhimurium ATCC 14028 bacterial culture ranging from 10¹–10⁸ CFU/mL, which were confirmed by plating in SS Agar. Thermal lysates prepared from the bacterial culture dilutions using the procedure described in Section 2.1 were used as DNA template. The assay was carried out in triplicates and sterile distilled water was used as no-template negative control. The experiment was repeated independently two times. The standard curve was constructed by plotting mean Ct values against bacterial culture concentrations [expressed in log₁₀(CFU/mL)] while amplification efficiency was calculated with the equation E = (10^−1/slope) −1 using the slope of the standard curve.

Four food types-RTE chicken biriyani, RTE chicken pulao, frozen chicken and raw egg were used for the assessing the LOD of real-time PCR in artificially contaminated food samples as described in Section 2.2. All food samples were confirmed to be Salmonella-free by standard microbiological methods. Prior to testing, the frozen chicken was allowed to thaw at refrigeration temperatures for 1 h. The background flora level of the samples was evaluated by Aerobic Plate Count and the results are presented in Supplementary Table 1. Total aerobic counts (TAC) were measured using Petrifilm^™ Aerobic Count Plates (3 M^™, United States). Following homogenization of a 10 g food sample in 90 mL of sterile diluent (0.9% NaCl), 1 mL of the homogenate was drawn, serially diluted, and 1 mL of each dilution was inoculated onto Petrifilm^™ plates according to the manufacturer’s instructions. After incubation at 35°C for 48 h, red colonies were counted to determine the TAC.

For the artificial contamination experiments, the food samples were artificially spiked with Salmonella Typhimurium ATCC 14028 such that the net inoculum was 10⁰–10³ CFU/mL of food homogenate. Artificially spiked food sample (25 g) incubated overnight at 4°C was aseptically transferred into a homogenizer bag (Nasco Sampling Bag, Himedia, India) and homogenized in 225 mL buffered peptone water (BPW) (Himedia, India) using a lab blender (BagMixer 400, Interscience, France) at 6 strokes/s for 2 min. The homogenates were kept for enrichment at 37°C and shaking (160 rpm). At 6 h, six technical replicates of 1 mL enriched samples were drawn from the homogenates, pelleted at 10,000 rpm for 5 min, and were subsequently prepared for real-time PCR testing. Three samples each were prepared by two different methods-thermal lysis (Section 2.1) and kit-based DNA extraction (Qiagen DNeasy Blood and Tissue Kit, 69506, Germany) with slight modifications. Briefly, the pelleted 1 mL enriched samples were resuspended in 180 μL ATL buffer and 5 μL Proteinase K provided with the kit for lysis. After 1 h, 200 μL of AL buffer was added to the samples and pelleted at 10,000 rpm for 2 min. Absolute ethanol (200 μL) was added to the supernatant followed by addition of the entire mixture to spin column. DNA was eluted in 100 μL of MilliQ water and was used as template for real-time PCR (Section 2.4).

Limit of Salmonella detection in various food matrices after 6 h enrichment was obtained from the Ct values of real-time PCR. Viable Salmonella count after 6 h enrichment in the spiked samples was enumerated by plating the serial dilutions (1X PBS) in SS agar and used to understand the effect of food matrices in Ct values. The experiment was performed independently three times with the food matrices.

Detection limit of Salmonella in the presence of natural background flora in food was determined using background flora from frozen chicken. The chicken sample was confirmed to be Salmonella-free by standard microbiological methods prior to use for experiments. Chicken sample (25 g) after homogenization in 225 mL BPW, were kept for incubation for 6 h at 37°C with 160 rpm shaking. Chicken homogenates (900 μL) drawn from 6 h enrichment (triplicates) were used as the source of background flora. To each replicate, 100 μL of Salmonella enterica ser. Typhimurium ATCC 14028 culture dilutions (10⁸–10¹ CFU/mL) was added, mixed and pelleted for thermal lysate preparation. The thermal lysate was prepared as per Section 2.1. The supernatant from the thermal lysate was used for real-time PCR for assessing the detection limit of Salmonella in the presence of natural background flora from the chicken homogenate. One independent experiment was performed. Total aerobic count of the chicken homogenates at 6 h was assessed using Petrifilm^™ aerobic count plates as described above in Section 2.7.

Presence of Salmonella in 52 natural food samples (raw and processed) (Section 2.2) was examined by the real-time PCR assay and compared with the ISO 6579-1:2017 method of Salmonella detection. For both methods of detection, 25 g of food sample was homogenized with 225 mL of buffered peptone water (BPW). The samples were enriched at 37°C with shaking at 160 rpm. At 6 h, 1 mL enriched samples were drawn in triplicate for real-time PCR analysis. Thermal lysate was prepared from these drawn samples as described in Section 2.1, and was subsequently analysed by real-time PCR.

Following sample collection for real-time PCR, the remaining enrichment cultures were incubated overnight to proceed with detection according to the standard ISO 6579-1:2017 protocol. This method was followed exactly up to the selective plating step, where samples were streaked onto Xylose Lysine Deoxycholate (XLD) agar and Hektoen Enteric (HE) agar. Presumptive colonies in XLD and HE agar were then sub-cultured onto non-selective Brain Heart Infusion (BHI) agar and incubated overnight at 37°C. The resulting colonies were subjected to automated biochemical identification using the BD Phoenix^™ M50 system with NID panels, following the manufacturer’s instructions.

The real-time PCR data obtained from the instrument were analyzed using QuantStudio Design and Analysis Software v1.4.3. The cycle threshold (Ct) values of amplification were determined by the software. Construction of standard curve and other data analysis was done using Microsoft Excel, 2007. Relative sensitivity, specificity, and diagnostic accuracy describe the ability of the PCR method for detection of analyte as compared to the reference method. The relative sensitivity, specificity, and accuracy were calculated as described:

where TP, true positive, FP, false positive, TN, true negative, FN, false negative as compared to the ISO 6579-1:2017 method.

BLAST analysis of the invJ gene sequence (Genbank accession ID NC_003197.2:3034342–3035253) with the default parameters showed hits with other Salmonella invJ nucleotide sequences with 84 to 100% sequence identity and 100% query coverage. The BLAST hits with other Enterobacteriaceae strains showed less than 15% query coverage indicating the absence of invJ homologs in Enterobacteriaceae strains. Multiple sequence alignment (MSA) of the 122 invJ gene nucleotide sequences from Salmonella strains obtained from NCBI was used to determine the conserved nature of the invJ gene among Salmonella strains. Among the 122 sequences that were analyzed, the Salmonella invJ gene was found to have 676 conserved sites and 335 variable polymorphic sites (289 parsimonious informative sites) that resulted in 195 conserved, and 141 variable amino acid polymorphic sites (121 Pi sites). The type of selection acting on Salmonella invJ gene was understood using dN/dS values. The dN-dS was observed to be −0.072, while dN/dS ratio was observed as 0.27 which is indicative of purifying selection. The rate of non-synonymous substitution (dN −0.027) was 3.7-fold lesser than of synonymous substitution (dS −0.099). This evolutionarily conserved nature of Salmonella invJ gene qualified it as a viable molecular target for specific detection of Salmonella in food samples.

Additionally, phylogenetic analysis of invJ by MEGA-11 revealed a phylogenetic tree with five clusters based on the species and subspecies variation in Salmonella which has been highlighted in Figures 2A,B using multi-colored symbols.

Degenerate real-time PCR primers and the TaqMan^® probe for the specific detection of Salmonella were designed targeting the conserved regions identified by multiple sequence alignment of Salmonella invJ gene sequences. The forward, reverse primers and probe were derived from three conserved regions viz. i.e., 802–819 bp, 934–952 bp and 866–881 bp, respectively, to provide an amplicon of 151 bp and spanning 802 bp to 952 bp region of Salmonella invJ gene (Figure 2C). The highly conserved 866–881 region without any nucleotide variability among serovars, was used for designing the TaqMan^® MGB probe. To sum up, the primers FW: 5′-CARCGCTGGGGAAATGAC-3′, RV: 5′-GATCRTCTCGCRYCAGGTG-3′, and the invJ gene specific probe 6 FAM-TACCGTCAAATACGCA-MGB NFQ were designed, custom synthesized and used for optimizing the real-time PCR assay conditions for the specific detection of Salmonella in food samples. Virtual PCR results indicate that all assessed 690 Salmonella invJ sequences, including S. bongori, S. enterica subsp. enterica, S. enterica subsp. houtenae, S. enterica subsp. arizonae, S. enterica subsp. diarizonae and S. enterica subsp. salamae (Supplementary Data), showed an amplicon of 151 bp.

Specificity of the invJ real-time PCR involving primers and probe designed in this study was determined using 34 bacterial test strains [Salmonella: 16; non-Salmonella: 18 (Table 1)]. Amplification was observed from all the Salmonella strains with Ct values ranging from 16.7 to 21.6 for the invJ target while Ct values of S. Typhimurium strains ranged from 16.8 to 19.3 in the real-time PCR assay (Table 1 and Figure 3A). No amplification was observed for the non-Salmonella strains used in the assay (Figure 3B). The specificity studies confirmed that only Salmonella strains were detected by the assay and no cross-reactivity was observed with non-Salmonella strains used in the study.

The analytical limit of Salmonella detection by the real-time PCR method with respect to CFU/mL count was determined using logarithmic dilutions of S. Typhimurium (ATCC 14028). A linear relationship was observed over the range of 10² to 10⁸ CFU/mL with Ct values ranging from 17.2 to 38.4 (Figure 4A and Table 3). The resulting equation for the straight line was −3.5857x + 45.714, which showed a good correlation between Ct values and Log₁₀ CFU/mL with an R² value of 0.9975 (Figure 4B). Results demonstrated that the LOD was 10² CFU/mL of Salmonella. The efficiency of amplification determined using the slope of the standard curve was found to be 90%.

The limit of direct Salmonella detection (LOD) from food matrices was determined using 6 h enrichments obtained by spiking Salmonella Typhimurium ATCC 14028 in food (initial inoculum—10⁰–10³ CFU/mL). Irrespective of the food matrices used, Salmonella could be detected from all the culture dilution enrichments with detection probability 1 (100%) even at the lowest level of initial inoculum (10⁰ CFU/mL).

In food homogenates viz. egg, frozen chicken, RTE chicken biriyani, and RTE chicken pulao having an initial inoculum of 10³ CFU/mL, the 6 h enrichment yielded Ct values of 15.6, 23.7, 15.4, and 14.6, respectively using kit extracted DNA (Table 4). The Ct values for other 6 h enrichments are shown in Table 4. No amplification was observed in the no-template control. It is observable that thermal lysate derived DNA showed a higher Ct value than kit derived DNA (Table 4). In all cases, the replicates showed uniform detection leading to a probability of detection (p/n) value equal to 1. The Salmonella CFU counts after 6 h enrichment was determined by plating in SS agar. At the end of 6 h of enrichment, ~10⁵ CFU/mL Salmonella was observed in all food matrices that were initially spiked with the lowest 10⁰ CFU/mL of food, except RTE chicken biriyani. Alternatively, 10³ CFU/mL spiking multiplied to 10⁸ CFU/mL (or 10⁷ CFU/mL for RTE chicken biriyani). Collectively, at the end of 6 h enrichment the CFU/mL values ranged from 10⁵–10⁸ CFU/mL for the initial inoculum levels of 10⁰–10³ CFU/mL, respectively, in all food matrices except in RTE chicken biriyani (Table 4). RTE chicken biriyani had the largest dynamic range of Ct values ranging from 15.4 to 27.6 for kit derived DNA as well as 19.1 to 32.4 for thermal lysate derived DNA, as compared to other food matrices (Table 4C).

The limit of Salmonella detection by real-time PCR in the presence of background flora was determined by mixing Salmonella Typhimurium ATCC 14028 culture dilutions (10⁰–10⁷ CFU/mL) with 6 h enrichment culture obtained from frozen chicken (Figure 5). The total aerobic plate count from the 6 h enrichment culture was 9 × 10⁷ CFU/mL (Table 5). In the presence of this high background flora, the LOD was estimated at 10⁴ CFU/mL with Ct values ranging from 27.1 to 36.6 for 10⁷ to 10⁴ CFU/mL Salmonella culture dilutions (Table 5) using thermal lysate derived DNA. No Ct value was observed in the No-template control. It was also observed that the linearity of the Ct values is affected by the presence of high background flora level as indicated by the standard curve that yielded an R² value of 0.51 (Supplementary Data).

A total of 52 natural food samples (raw and processed) were analyzed by the real-time PCR assay and it was compared with ISO 6579-1:2017 for Salmonella detection. Out of 52 samples, 14 samples were found to be positive for Salmonella by real-time PCR assay after 6 h enrichment (Section 2.8). Similar results were observed with the automated microbial identification as well (Table 6). Raw egg and raw meat yielded a Salmonella positivity rate of 80 and 64.7% respectively, followed by raw milk and raw chicken (40 and 37.5%), which was corroborated by the automated microbial identification (Table 6). Processed food, fresh vegetables and frozen chicken was found to be negative for Salmonella. It is to be noted that the selective agar plating step in the ISO 6579-1:2017 method yielded Salmonella-like colonies in 100% of raw chicken, 50% of frozen chicken, 88.2% of raw meat, and 100% of fresh fruits and vegetables. However, the subsequent automated microbial identification step provided the actual result revealing 37.5, 0, 64.7, and 0% Salmonella contamination, respectively, in the samples (Table 6). The real-time PCR assay was found to have 100% relative specificity, 100% relative sensitivity and 100% relative accuracy as shown in Table 7.