The “Rapid’Salmonella” Method: Estimation of the Limit of Detection for Salmonella Strains Typhimurium and Enteritidis Isolated from Frozen Poultry Meat

As a part of evaluation the surveillance system of Salmonella in frozen imported poultry meat into Jordan, we conducted a study to estimate the limit of detection (LOD50% and LOD95%) of Salmonella Typhimurium and Salmonella Enteritidis based on chromogenic media of Rapid’Salmonella method. Salmonella-free chicken meat samples was inoculated with 1 to 100 CFU of 11 wild strains that originated from frozen imported poultry meat and 2 reference strains. In the experiment, the observed lowest concentration for Salmonella Typhimurium and Salmonella Enteritidis using Rapid’Salmonella method were from 1 to 50 CFU/25 g. Based on these results, probability of detection (POD) curve was estimated according to the model described in EN ISO 16140-4. From the estimated POD functions, the LOD50% and LOD95% was determined for the Rapid’Salmonella method. The LOD50% of the different strains varied from 0.9 to 21.2 CFU/25 g. The two reference strains and 9 wild strains had a LOD50% less than 2 CFU/25 g, one wild strain of Salmonella Enteritidis had a LOD50% of 6.8 CFU/25 g and another one had a LOD50% of 21.2 CFU/25 g. The majority of Salmonella strains has a LOD50% of 1-4 CFU/25 g in poultry meat, but also that there are some Salmonella strains which will first be detected at 10 CFU/25 g and higher.


Introduction
Non-Typhoidal Salmonella (NTS) including Salmonella Typhimurium and Enteritidis are the most frequent causes of foodborne salmonellosis in the Middle East and North Africa (MENA). In the MENA countries including Jordan, the presence of NTS strains Typhimurium and Enteritidis in domestic and imported poultry meat is one of the main concerns of the food safety authorities (Malaeb, Bizri, Ghosn, Berry, & Musharrafieh, 2016;Nimri, Abu AL-Dahab, & Batchoun, 2014;Osaili et al., 2014).
According to the Jordan Food and Drug Administration (JFDA) guidelines, all imported frozen poultry meat at customs ports requires a scheduled sampling to test for Salmonella strains Typhimurium and Enteritidis (Jordan Food and Drug Administration, 2015). From each batch, one sample is collected. A batch of poultry meat is equivalent to one type of product produced on a specific date in one establishment (Jordan Food and Drug Administration, 2015). Sample units are transported to the JFDA laboratories where 25 g is apportioned, thawed and analyzed according to the Rapid'Salmonella method (Bio-Rad, Marnes-la-Coquette, France). This method was introduced in 2015 as an alternative to the reference method "ISO 6579:2017" for rapid detection of Salmonella spp. including strains of Salmonella Typhimurium and Enteritidis and followed by real time (RT)-PCR (Maurischat, Baumann, Martin, & Malorny, 2015) for confirmation and identification.
In 2015, with the use of Rapid'Salmonella method, 29 batches of poultry meat (representing approximately 200 tons) out of 3,109 examined (representing approximately 50,000 tons) were rejected at the Jordan border because they were found positive for Salmonella strains Typhimurium and Enteritidis. However, it is expected that a number of contaminated batches of poultry meat were not detected by the method used. The likelihood to detect a contaminated batch depends on the actual occurrence of Salmonella in the batch (prevalence of contaminated items and concentration of Salmonella in those items) and limit of detection (LOD) of laboratory method used. This likelihood can be described by the probability of detection (POD) function (Wilrich & Wilrlich, 2009), and is a useful quantitative measurement of the overall performance of a surveillance program.
However, the LODs of the Rapid'Salmonella method for Salmonella strains Typhimurium and Enteritidis in frozen poultry contaminated with relevant field strains for poultry meat imported to Jordan has never been studied.
The objective of this study was to determine the lowest number of cells of different Salmonella Typhimurium and Enteritidis strains isolated from imported frozen poultry meat that can be detected using the commercial laboratory method Rapid'Salmonella. This was done in a spiking experiment using a serial dilution of concentration of several field strains. Subsequently, a probability function of detection (POD) was fitted to the observed LOD values, from where LOD 50% and LOD 95% was determined.
The overall aim of the border control is to protect the consumers against salmonellosis attributable to imported poultry meat. The POD functions estimated in this study will be important input for subsequent assessment of the border control using quantitative risk assessment. In addition, the experimental setup and the estimation of the POD function can be used when assessing the effect of improved laboratory methods and sampling strategies used at border control.

Method
In the spiking experiment known numbers of different strains of Salmonella Typhimurium and Enteritidis were duplicate inoculated on Salmonella-free chicken meat samples, and subsequently the samples were analyzed using the Rapid'Salmonella method for detection. The observed data from the spiking experiment (concentration and positive/negative) was used for estimating the POD, LOD 50% , and LOD 95% .

Chicken Meat Samples
The Salmonella-free chicken meat samples (whole chicken carcasses and boneless chicken breast fillet with skin) used in this study were equally brought from Denmark and Brazil in 2018. The European Commission regulation 2018/307 declared Danish broiler meat as Salmonella-free. Before conducting the study, the Brazilian chicken meat samples were collected from a batch of boneless breast fillet chicken meat with skin of 2.5-kg packages. From this batch, five samples of 25 g were collected and tested for the presence of Salmonella using the ISO 6579:2017 method (Anynomous, 2017). All five samples were negative.
The Salmonella-free chicken meat was cut into 25-g portions representing samples, and these samples (n=132) were stored at -18°C for a maximum of 30 days. The samples were thawed at 4°C for 24 h before use.

Bacterial Strains and Inoculum Preparation
The samples were spiked with 13 strains of Salmonella Typhimurium and Salmonella Enteritidis from the JFDA surveillance collection (see Table A1 in Appendix). These strains were grown in nutrient broth (Oxoid, Basingstoke, UK) and incubated at 37°C ± 1°C for 24 h to obtain expected bacterial concentrations 10 9 CFU/ml. Using 10-ml volumes, serial dilutions established five levels with expected bacterial concentrations of 100, 50, 10, 5, and 1 CFU/ml. The number of cells in each established level of inoculation were enumerated and recorded to calculate the initial bacterial concentrations as described below.

Total Count of Inocula
One ml of each of the above five levels of established serial dilutions was poured on duplicate plates of aerobic Plate Count Agar (PCA, Scharlau, Barcelona, Spain). These two plates were used for counting the total count of the bacteria after incubation at 37°C ± 1°C for 48 h. Colony counts from 0-250 CFU/plate were used for estimating the total viable count. The same person throughout the study performed the counting. The estimated total count was used to determine the apparent concentration of Salmonella in the 25-g of spiked samples (see Table 1). The total count of bacterial concentration in each inoculum was estimated according to the formula: Note. Plate=1...n is the plates with colony numbers between 0-250 CFU for a specific strain.

Spiking Samples
For each dilution and strain, we performed duplicate spikes on two separate chicken meat samples. Each 25-g sample was spiked individually with 1 ml of each of the 13 strains of Salmonella Typhimurium and Salmonella Enteritidis with the established five levels of expected bacterial concentrations 100, 50, 10, 5, and 1 CFU/25 g. In addition, two samples were not spiked and served as negative controls. All samples were analyzed for Salmonella presence as described below.

Determining the Limit of Detection
The observed lowest concetration that was detected from the inoculation experiment by comparing the observed results in terms of presence/absence of growth with the number of Salmonella in the 25-g spiked samples. We used two measures for "the number of Salmonella in the spiked samples": the number based on the total count as calculated in formula 1, and the expected number of bacteria based on the dilution series. Based on the qualitative results, a probability function for detecting the strain at different concetrations (d) was estimated under the statistical analysis was carried out by application of the EXCEL sheet PODLOD.xls (Wilrich & Wilrlich, 2009). This model described in EN ISO 16140-4, using a program in Excel, freely available on the Internet, version 9, dated 2017-09-23 (Anonymous, 2016). The LOD p was defined as the lowest contamination level (CFU/25 g) where the Rapid'Salmonella method is positive with specified probability, p.
Based on the function, the LOD 50% and LOD 95% was calculated for each strain, specifying the lowest concentration of Salmonella in the meat matrix that can be detected with a probability of 50% or 95%, respectively. The LOD 50% and LOD 95% with confidence limits for each strain were calculated (Table 2 and Table  3). Finally, The obtained estimates is used to express the POD function as p (d) of wide range of assumed known contamination d according to inoculated levels from 0 to 100 CFU/25 g, and as the following formula: Where A 0 is the sample size =25-g, Fi is the matrix effect that is < 1 (estimated the deviation of the POD curve from the ideal POD curve that has estimated LOD =1 by application of the EXCEL sheet PODLOD.xls (Wilrich & Wilrlich, 2009)), and d the contamination in CFU/25 g.

Concentrations of Bacterial Inocula
The validity of the estimted LOD p is strongly depending on that the number of bacteria in the inocula is known.
In this study, we performed the estimation of the probability function using both expected number of bacteria and apparent number of bacteria. The total counts of bacteria were about 50%-100% of the expected bacterial concentration that was established for Salmonella pure cultures and spiked chicken meat samples (see Table 1). The relatively low apparent counts may be due to bacterial clustering features, and some organisms may have been stressed and died during handling of the sample (Capozzi, Fiocco, Amodio, Gallone, & Spano, 2009;Sutton, 2011). Most likely, the actual bacterial concentrations in this study were in-between the total apparent concentrations and the expected bacterial concentrations based on the dilution series. Accordingly, the observed lowest concentration was assigned to both apparent and expected bacterial concentration. The differences bewteen LOD p based on apparent and expected bacterial concentration were negligible (see Table A3 and Table  A4 in Appendix).

Limit of Detections
The observed lowest concentration for Salmonella Typhimurium and Salmonella Enteritidis using Rapid'Salmonella method were from 1 to 50 CFU/25 g for spiked chicken meat samples (Table 1). The estimatd LOD 50% for Salmonella Typhimurium were from 0.9 to 1.8 CFU/25 g (Table 2) and for Salmonella Enteritidis were from 0.8 to 21.2 CFU/25 g ( Table 3). The LOD 95% for Salmonella Typhimurium were from 3.7 to 7.6 CFU/25 g (Table 2) and for Salmonella Enteritidis were from 3.7 to 91.7 CFU/25 g ( Table 3). The LOD 50% combined results for Salmonella Typhimurium and Salmonella Enteritidis were 1.1 CFU/25 g (95% CI: 0.6-1.8 CFU/25 g) and 4.2 CFU/25 g (95% CI: 2.3-7.3 CFU/25 g), respectively, indicating a significant difference in general between Salmonella Typhimurium and Salmonella Enteritidis detection level. realistic situations where the bacteria in the imported meat has been frozen, and thereby, they are stressed and injured causing a lag in the growth. However, due to controlled conditions in the enrichment, the lag-phase can be assumed to last for only 1-2 h (Oscar, 1998), which is equivalent with 3-6 generations of growth for Salmonella in optimal growth conditions in broth at 37 ℃, assuming a generation time of 20-30 min. The loss of 3-6 generations due to lag-phase is proportionally a low number compared with the approximately 100 generations that can be expected for Salmonella in 18-20 h given experimental conditions (Oscar, 1998). Thereby, the effect of using non-frozen isolates on the estimated LOD p is expected to be minor. Contemporary, by not freezing the samples after spiking, we know how many viable bacteria that are actually present in the sample, which strengthens the validity of the estimated LOD p .
In the spiking experiment, one strain of each Salmonella Typhimurium and Enteritidis from the ATCC and 11 from the JFDA's isolates in imported frozen poultry meat were used. Even though, the estimated LOD p cannot be generalized to all strains of Salmonella in all types of food items, the estimated LOD p in this study indicate the LOD p that can be expected when using of Rapid'Salmonella method at the border control of frozen poultry meat.
In the spiking study performed by AFNOR, they used 152 Salmonella strains at contamination levels between 5-25 CFU/sample, but only 5 of those represent Salmonella Typhimurium and Enteritidis spiked into poultry meat (ADRIA Development, 2017). In our study, 12 out of 13 Salmonella strains are at observed lowest concentration ≤5 CFU/25 g, which is in alignment with AFNOR spiking study finding approximately 70% of tested Salmonella strains with the same observed lowest concentration (ADRIA Development, 2017). In addition, they compare the performances of reference method and alternative method by estimating LOD 50% , which were between 0.1-5.6 and 0.1-1.8 CFU/25 g, respectively (Norli & Nielsen, 2018). In our study, the LOD 50% for Salmonella Typhimurium and Salmonella Enteritidis were between 0.6-7.3 CFU/25 g, which is in alignment with AFNOR validation certification (Norli & Nielsen, 2018).
There are many surveillance programs that employ rapid immunoassays and PCR methods instead of conventional culture methods for detecting Salmonella in poultry meat to cope with the enormous volume of samples (Brooks, Lutze-Wallace, Devenish, Elmufti, & Burke, 2012;Cheung & Kam, 2012;Hitchins, 2012;Tomás Fornés, McMahon, Moulin, & Klijn, 2017). The observed detection level of the conventional pre-enrichment step that directly coupled with the PCR methods is 100-200 CFU/25 g of Salmonella Typhimurium and Enteritidis in poultry meat (Mohd Afendy & Son, 2015; Paião et al., 2013;Siala et al., 2017). Compared to this, the Rapid'Salmonella method can be considered relatively sensitive for most strains of Salmonella Typhimurium and Enteritidis.

Conclusion
In the spiking experiment, we found that the observed level of detecting Salmonella Typhimurium and Salmonella Enteritidis from poultry meat using the commercial Rapid'Salmonella method varies between 1 and 50 CFU/25 g. The most natuarally wild Salmonella strains and laboratory-adapted ones have LOD 50% between 1 and 4 CFU/25 g. However, due to the studied Salmonella strains are limited in numbers and serotypes, their results can't be generalized to all Salmonella spp. without caution. Future studies should focus on including more serotypes that representing different countries of origin and interlaboratory comparison for robustness.
Referring to the POD curves in figure 1, it can be concluded that most studied Salmonella strains has a LOD 50% of 1-4 CFU/25 g in poultry meat, but also that there are some Salmonella strains which will be detected at concentrations around 10 CFU/25 g and higher. The concetration in the matrix, which gives a 95% likelihood for detection (LOD 95% ) was for most Salmonella Typhimiurm strains around 5 CFU/25 g, whereas for most Salmonella Enteretidis strains it was around 50 CFU/25 g.
This study is the intial step in evaluating and optimizing current Salmonella surveillance of poultry meat in the MENA region.