Sixty Salmonella strains and thirty non-Salmonella strains used in the selectivity testing in this study are listed in Tables 1 and 2, respectively. Salmonella Enteritidis ATCC 13076, Salmonella Typhimurium ATCC 14028, Salmonella Kentucky NCTC 05799 and Salmonella Bredeney NCTC 05731 were used to artificially contaminate chicken samples. These serotypes were chosen for artificial inoculation experiments because they have been associated with poultry-related outbreaks of gastroenteritis in Ireland 2829.

All Salmonella, non-Salmonella strains and chicken samples used in this study were tested for Salmonella by the ISO 6579:2002 European International Standard Method 6 (Figure 1). Biochemical testing was performed using api20E test strips (bioMerieux, Marcy l'Etoile, France). Serotyping was performed according to the Kauffmann-Whyte typing scheme using slide agglutination with standard antisera (Murex, Dublin, Ireland). The real-time multiplex PCR method comprised an 18 h pre-enrichment in buffered peptone water (BPW CM1049, Oxoid, Basingstoke, UK) at 37°C, followed by 6 h incubation in Rappaport-Vassiliadis Soya (RVS) Peptone Broth (CM0866, Oxoid) at 41.5°C, with subsequent DNA extraction (Figure 1).

DNA was extracted from pure cultures of bacteria and chicken samples that had been enriched at 37°C for 18 h in BPW and at 41.5°C for 6 h in RVS. One ml of the RVS broth was collected in an eppendorf tube, centrifuged at 15,339 × g for 5 min and the supernatant discarded. The cell pellets were resuspended in 1 ml phosphate buffered saline (PBS, Oxoid), centrifuged at 15,339 × g for 5 min and the supernatant discarded. DNA extractions were performed on the cell pellet using the DNeasy Blood and Tissue kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. DNA preparations were stored at -20°C until use.

The design of the primers was based on the multiple alignment of Salmonella and non-Salmonella sequences of the individual genes using MegAlign Lasergene software (DNAstar, Madison WI, USA). Sequences were taken from GenBank accession numbers in the literature and nucleotide entries on NCBI website http://www.ncbi.nlm.nih.gov. The gene targets used for primer design and for the real-time multiplex PCR assay and the Salmonella serotypes they detect are included in Figure 2. Four primer pairs were designed using Primer Select DNAStar Lasergene software. The specificity of the primer sequences was tested by homology searches in the nucleotide database (NCBI, nucleotide BLAST (blastn)). TIB Molbiol (Berlin, Germany) designed the TaqMan probes and the probes were labeled with reporter dyes FAM (aceK), YAK (fliC), ROX (sefA) and Cy5 (sdf). Primers and probes were purchased from TIB Molbiol and their sequences are presented in Table 3.

Real-time multiplex PCR reactions were performed with the Rotor Gene 3000 (Corbett Research, Australia) using a total volume of 25 μl, which was contained within a 0.2 ml thin-walled PCR tube. The optimal amplification reaction mixture contained 12.5 μl 2 × QuantiTect Multiplex PCR No ROX mastermix (Qiagen) containing HotStar Taq polymerase, QuantiTect multiplex PCR buffer, dNTP mix including dUTP and 11 mM MgCl₂, a final concentration of 0.4 μM for each primer and probe, 4.5 μl sterile RNase-free water and 2 μl of test or control DNA. Each real-time PCR assay systematically included six control reactions performed in parallel with the test samples. These positive and negative controls used as template, genomic DNA of five Salmonella strains or water respectively. Incubation conditions consisted of an initial denaturation at 95°C for 15 min; followed by 35 cycles of 94°C for 60 sec and 60°C for 90 sec. Fluorescence signals were detected in FAM, YAK, ROX and Cy5 channels. The Rotor Gene 3000 analyzer is capable of detecting four dyes simultaneously in a multiplex assay. The line for calculating the threshold cycle number (C_T) for each channel was assigned to a fixed value intersecting the amplification curves in the linear region of the logarithmic plot. Any sample showing a fluorescence signal above this line was regarded as positive. Each PCR reaction gave a positive or negative result at this threshold line. Samples with C_T values of less than 35 cycles were considered as positive from preliminary optimization experiments.

Sixty Salmonella strains (10 CFU/ml) and thirty non-Salmonella (1,000 CFU/ml) strains were tested by both the traditional cultural method and real-time multiplex PCR assay. The strains were cultured in 10 ml BPW at 37°C for 18 h prior to decimal dilution in Maximum Recovery Diluent (MRD CM0733, Oxoid) to obtain the required inoculation level. One ml of this dilution was then transferred to 9 ml BPW at 37°C for 18 h and a 0.1 ml aliquot of this was added to 10 ml RVS broth at 41.5°C for 6 h. One ml of the RVS broth was then collected in an eppendorf tube for DNA extraction as previously described and subsequent real-time multiplex PCR. A 2 μl aliquot of the DNA extract was used as template in the real-time multiplex PCR assay.

Pre-packed whole chickens were purchased from local supermarkets and were tested for the presence of Salmonella sp. by enrichment of 25 g chicken skin and chicken meat in 225 ml BPW (Oxoid) at 37°C for 18 h, and subsequent plating on Xylose Lysine Desoxycholate agar (XLD CM0469, Oxoid) at 37°C for 18 h. As no Salmonella sp. was detected, the chicken was used for subsequent artificial contamination experiments.

Chicken skin and chicken meat samples (25 g portions) were artificially contaminated individually with four Salmonella serotypes (Enteritidis ATCC 13076, Typhimurium ATCC 14028, Kentucky NCTC 05799 and Bredeney NCTC 05731) at four levels of contamination (0, 1–10, 10–100, 100–1,000 CFU/25 g). The four strains were cultured in 10 ml BPW at 37°C for 18 h prior to decimal dilution in MRD (Oxoid) to obtain the four different levels of contamination. Twenty-five gram portions of chicken samples were inoculated with 1 ml of each level of inoculation, placed in stomacher bags and 225 ml of BPW was added. The mixture was homogenized in a stomacher (Seward, London, United Kingdom) for 1 min and immediately incubated at 37°C for 18 h. The precise numbers of CFU introduced into the chicken skin and chicken meat was determined by plating each level of inoculation on Tryptone Soya Agar plates (TSA CM0131, Oxoid), followed by incubation at 37°C for 24 h. After pre-enrichment in BPW, the cultures were subdivided into two aliquots of RVS broth: one for analysis by the traditional culture method and the second for the real-time multiplex PCR assay. A total of thirty-two chicken skin and chicken meat samples, were analyzed in triplicate by both the ISO 6579 and real-time multiplex PCR methods.

A total of sixty-three 25 g chicken skin samples from naturally contaminated whole chickens, crowns, legs and breasts were obtained from local butchers in the Dublin area. All samples were tested for Salmonella by both the ISO 6579 method and real-time multiplex PCR assay.

Thirty cultures on solid media were received for blind testing from a collaborating group to assess the real-time multiplex PCR assay. One ml of an overnight BPW culture of the blind samples was transferred to 9 ml BPW at 37°C for 18 h and a 0.1 ml aliquot was added to 10 ml RVS broth at 41.5°C for 6 h. One ml of the RVS broth was then collected in an eppendorf tube for DNA extraction as described previously and subsequent real-time mutiplex PCR.

Relative sensitivity, specificity and accuracy were calculated according to the MICROVAL protocol 30. The formulas used for the analysis were:

(1) Relative accuracy: AC=(PA+NA)N×100% MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGaciGaaiaabeqaaeqabiWaaaGcbaGaeeyqaeKaee4qamKaeyypa0tcfa4aaSaaaeaacqGGOaakcqqGqbaucqqGbbqqcqGHRaWkcqqGobGtcqqGbbqqcqGGPaqkaeaacqqGobGtaaGccqGHxdaTcqaIXaqmcqaIWaamcqaIWaamcqGGLaqjaaa@3D61@

(2) Relative specificity: SP=NAN-×100% MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGaciGaaiaabeqaaeqabiWaaaGcbaGaee4uamLaeeiuaaLaeyypa0tcfa4aaSaaaeaacqqGobGtcqqGbbqqaeaacqqGobGtcqqGTaqlaaGccqGHxdaTcqaIXaqmcqaIWaamcqaIWaamcqGGLaqjaaa@39BC@

(3) Relative sensitivity: SE=PAN+×100% MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGaciGaaiaabeqaaeqabiWaaaGcbaGaee4uamLaeeyrauKaeyypa0tcfa4aaSaaaeaacqqGqbaucqqGbbqqaeaacqqGobGtcqqGRaWkaaGccqGHxdaTcqaIXaqmcqaIWaamcqaIWaamcqGGLaqjaaa@39A6@

[ Where;

PA is the positive agreement between culture and PCR methods;

NA is the negative agreement between culture and PCR methods;

PD are the false positives by PCR method;

ND are the false negatives by PCR method;

N is the total number of samples (NA+PA+PD+ND);

N- is the total number of negative results with the reference method (NA+PD) and

N+ is the total number of positive results with the reference method (PA+ND)].

Relative sensitivity is the ability of the alternative method (real-time multiplex PCR) to detect the analyte when it is detected by the reference method (traditional culture method ISO 6579) in the presence of a biological matrix. Relative specificity is the ability of the real-time multiplex PCR to not detect the target organism when it is not detected by the reference method. The relative accuracy is the degree of correspondence between the response obtained by the alternative method and the reference method on identical samples 30.

The sixty Salmonella strains were confirmed as Salmonella positive by the traditional cultural method. These strains, comprising 26 different Salmonella serotypes, all demonstrated positive results for the aceK target in the real-time multiplex PCR assay (Table 1). Testing of the fliC serotype specific target yielded positive results for serotypes Typhimurium and Kentucky only and negative results for all other Salmonella serotypes tested. The sefA serotype specific target was positive for serotypes Enteritidis, Dublin, and Gallinarum and gave negative results for all other Salmonella serotypes. The sdf serotype specific target was specific for all Salmonella Enteritidis strains only and was not amplified in the other Salmonella serotypes tested. The thirty pure culture non-Salmonella strains listed in Table 2 yielded negative results by the traditional cultural method. The non-Salmonella strains also yielded negative results by the four gene targets aceK, sefA, sdf and fliC in the real-time multiplex PCR assay. The selectivity study verified that only Salmonella strains were detected by both methods and that the gene targets sdf, sefA and fliC also correctly identified the serotype where appropriate. The selectivity study also confirmed that no cross reactivity ocurred with the non-Salmonella strains tested.

All artificially contaminated chicken samples tested positive for Salmonella by both the traditional cultural method and real-time multiplex PCR and all un-inoculated control samples tested negative by both methods (Table 4). The Salmonella-positive chicken samples were identified correctly at the serotype level by the real-time multiplex PCR assay. Mean C_T values obtained by the detection channels ranged from 19 to 28 cycles. The C_T values given cannot be used for quantitative interpretations because of the enrichment of each sample before the real-time multiplex PCR assay. The results from the artificially contaminated samples indicated that both methods are sensitive and able to detect to a level as low as 1–10 CFU of Salmonella in 25 g of chicken skin and chicken meat. The results from the multiplex real-time PCR assay were in complete agreement with the ISO 6579 method in three independent artificial inoculation experiments.

Fifteen of the sixty-three naturally contaminated chicken skin samples were Salmonella positive by both methods (PA) and all fifteen samples were identified as either serotypes Typhimurium or Kentucky by the real-time multiplex PCR (Table 5). Fourty-one of the sixty three samples yielded negative results for Salmonella by both methods (NA). One sample was Salmonella-positive by the traditional cultural method and Salmonella negative by the real-time multiplex PCR (ND). Six samples were positive for Salmonella by real-time multiplex PCR and negative by the traditional cultural method (PD). Of these six samples positive for Salmonella only by real-time multiplex PCR, four were identified to the serotype level as Typhimurium or Kentucky as positive signals were detected in the YAK channel as well as the FAM channel. Mean C_T values obtained by the detection channels ranged from 19 to 32 cycles. The C_T values given cannot be used for quantitative interpretations because of the enrichment of each sample before real-time PCR. Relative accuracy, relative sensitivity and relative specificity of the real-time PCR method were determined to be 89, 94 and 87%, respectively.

As part of the blind testing, there was a discrepancy in results with one of the blind samples (BS11) which was identified correctly by the real-time multiplex PCR (Table 6). BS11 was identified as Salmonella-positive in the FAM channel of the real-time multiplex PCR assay and confirmed by biochemical and serology testing. However the colloborating group reported that it was a non-Salmonella isolate.