|Home | About | Journals | Submit | Contact Us | Français|
Certified Campylobacter-free poultry products have been produced in Denmark since 2002, the first example of fresh (unprocessed and nonfrozen) chickens labeled “Campylobacter free.” This success occurred partly through use of a 4-hour gel-based PCR testing scheme on fecal swabs. In this study, a faster, real-time PCR approach was validated in comparative and collaborative trials, based on recommendations from the Nordic system for validation of alternative microbiological methods (NordVal). The comparative real-time PCR trial was performed in comparison to two reference culture protocols on naturally contaminated samples (99 shoe covers, 101 cloacal swabs, 102 neck skins from abattoirs, and 100 retail neck skins). Culturing included enrichment in both Bolton and Preston broths followed by isolation on Preston agar and mCCDA. In one or both culture protocols, 169 samples were identified as positive. The comparative trial resulted in relative accuracy, sensitivity, and specificity of 98%, 95%, and 97%, respectively. The collaborative trial included nine laboratories testing neck skin, cloacal swab, and shoe cover samples, spiked with low, medium, and high concentrations of Campylobacter jejuni. Valid results were obtained from six of the participating laboratories. Accuracy for high levels was 100% for neck skin and cloacal swab samples. For low levels, accuracy was 100% and 92% for neck skin and cloacal swab samples, respectively; however, detection in shoe cover samples failed. A second collaborative trial, with an optimized DNA extraction procedure, gave 100% accuracy results for all three spiking levels. Finally, on-site validation at the abattoir on a flock basis was performed on 400 samples. Real-time PCR correctly identified 10 of 20 flocks as positive; thus, the method fulfilled the NordVal validation criteria and has since been implemented at a major abattoir.
Human campylobacteriosis is a widespread zoonotic food-borne infection (9). The majority of infections are sporadic and the sources rarely determined (20). Humans can be infected by eating insufficiently cooked meat products (poultry, pork, and beef), by drinking raw milk and polluted water, and by coming into contact with pets (2). A Danish case-control study has identified consumption of undercooked poultry as one of the main causes of human infections (19). In 2003, approximately 35% of Danish broiler flocks were infected with Campylobacter jejuni (6).
Rapid detection methods are an important part of any intervention strategy, which may also include separating Campylobacter-positive from Campylobacter-negative flocks at slaughter. The production of certified, fresh Campylobacter-free poultry products has been taking place in Denmark since 2002 and is the world's first example of fresh chickens labeled for sale as being “Campylobacter free.” This certification has been based on a conventional gel electrophoresis PCR. In general, the superiority of PCR testing lies in its ability to detect the pathogen in cloacal swab samples within the same working shift, making it possible to slaughter chicken flocks with a known Campylobacter-negative status before dealing with positive flocks. The purpose of the present study was to validate a newly developed TaqMan PCR method (12) in comparative and collaborative trials according to the recommendations of the Nordic validation organization, NordVal (5). The PCR method used includes an internal amplification control (IAC), reinforcing its value as a diagnostic tool.
The comparative validation was performed against two reference culture methods, applying different combinations of selective enrichment media and selective plating media. To our knowledge, none of the published, noncommercial, real-time PCR methods for the detection of thermotolerant Campylobacter have been evaluated through a collaborative trial (17, 21, 22, 23, 24). It is essential for routine application that an analytical method is reliable and consistent. To establish these features, extensive testing of the robustness and performance characteristics of the method is required, preferably through a collaborative trial, including end-use laboratories (10) and on-site on a flock basis.
The Preston broth was prepared according to the recommendations of the Nordic Committee on Food Analysis (1). Briefly, 25 g of nutrient broth no. 2 (catalog no. CM67; Oxoid, Basingstoke, United Kingdom) was dissolved in 1,000 ml of distilled water, autoclaved for 15 min at 121°C, and, after cooling to 50°C, was combined aseptically with a mixture of 50 ml lysed horse blood (Danish Institute for Food and Veterinary Research [DFVF], Copenhagen, Denmark), 4 ml of modified Preston Campylobacter selective supplement (catalog no. SR204E; Oxoid), and 4 ml of Campylobacter growth supplement (catalog no. SR084E; Oxoid).
The Bolton broth was prepared according to the manufacturer's recommendations. Briefly, 13.8 g of Bolton broth (catalog no. CM0983; Oxoid) was dissolved in 500 ml of distilled water, autoclaved for 15 min at 121°C, and, after cooling to 50°C was combined aseptically with a mixture of 25 ml lysed horse blood (DFVF) and one vial of Bolton broth selective supplement (SR0183E) reconstituted as directed.
The modified charcoal cefoperazone deoxycholate agar (mCCDA) was prepared according to the manufacturer's recommendations. Briefly, 45.5 g of Campylobacter blood-free selective agar (catalog no. CM0739; Oxoid) was suspended in 1,000 ml of distilled water and brought to boil to dissolve completely. The agar was autoclaved at 121°C for 15 min and, after cooling to 50°C, was combined aseptically with two vials of CCDA selective supplement (catalog no. SR0155; Oxoid) reconstituted twice with 2 ml sterile distilled water.
The Preston agar was prepared according to the manufacturer's recommendations. Briefly, 18.5 g of Campylobacter agar base (catalog no. CM0689; Oxoid) was suspended in 475 ml of distilled water and brought to boil to dissolve completely. The agar was autoclaved at 121°C for 15 min, and after cooling to 50°C was combined with a mixture of 25 ml lysed horse blood (DFVF) and one vial of Preston Campylobacter selective supplement (catalog no. SR0117; Oxoid) reconstituted with 2 ml of 50/50 acetone/sterile distilled water.
The brain heart infusion medium (BHI) was prepared according to the manufacturer's recommendations and consisted of 37 g/liter BHI (Difco, Sparks, MD) with 5% (vol/vol) calf blood (DFVF) and 0.5% agar added.
Real-time TaqMan PCR was performed in a RotorGene 3000 (Corbett Research, Australia) in 0.2-ml PCR tubes as described previously (12), except for the addition of 2.0 μl/reaction of 87% pure glycerol (Merck A/S, Denmark) and 1.0 μl/reaction of 12.5 mM dNTP mix with dUTP (Applied Biosystems, Foster City, CA), enabling uracil-N-glycosylase treatment to prevent carryover contamination. Each PCR analysis included a positive DNA control, a negative DNA control, a nontemplate control (NTC), and an IAC. The cutoff level and definition of positive/negative responses were exactly as described previously (12). Samples with a threshold cycle (CT) response below 40 were considered to be positive.
As shown in Table Table1,1, the following samples were collected in Denmark: 99 pooled fecal samples on shoe covers from rearing houses, 101 cloacal swab samples from an abattoir, 102 samples of neck skin from an abattoir, and 100 samples of neck skin from the retail sector. The shoe cover, cloacal swab, and neck skin samples from the abattoir originated from flocks at 34 different farms. Approximately half of the samples were taken in the spring, when the prevalence of Campylobacter spp. in chicken flocks is expected to be low in Denmark (approximately 20%), and the other half were taken in the early autumn, when the prevalence is expected to be high (approximately 60%). Approximately half of the neck skin samples from the retailers were taken from chicken flocks that were reported to be Campylobacter positive by the supplier, and the other half from flocks that had tested negative.
Fecal samples were collected on disposable shoe covers in the rearing houses and shipped at ambient temperature to DFVF. On arrival, the shoe cover samples were weighed and added to 1:10 (wt/vol) physiological saline and homogenized for 60 s in a stomacher.
The cloacal swab samples were taken at the abattoir from 12 or 13 individual broilers on arrival at the abattoir. The swabs were stored in tightly capped 15-ml plastic tubes with BHI and shipped to DFVF. On arrival, 20 samples were pooled into 1 sample and homogenized manually in 30 ml physiological saline for 60 s. Because of the high background flora in the fecal samples, growth during transportation (one day) was regarded as insignificant to the outcome of the test.
Approximately 40 g of chicken neck skin was cut and pooled into a plastic bag at the slaughter line. The samples were transferred to a stomacher bag, sealed, and shipped on ice packs to DFVF by mail. On arrival, the samples were diluted 1:1 in physiological saline and homogenized for 60 s in a stomacher.
For retail samples, fresh (nonfrozen) chickens were purchased; 40 g neck skin was cut at the laboratory and diluted 1:1 in physiological saline and homogenized for 60 s in a stomacher.
The detection of Campylobacter spp. was conducted in accordance with the recommendations from the Nordic Committee on Food Analyses (1) and the International Organization for Standardization (3). All samples were enriched in Bolton and Preston broths (1:10) at 42 ± 0.5°C in a microaerobic atmosphere (6% O2, 7% CO2, 7% H2, and 80% N2) for 24 h before 100 μl was plated onto Preston agar and mCCDA. The agar plates were incubated at 42 ± 0.5°C in a microaerobic atmosphere for 48 h. From the selective agar, five typical thermotolerant Campylobacter colonies were selected for verification by an internationally validated gel-based PCR assay for the identification of Campylobacter jejuni, C. coli, and C. lari (15, 16). In the absence of typical colonies, five nontypical colonies were selected for PCR verification.
From the fecal sample suspensions (cloacal swabs and shoe covers), 1-ml aliquots were drawn for DNA extraction before enrichment. From the neck skin samples, 1-ml aliquots were drawn for DNA extraction after overnight enrichment in Bolton broth. The samples were centrifuged at 16,000 × g for 7 min at 4°C, and DNA extraction was performed on a KingFisher processor (Thermo Lab Systems, Helsinki, Finland) using an automated, magnetically based separation and DNA isolation kit for blood, cells, and tissue (Thermo Lab Systems) as specified by the manufacturer. Briefly, the sample pellet was resuspended in lysis buffer and transferred to a 96-well plate (Thermo Lab Systems) containing magnetic particles, washing buffers, and elution buffer. The DNA extraction program consisted of two salt-buffer washing steps and two alcohol-buffer washing steps, followed by a final elution step (for a detailed protocol, see the Rapid Diagnostic Group website [http://www.pcr.dk/Innovations_pcr/innovations_pcr_startside.htm]). A total of 5 μl of the extracted DNA was used as the template in the real-time PCR.
The comparative validation study included three test characteristics: relative accuracy, sensitivity, and specificity (5) (see Table Table2).2). The relative accuracy is defined as the degree of correspondence between the response obtained by the alternative method and the reference method on identical samples, as follows: (PA + NA + FP) × 100/(PA + NA + TP + FN + FP), where PA refers to positive agreement, NA to negative agreement, FP to false positives, TP to true positives, and FN to false negatives. The relative sensitivity is defined as the ability of the alternative method to detect the target microorganism compared to the reference method, as follows: (PA + TP) × 100/(PA + FN). The relative specificity is defined as the ability of the alternative method not to detect the target microorganism when it is not detected by the reference method, as follows: (NA × 100)/(NA + FP).
To compare the performances of Bolton and Preston broths, the number of positive responses obtained from each was subtracted, giving one difference for each combination of sample type and agar type. The Wilcoxon signed-rank test with a continuity correction was applied to these differences to test whether the two enrichment broths differed significantly from each other (14). The calculations where performed using Splus software, professional edition version 6.1.
A collaborative trial involving nine national laboratories was performed to evaluate the robustness and reproducibility of the real-time PCR method testing identical samples.
The collaborative trial was designed and conducted according to the recommendations from NordVal (5). The nine participating laboratories received pellets from 18 coded 1-ml samples, including 6 chicken neck skin samples, 6 shoe cover samples, and 6 cloacal swab samples (see Table Table3).3). The samples were spiked in duplicate with C. jejuni CCUG 11284 at three levels, making it possible to assess the usefulness of the method at various infection levels. The shipment included a positive DNA control (1 μg/ml C. jejuni CCUG 11284) and a negative DNA control (1 μg/ml Arcobacter butzleri CCUG 30485), a ready-to-use PCR mixture with added IAC, and reagents for the magnetically based DNA extraction. To minimize any interlaboratory variability (not attributable to the method performance), we supplied all the reagents necessary. Each participant received a detailed protocol describing the DNA extraction, real-time PCR setup, real-time PCR run, and data analysis and a reporting form to record the obtained PCR results to return to DFVF. The participants were also asked to return a file containing the real-time PCR runs.
A second collaborative trial, comprising eight participating laboratories, was subsequently performed only on shoe cover samples. The second trial was performed exactly as the first one, except for using a modified DNA extraction protocol, with an increased amount of paramagnetic particles.
The samples for the collaborative trial were prepared as described above (“Sample preparation”). Regarding the neck skin samples, one broth was left unspiked, one was spiked with 1 to 10 CFU/100 ml, and one with 10 to 100 CFU/100 ml, and incubated at 42 ± 0.5°C for 24 h in a microaerobic atmosphere. After the enrichment, 1-ml aliquots were drawn and centrifuged at 16,000 × g for 7 min at 4°C. The supernatant was discarded, and the pellet kept at −80°C until shipped on ice to the trial participants.
From both shoe cover and cloacal swab samples, 1-ml aliquots were drawn, spiked with 0, 100 to 500, or 1,000 to 2,000 CFU/ml, and centrifuged at 16,000 × g for 7 min at 4°C. The supernatant was discarded and the pellet kept at −80°C until shipped on ice to the trial participants.
The Campylobacter status of all samples was confirmed at DFVF by the reference culture method according to International Organization for Standardization publication no. 10272-1 (3) and Nordic Committee on Food Analyses publication no. 119 (1) prior to and after spiking. The stability of the samples was examined using real-time PCR (12) immediately after spiking, prior to commencement of the collaborative trial, and during the period of analysis, to verify the continued detection of Campylobacter. The possible detrimental effect of shipping time at ambient temperature on the real-time PCR results was investigated, and no detrimental effect was found.
At the participating laboratories, DNA was extracted from the samples by using the Magnesil KF genomic system (Promega, Madison, WI) on an automated DNA extraction platform of the laboratories choosing but following the KingFisher protocol described above. In the second trial, the amount of paramagnetic particles was increased from 20 μl/sample to 75 μl/sample. A total of 5 μl purified DNA was used as the template in the real-time PCR.
Real-time PCR at the participating laboratories was performed on a Mx3000 or Mx4000 real-time PCR system (Stratagene, La Jolla, CA), ABI-PRISM 7000 or 7900 (Applied Biosystems, Foster City, CA), RotorGene 3000 (Corbett Research, Mortlake, Australia), or an iCycler thermal cycler (Bio-Rad, Hercules, CA). The participating laboratories were asked to use the blank (NTC), the process blank (a Campylobacter-negative sample processed throughout the entire protocol), and the negative control to assign the cutoff line and report back the CT values.
The test reports and the real-time PCR analyses from the participating laboratories were carefully evaluated on return to the expert laboratory, and the results were approved for inclusion in the statistical analysis, unless they fell into at least one of the following two categories: (i) obvious performance deviation from the protocol and (ii) presence of target amplicons in the negative control results, indicating cross contamination at the participating laboratory.
The results obtained in the collaborative trial were analyzed according to the recommendations from NordVal (5). The relative specificity was calculated for the unspiked samples by the following equation: (1 − [FP/N−]) × 100%, where N− refers to the total number of unspiked samples. The relative sensitivity was calculated for each level of spiking by the following equation: (TP/N+) × 100%, where N+ refers to the number of spiked samples. The relative accuracy was calculated for all levels of spiking by the following equation: ([PA + NA + FP]/N) × 100%, where N refers to the number of samples tested.
At a major abattoir, the real-time PCR method was validated against a routinely used gel-based PCR method approved previously by the Danish authorities to monitor production. The two methods were performed in parallel on 400 pooled cloacal swab samples collected from 20 chicken flocks. Each cloacal swab sample was a pool of 25 swabs. Thus, 20 pooled samples represented 500 chickens per flock. The samples giving a positive signal had CT values between 17 and 39. All samples giving negative signals had CT values of >40 (see Table Table55).
Out of 402 samples, 169 were found to be culture positive by at least one of the detection methods used (Table (Table1).1). The single highest number of positive samples was obtained by the combination of selective enrichment in Bolton or Preston broth followed by isolation on mCCDA, resulting in 167 positive samples. The combination of Bolton broth and Preston agar was slightly less effective (164 positive samples). Applying Preston broth followed by Preston agar resulted in 156 samples being found positive. The obtained P value was 0.32, indicating no significant difference between the Bolton and Preston broths.
A total of 167 out of 402 samples gave positive results by real-time PCR, compared to a total of 169 by the culture methods. Nine false-negative and 7 false-positive results were obtained by the real-time PCR method. This resulted in a relative accuracy of 98%, a relative sensitivity of 95%, and a relative specificity of 97% (Table (Table22).
In agreement with the predefined criteria, results from three laboratories were excluded because of obvious deviation from the protocol. The results from the remaining laboratories were accepted; thus, the final statistical analysis was performed on six sets of results (Table (Table33).
The relative specificity, sensitivity, and accuracy were 100% in the testing of neck skin samples. In the testing of cloacal swab samples, a relative specificity of 100% was obtained. A relative sensitivity of 92% was achieved for cloacal swab samples spiked with 100 to 500 CFU/ml, and 100% for the samples spiked with the higher levels, i.e., 1,000 to 2,000 CFU/ml (Table (Table44).
The real-time PCR method failed to detect any of the shoe cover samples, regardless of the level of spiking, except for one positive signal obtained from a sample spiked with 100 to 500 CFU/ml.
In the second collaborative trial on shoe cover samples, results from all eight participating laboratories were included in the statistical analysis. Only results from six laboratories are shown in Table Table3.3. The relative specificity, sensitivity, and accuracy for shoe cover samples were 94%, 100%, and 100%, respectively (Table (Table44).
Table Table55 shows the real-time PCR results obtained, compared to those from a routinely used gel-based PCR approved by the Danish authorities. The real-time PCR was performed at the testing laboratory of one of the major poultry producers in Denmark. The real-time PCR method correctly identified 10 out of 10 Campylobacter-positive chicken flocks and 10 out of 10 Campylobacter-negative chicken flocks.
Enrichment in selective broth will always be a compromise between the inhibition of competitive flora and the recovery and growth of the target microorganism. The results of the present study did not detect any difference (P = 0.32) in the ability of Preston and Bolton broth to support growth of Campylobacter. Martin et al. reached the same conclusion, testing 100 samples of chicken meat, sausage meat, pig offal, unpasteurized milk, and untreated water (18). However, these results differ from the findings of Baylis et al., who found Bolton broth superior to Preston broth, testing 100 raw foods, including chicken carcass, chicken meat, chicken liver, turkey, duck, beef, lamb liver, and pork sausage meat (7). In opposite findings, Borck et al. reported that Preston broth was better in tests of 41 turkey neck skin samples enriched in Preston broth and Campylobacter enrichment broth (same formula as Bolton broth) (8). Finally, Josefsen et al. have shown in a comprehensive study that Bolton broth and Preston broth equally support the growth of C. jejuni, while Preston broth was less effective in supporting the growth of Campylobacter coli (13). These differences in results can be attributed to the matrices and the background flora of the samples; thus, the two enrichment broths can be used equally for routine testing.
In the comparative study, the divergence in the number of negative results between the culture method and real-time PCR could be ascribed to the fact that in the present study, the fecal samples were enriched in Bolton and Preston broths for 24 h and consequently grown to a high concentration of Campylobacter spp. in the culture method. Real-time PCR, however, was performed directly on the fecal samples without any preceding enrichment. Seven samples were identified as positive by real-time PCR but were identified as negative using the culture method. This difference may be attributable to the presence of Campylobacter spp. that were viable but not culturable or dead in these samples.
In the first collaborative study, complete agreement between the real-time PCR method and the microbiological reference method was obtained for all test characteristics for neck skin samples. Compared to the reference culture method, real-time PCR detected 11 out of the 12 cloacal swab samples spiked with the target microorganism at a low level, corresponding to a 92% relative sensitivity. However, the level of thermotolerant Campylobacter in fecal samples from infected chicken flocks is usually in the range of 4 to 8 log10 CFU/g feces and will most probably be between 5.5 and 6.5 log10 CFU/g feces (22). The number of CFU to be analyzed in a 1-ml sample by the real-time method is estimated to be 2 × 104 to 2 × 105 CFU.
The considerable difference in the CT values observed among the participating laboratories (Table (Table3)3) for the same samples can be attributed to variation in the sensitivity of the real-time PCR platforms used. It has been shown that transferring a PCR method from one type of real-time instrument to another can result in a substantial shift in CT values (12).
In the first collaborative trial, the real-time PCR method failed to detect any of the shoe cover samples regardless of the level of spiking. Detection of C. jejuni in the shoe cover samples was possible at our laboratory at both low and high spiking levels. Posttrial investigations have shown that the reason for this discrepancy lies with the DNA extraction procedure. The extraction protocol provided to the participating laboratories was modified for compatibility with several DNA extraction platforms, and this approach resulted in a markedly reduced amount of total DNA recovery. The shoe cover samples can be more PCR inhibitory and yield less DNA than the other sample matrices, and it is likely that these features were the cause of the absence of positive signals.
In the second collaborative trial, the amount of paramagnetic particles was more than tripled, rectifying the problem with DNA extraction from this matrix. Shoe cover samples contain substantial amounts of extraneous material, and the results indicate that clotting of the magnetic particles hindered the absorption of DNA.
A relative specificity of 94% was obtained for shoe cover samples in the second collaborative trial, since one laboratory obtained a positive signal from one of the nonspiked samples (Table (Table3).3). This unexpected result could be due to cross contamination during DNA extraction or miscoding of the sample at DFVF.
The samples tested in the collaborative study were not naturally infected. Because the concentrations of the target microorganism in naturally infected samples are unknown, it would have been difficult to assess the method performance on samples with low levels of thermotolerant Campylobacter spp.
The on-site, flock-based validation of real-time PCR at the abattoir against an existing conventional gel-PCR method was successful. It should be noted that the Campylobacter status of the flocks had been previously determined by a reference culture method (1) on shoe cover samples from the rearing houses. This practice allows the poultry manufacturer to perform separated slaughtering. Pooled cloacal swab samples are taken out immediately after the killing and tested to determine the Campylobacter status of the chicken flocks and subsequently to label and mark the chickens as being Campylobacter free in accordance with Danish regulations. The regulation states that if one or more of the 20 pooled samples are Campylobacter positive, the whole flock should be regarded as being positive. Although some minor variations were seen in single samples, the results from the conventional gel-based PCR and the microbiological history of the flocks confirmed the real-time PCR results, emphasizing that the faster and less work demanding real-time PCR would be a practical alternative to the gel-based PCR.
In addition, the combination of an automated DNA extraction and the closed system of the real-time PCR provides a faster and less work-intensive method with a minimized risk of contamination compared to the gel-based PCR. Furthermore, the real-time PCR method includes the dUTP-uracil-N-glycosylase system, minimizing the risk of carryover contamination. The PCR reagents used in the method can be mixed in advance, distributed in smaller quantities, and frozen at −20°C for up to 6 months and be ready to use. These features are a major benefit for on-site use of the test. The method is an open-formula technique, i.e., the reagents and target gene, etc., are known, in contrast to commercial kits.
In conclusion, the real-time PCR method complied with the criteria for the validation of alternative microbiological methods and has been approved by NordVal as an alternative method for detection of thermotolerant Campylobacter spp. in chicken samples. The method is currently implemented for use in separated-slaughtering practice by the leading poultry producers in Denmark as part of a risk management program, and for the certified production of Campylobacter-free chicken.
This work was supported in part by the Danish Directorate for Food, Fisheries, and Agri-Business (DFFE) (grant no. 3401-66-03-5); in part by the CampyFood project of the Nordic Innovation Centre (grant no. F040301); in part by the Nordic Council of Ministers (grant no. 681048-05317); and in part by the European Union through the Food-PCR 2 research project as part of the Network of Excellence MED-VET-NET (FOOD-CT-2004-506122) under the 6th RTD Framework.
We thank Mette Skafte Thomsen, Sarah Omø Nielsen, and Stefan Jensen for excellent technical assistance and Danpo Ltd. for providing the sample material.