Evaluation of alternative sampling methods for Salmonella critical control point determination at broiler processing

Comparison of Stomaching versus Rinsing for Recovering Bacterial Communities from Rainbow Trout (Oncorhynchus mykiss) Fillets

ABSTRACT

The use of high-throughput methods allows a better characterization of food-related bacterial communities. However, such methods require large amounts of high-quality bacterial DNA, which may be a challenge when dealing with a complex matrix that has a low concentration of bacteria, such as fresh fish fillets. Therefore, the choice of method used to recover bacteria from a food matrix in a cost-effective way is critical, yet little information is available on the performance of commonly used methods. We assessed the recovery capacity of two such methods: stomaching and mechanical rinsing. The efficiency of the methods was evaluated through quantitative recovery and compatibility with end-point quantitative PCR (qPCR). Fresh rainbow trout (Oncorhynchus mykiss) fillets were inoculated with a bacterial marker, Brochothrix thermosphacta, at different concentrations (7.52 to 1.52 log CFU/g). The fillets were processed by one of the two methods, and the recovery of the marker in the suspensions was assessed by plate counting and qPCR targeting B. thermosphacta-rpoC. The same analyses were performed on six noninoculated fresh fillets. Stomaching and mechanical rinsing allowed efficient and repeatable recovery of the bacterial communities from the 42 inoculated fillets. No significant differences in recovery ratios were observed between the marker enumerated in the inoculation suspensions and in the corresponding recovery suspensions after rinsing and stomaching. However, the stomaching method allowed too many particles to pass through the filters bag, making necessary a limiting supplementary filtration step. As a consequence, only the rinsing recovery method allowed proper PCR quantification of the inoculated B. thermosphacta. The mean recovered bacterial level of the fillets was approximately 3 log CFU/g. It seems more relevant and cost-effective to recover the endogenous bacterial microbiota of a fish fillet structure using the rinsing method rather than the stomaching method.

HIGHLIGHTS

A farm-to-fork quantitative risk assessment model for Salmonella Heidelberg resistant to third-generation cephalosporins in broiler chickens in Canada

Salmonella Heidelberg resistant to ceftiofur (a third-generation cephalosporin antimicrobial agent) in broiler chicken products pose a risk to public health in Canada. The objective of this study was to assess the extent of that risk and to evaluate the effect of intervention measures along the agri-food chain. A stochastic farm-to-fork quantitative microbial risk assessment model was developed following the Codex Alimentarius Guidelines for Risk Analysis of Foodborne Antimicrobial Resistance. Different scenarios were analyzed to assess the individual relative effects of 18 possible interventions in comparison to a baseline scenario. The baseline scenario represented the first year of on-farm antimicrobial use surveillance in the Canadian broiler industry and the year before an industry-imposed ban on the preventive use of antimicrobials of very high importance to human health (2013), where 31.3% of broiler flocks consisted of birds to which ceftiofur was administered. The baseline scenario predicted an average probability of illness of 1.1 per 100,000 servings (SE: 0.064 per 100,000), corresponding to an average of 22,000 human infections (SE: 1900) with ceftiofur-resistant S. Heidelberg per year, which is likely an overestimation. This risk was reduced by 90% or 20% when two separate scenarios designed to capture the effect of withdrawing preventive ceftiofur use from poultry production were simulated using different approaches; data used for the former scenario were confounded by other potential concomitant control measures (e.g. Salmonella vaccination programme), so the true effect likely lies somewhere between the two estimates. A theoretical 'worst case' scenario where all flocks had birds exposed to ceftiofur increased the risk by 107%. A 50% reduction in the probability of human prior exposure to antimicrobials, which has a selective and competitive effect for Salmonella spp. following ingestion of contaminated products, reduced the risk by 65%. Other promising measures that could be considered for further risk management included improved cleaning and disinfection between broiler flocks on farm (risk reduction by 26%), exclusive use of air chilling (risk reduction by 34%), and the improvement of meat storage and preparation conditions, e.g., no temperature abuse at retail (risk reduction by 88%). These findings showed the importance of a structured approach to assessing and potentially implementing effective interventions to reduce the risk associated with ceftiofur-resistant S. Heidelberg at different steps along the agri-food chain. Major data gaps included information on concentrations of resistant bacteria, cross contamination at processing and how ceftiofur-resistant S. Heidelberg behave in comparison with susceptible ones, e.g., in terms of growth and survival ability, as well as pathogenicity and virulence.

Prevalence and serotypes of Salmonella spp. on chickens sold at retail outlets in Trinidad

BACKGROUND

This cross-sectional study determined the prevalence of Salmonella spp. and their serotypes on dressed chicken sold at retail outlets in Trinidad. The study also investigated the risk factors for contamination of dressed carcasses by Salmonella spp. at cottage poultry processor outlets where chickens are slaughtered and processed for sale.

METHODS

A total of 133 dressed, whole chickens and 87 chicken parts from 44 cottage poultry processors and 36 dressed, whole chickens and 194 chicken parts from 46 supermarket outlets were randomly collected throughout the country. Isolation and identification of Salmonella spp. were performed using standard bacteriological techniques. Serotyping was performed by a regional reference laboratory.

RESULTS

The prevalence of Salmonella spp. in chicken carcasses sampled from cottage poultry processors and supermarkets was 20.5% and 8.3% respectively (p <0.001). The frequency of isolation of Salmonella spp. at cottage poultry processors was 22.4%, 23.0%, 7.1%, and 10.0% for non-chilled whole chicken, non-chilled chicken parts, chilled whole chicken and chilled chicken parts respectively. Fresh, non-chilled chicken (22.6%) yielded a higher frequency of isolation of Salmonella spp. than chilled chickens (8.3%). For supermarket samples, the frequency of isolation of Salmonella spp. was 19.0%, 8.1%, 0.0% and 7.6% for chilled whole chickens, chill chicken parts, frozen whole chicken and frozen chicken parts respectively. The swab method of sampling yielded a statistically significantly (p = 0.029) higher frequency (3.2%) of Salmonella spp. than the rinse method (1.6%). The predominant serotypes isolated were Kentucky (30.9%) and Javiana (22.7%). Use of chilled water-bath to cool carcasses was the only risk factor significantly (p = 0.044) associated with isolation of Salmonella spp.

CONCLUSION

Raw chicken carcasses purchased from cottage poultry processors pose a significantly higher risk of contamination with Salmonella spp. than those sold at supermarkets.

Quantification of loosely associated and tightly associated bacteria on broiler carcass skin using swabbing, stomaching, and grinding methods

This research was conducted to quantify bacterial populations after swabbing or stomaching, followed by grinding the swabbed or stomached broiler skins. For each of 3 replications, 3 eviscerated broilers were randomly taken from a processing line in a local broiler processing plant. Ten swabs and 10 stomachs per bird were conducted on the left- and the right-side skins (10×7 cm), respectively, which were then finally ground. Results indicated that mesophilic aerobic bacteria (MAB) in the first swabbed sample were significantly lower than those in the first stomached sample (P<0.05), with no difference seen for the remaining sampling times (P>0.05). During 10 swabbings followed by final grinding, 8, 9, and 83% of MAB were detected after the first swabbing, after the second through 10th swabbings, and after final grinding of the skin, respectively. During 10 stomachings followed by the final grinding, 17, 18, and 65% of MAB were detected after the first stomaching, after the second through 10th stomachings, and after final grinding of the skin, respectively. Escherichia coli (E. coli) and coliforms were significantly higher in the first stomaching than those in the first swabbing (P<0.05), with no difference seen between the 2 sampling methods for the rest sampling times (P>0.05). Populations of E. coli and coliforms decreased step-wisely from the highest after grinding to the intermediate after first and second sampling, and to the least after 10th sampling (P<0.05), regardless of swabbing or grinding. In this study, less than 35% of MAB seemed loosely associated in the skin of eviscerated broiler, whereas more than 65% of MAB looked tightly associated, which were not recovered by stomaching or swabbing even 10 times but were recovered by grinding the skin.

Effect of hot water spray on broiler carcasses for reduction of loosely attached, intermediately attached, and tightly attached pathogenic (Salmonella and Campylobacter) and mesophilic aerobic bacteria

Chickens are known to harbor many bacteria, including pathogenic microorganisms such as Salmonella and Campylobacter. The objective of this study was to evaluate the efficacy of hot water spray (HWS, 71°C for 1 min) in reducing bacterial contamination of prechilled broiler carcasses. For each of 4 replications, skin samples from 5 broilers were collected at 3 processing stages: after bleeding (feathers removed manually), after evisceration (with/without HWS), and after water chilling. Broiler skin was quantitatively assessed for loosely attached (by rinsing the skin), intermediately attached (by stomaching the rinsed skin), and tightly attached (by grinding the rinsed/stomached skin) mesophilic aerobic bacteria (MAB) and Campylobacter as well as for the prevalence of Salmonella and Campylobacter. Broiler skins possessed 6.4 to 6.6 log cfu/g, 3.8 to 4.1 log cfu/g, and 2.8 to 3.5 log cfu/g of MAB populations after bleeding, evisceration, and chilling, respectively. The HWS resulted in more than 1 log unit of reduction in MAB immediately after evisceration and immediately after chilling regardless of microbial sampling method. Compared with MAB, the contamination of Campylobacter was low (1.7 to 2.6 log cfu/g) after bleeding, but the level was not reduced throughout the processing steps regardless of HWS. The application of HWS reduced the prevalence of Salmonella after chilling, but not for Campylobacter except for loosely attached cells. After hot water exposure, a partially cooked appearance was seen on both broiler skin and skinless breast surface. More research is required to effectively eliminate pathogenic organisms during processing and suppress any recovery of bacteria regardless of attachment type after chilling.

Scientific and technical factors affecting the setting of Salmonella criteria for raw poultry: a global perspective

Concerns about foodborne salmonellosis have led many countries to introduce microbiological criteria for certain food products. If such criteria are not well-grounded in science, they could be an unjustified obstacle to trade. Raw poultry products are an important part of the global food market. Import and export ambiguities and regulatory confusion resulting from different Salmonella requirements were the impetus for convening an international group of scientific experts from 16 countries to discuss the scientific and technical issues that affect the setting of a microbiological criterion for Salmonella contamination of raw chicken. A particular concern for the group was the use of criteria implying a zero tolerance for Salmonella and suggesting complete absence of the pathogen. The notion can be interpreted differently by various stakeholders and was considered inappropriate because there is neither an effective means of eliminating Salmonella from raw poultry nor any practical method for verifying its absence. Therefore, it may be more useful at present to set food safety metrics that involve reductions in hazard levels. Such terms as "zero tolerance" or "absence of a microbe" in relation to raw poultry should be avoided unless defined and explained by international agreement. Risk assessment provides a more meaningful approach than a zero tolerance philosophy, and new metrics, such as performance objectives that are linked to human health outcomes, should be utilized throughout the food chain to help define risk and identify ways to reduce adverse effects on public health.

Development of stable reporter system cloning luxCDABE genes into chromosome of Salmonella enterica serotypes using Tn7 transposon

BACKGROUND

Salmonellosis may be a food safety problem when raw food products are mishandled and not fully cooked. In previous work, we developed bioluminescent Salmonella enterica serotypes using a plasmid-based reporting system that can be used for real-time monitoring of the pathogen's growth on food products in short term studies. In this study, we report the use of a Tn7-based transposon system for subcloning of luxCDABE genes into the chromosome of eleven Salmonella enterica serotypes isolated from the broiler production continuum.

RESULTS

We found that the lux operon is constitutively expressed from the chromosome post-transposition and the lux cassette is stable without external pressure, i.e. antibiotic selection, for all Salmonella enterica serotypes used. Bioluminescence expression is based on an active electron transport chain and is directly related with metabolic activity. This relationship was quantified by measuring bioluminescence against a temperature gradient in aqueous solution using a luminometer. In addition, bioluminescent monitoring of two serotypes confirmed that our chicken skin model has the potential to be used to evaluate pathogen mitigation strategies.

CONCLUSIONS

This study demonstrated that our new stable reporting system eliminates bioluminescence variation due to plasmid instability and provides a reliable real-time experimental system to study application of preventive measures for Salmonella on food products in real-time for both short and long term studies.

Development of bioluminescent Salmonella strains for use in food safety

BACKGROUND

Salmonella can reside in healthy animals without the manifestation of any adverse effects on the carrier. If raw products of animal origin are not handled properly during processing or cooked to a proper temperature during preparation, salmonellosis can occur. In this research, we developed bioluminescent Salmonella strains that can be used for real-time monitoring of the pathogen's growth on food products. To accomplish this, twelve Salmonella strains from the broiler production continuum were transformed with the broad host range plasmid pAKlux1, and a chicken skin attachment model was developed.

RESULTS

Salmonella strains carrying pAKlux1 constitutively expressed the luxCDABE operon and were therefore detectable using bioluminescence. Strains were characterized in terms of bioluminescence properties and plasmid stability. To assess the usefulness of bioluminescent Salmonella strains in food safety studies, we developed an attachment model using chicken skin. The effect of washing on attachment of Salmonella strains to chicken skin was tested using bioluminescent strains, which revealed the attachment properties of each strain.

CONCLUSION

This study demonstrated that bioluminescence is a sensitive and effective tool to detect Salmonella on food products in real-time. Bioluminescence imaging is a promising technology that can be utilized to evaluate new food safety measures for reducing Salmonella contamination on food products.

Number of Salmonella on chicken breast filet at retail level and its implications for public health risk

This study aimed to characterize the number of Salmonella on chicken breast filet at the retail level and to evaluate if this number affects the risk of salmonellosis. From October to December 2005, 220 chilled raw filets (without skin) were collected from five local retail outlets in The Netherlands. Filet rinses that were positive after enrichment were enumerated with a three-tube most-probable-number (MPN) assay. Nineteen filets (8.6%) were contaminated above the detection limit of the MPN method (10 Salmonella per filet). The number of Salmonella on positive filets varied from 1 to 3.81 log MPN per filet. The obtained enumeration data were applied in a risk assessment model. The model considered possible growth during domestic storage, cross-contamination from filet via a cutting board to lettuce, and possible illness due to consumption of the prepared lettuce. A screening analysis with expected-case and worst-case estimates for the input values of the model showed that variability in the inputs was of relevance. Therefore, a Monte Carlo simulation with probability distributions for the inputs was carried out to predict the annual number of illnesses. Remarkably, over two-thirds of annual predicted illnesses were caused by the small fraction of filets containing more than 3 log Salmonella at retail (0.8% of all filets). The enumeration results can be used to confirm this hypothesis in a more elaborate risk assessment. Modeling of the supply chain can provide insight for possible intervention strategies to reduce the incidence of rare, but extreme levels. Reduction seems feasible within current practices, because the retail market study indicated a significant difference between suppliers.

Validation of individual and multiple-sequential interventions for reduction of microbial populations during processing of poultry carcasses and parts

Changes in aerobic plate counts (APC), total coliform counts (TCC), Escherichia coli counts (ECC), and Salmonella incidence on poultry carcasses and parts and in poultry processing water were evaluated. Bacterial counts were estimated before and after individual interventions and after poultry carcasses were exposed to multiple-sequential interventions at various stages during the slaughter process. Individual and multiple-sequential interventions were evaluated at three processing plants: (i) plant A (New York wash, postevisceration wash, inside-outside bird washes 1 and 2, chlorine dioxide wash, chlorine dioxide wash plus chlorine chiller, chiller exit spray, and postchiller wash), (ii) plant B (New York wash, inside-outside bird washes 1 and 2, trisodium phosphate wash, and chlorine chiller), and (iii) plant C (trisodium phosphate wash and chlorine chiller). The majority of individual interventions effectively or significantly (P < 0.05) reduced microbial populations on or in carcasses, carcass parts, and processing water. Reductions in APC, TCC, and ECC due to individual interventions ranged from 0 to 1.2, 0 to 1.2, and 0 to 0.8 log CFU/ml, respectively. Individual interventions reduced Salmonella incidence by 0 to 100% depending on the type of process and product. Multiple-sequential interventions resulted in significant reductions (P < 0.05) in APC, TCC, ECC, and Salmonella incidence of 2.4, 2.8, and 2.9 log CFU/ml and 79%, respectively, at plant A; 1.8, 1.7, and 1.6 log CFU/ml and 91%, respectively, at plant B; and 0.8, 1.1, and 0.9 log CFU/ml and 40%, respectively, at plant C. These results enabled validation of in-plant poultry processing interventions and provide a source of information to help the industry in its selection of antimicrobial strategies.

Prevalence and molecular profiles of Salmonella collected at a commercial turkey processing plant

In this study, whole carcasses were sampled at eight stages on a turkey-processing line and Salmonella prevalence was determined using enrichment techniques. Recovered Salmonella was further characterized using serotyping and the molecular profiles were determined using pulsed-field gel electrophoresis (PFGE). Prevalence data showed that contamination rates varied along the line and were greatest after defeathering and after chilling. Analysis of contamination in relation to serotypes and PFGE profiles found that on some visits the same serotype was present all along the processing line while on other days, additional serotypes were recovered that were not detected earlier on the line, suggesting that the birds harbored more than one serotype of Salmonella or there was cross-contamination occurring during processing. Overall, this study found fluctuations in Salmonella prevalence along a turkey-processing line. Following washing, Salmonella prevalence was significantly reduced, suggesting that washing is critical for Salmonella control in turkey processing and has significant application for controlling Salmonella at the postdefeathering and prechill stages where prevalence increased.

Comparison of different sampling techniques and enumeration methods for the isolation and quantification of Campylobacter spp. in raw retail chicken legs

Comparable quantitative data of Campylobacter spp. on chicken products are a major data lack for quantitative risk assessment approaches. The objective of this study was to compare two different sampling techniques for the isolation and enumeration of Campylobacter spp. in chicken and to evaluate a suitable enumeration method comparing the most probable number (MPN) technique to the direct plating method. For this, 90 packages containing at least two raw chicken legs were examined for the comparison of sampling techniques, rinsing one leg and homogenizing 25 g of skin of the other leg of each package; both sample preparation types were examined by direct plating method and MPN technique in 40 out of 90 packages. Of the skin samples, 70% (63/90), and of the rinse samples, 77% (69/90), were Campylobacter-positive. Enumeration of Campylobacter spp. by direct plating revealed a median of log 4 cfu/leg surface in skin samples (S.D.=0.6) and a median of log 4.3 cfu/leg surface in rinse samples (S.D.=0.9) of the rinse samples; 73% (37/51) had higher numbers of Campylobacter spp. than the skin samples although the difference was not significant (p=0.08). The correlation coefficient of Campylobacter counts in skin and rinse samples was 0.43. The prevalence of Campylobacter spp. in rinse samples was 58% (23/40). In 5% (2/40) of the rinse samples, numbers of Campylobacter spp. could be detected only by the MPN technique due to the lower detection limit compared to the direct plating method. The MPN technique turned out to be unsuitable for the enumeration of Campylobacter spp. in skin samples because a layer formation on the top of the incubated MPN-tubes leads to irregular MPN results. Out of 80% (16/20) of the compared rinse samples, the direct plating detected higher numbers of Campylobacter spp., with a median count of log 4.2 cfu/leg surface (S.D.=1) compared the MPN technique where a median of log 4 cfu/leg surface (S.D.=1.1) was obtained. The difference was not significant (p=0.05). A highly positive correlation coefficient of 0.9 was observed between the direct plating and the MPN technique. Both sampling methods, rinsing the chicken leg and homogenizing the skin, are suitable for the detection and quantification of Campylobacter spp.; the direct plating method was superior to the MPN technique for enumerating Campylobacter spp. in raw chicken legs at retail level because enumeration is more rapid and less laborious.

An evaluation of sampling methods for the detection of Escherichia coli and Salmonella on Turkey carcasses

The efficacy of rinse, excision, and swab methods for the microbiological analysis of prechill turkey carcasses was investigated. Aerobic plate counts from a 50-cm2 area of the breast sampled by excision and by swabbing were compared. Escherichia coli and Salmonella recoveries were determined from turkeys sampled by a carcass rinse (CR), a modified rinse with the carcass supported in a swing (MCR), a two-site swab of 50 cm2 at the back and thigh (2S), a one-site swab of 50 cm2 beneath the wing (1S), a whole-carcass swab of the inner and outer carcass surface (WS), and excision of 25 g of neck skin tissue (NE). The effect of diluent volume (25, 50, and 100 ml) on E. coli counts from swab samples was also assessed. The aerobic plate count from breast tissue sampled by excision was greater than that by swabbing (P < 0.05). E. coli recoveries by the MCR method were similar to those by CR. E. coli counts from IS and WS samples were higher when swabs were stomached in 50 rather than 25 ml of diluent (P < 0.05). For swabs stomached in 50 ml of diluent, E. coli recoveries by the MCR, 2S, 1S, and WS methods were similar. For swabs stomached in 50 ml of diluent, Salmonella recoveries by the WS and MCR methods were higher than those by the 2S and 1S methods. Excision was more effective than swabbing for obtaining total bacterial counts from reduced turkey carcass areas. Whole-carcass sampling by rinsing or swabbing is necessary for optimum Salmonella recovery. Sampling a reduced area of the carcass is sufficient for E. coli analysis.

Predicted quantitative effect of logistic slaughter on microbial prevalence

Logistic slaughter is an intervention measure intended to reduce cross-contamination during slaughter by slaughtering contaminated units (=(groups of) animals) last. This paper describes a simple mathematical model which predicts the prevalences of contaminated units after logistic and random-order slaughter. The effect of logistic slaughter is the difference between these prevalences. The model assumes that uncontaminated units can become contaminated by contaminated units that were slaughtered before them; the contributions of contaminated units are independent. It also assumes that a slaughterhouse is uncontaminated at the start of the day and that a unit that is contaminated before slaughter also is contaminated after slaughter. The model was analysed using numerical simulations; for a selection of cases, analytical formulas can be derived and are presented. Contamination of broiler flocks with Salmonella was used as a case study. Even for this simple model, data availability is a problem leading to uncertain parameter estimates. An average cross-contamination scenario predicts that the beneficial effect of logistic slaughter is as low as 9.1%, which casts doubt on its usefulness as an intervention measure. The case study produced these general model results: the effect of logistic slaughter increases with the probability of cross-contamination between units; with the length of the slaughter queue; and with sensitivity (the probability of a positive test from a unit contaminated at the start of slaughter). However, the effect is small if the prevalence of contaminated units before slaughter is low or high.