Reduction of Salmonella contamination on the surface of chicken skin using bacteriophage

All of the phages used in this study (Eϕ151, Tϕ10 and Tϕ11) were isolated from poultry excreta and abattoir effluent as reported previously [24]. All of these phage exhibited a clear plaquing phenotype on their respective host strains of Salmonella used in this study, indicating a lytic lifecycle on these strains. A lytic spectrum for phage Tϕ11 is presented in Supplementary Table 1, and electron micrographs of this phage are presented in Supplementary Figures 1 and 2. Phages were propagated on their host strains in nutrient broth (NB, CM0001, Oxoid, Basingstoke, UK) using a modification of a previously described protocol [35]. Briefly, a volume of fresh overnight culture of the host strain (0.1 ml) was added to 10 ml of pre-warmed NB (37 °C) in a 30 ml tube. To this was added 0.1 ml of a 10⁶ PFU ml^− 1 suspension of phage. The suspension was incubated statically at 37 °C for up to 8 h until lysis was apparent (compared with an uninfected control culture). The lysate was then filtered through a 0.22 μm pore-size filter (Millipore) and the phage titer determined using the method described below. Phage stocks were stored at 4 °C until required, but for no longer than 48 h before application. Phages in crude liquid lysates were concentrated and purified using PEG precipitation, as described previously [36]. Phage titres were determined by decimally diluting the phage suspension in SM buffer (50 mM Tris-Cl [pH 7.5], 0.1 M NaCl, 8 mM MgSO₄.7H₂O, 0.01% w/v gelatin, reagents from Sigma) then adding duplicate 100 μl volumes of each dilution to equal volumes of approximately 8.0 log₁₀ CFU ml^− 1 of an overnight NB culture of Salmonella host strain and incubating for 15 min at 37 °C. Each of these suspensions was then added to 5 ml of molten overlay agar (NB containing 0.5% w/v bacteriological agar LP0011, Oxoid), gently shaken, and poured over pre-warmed (37 °C, 30 min) nutrient agar plates (NA, CM0003, Oxoid). After allowing the overlay to set, the plates were incubated at 37 °C for 24 h before examining for plaques.

Salmonella-free Ross broiler chickens (36 days old, n = 36) were obtained from a commercial supplier (Lloyd Maunder, Devon, United Kingdom). Upon arrival at the University, the birds were separated into two equal groups and housed in groups of three in floor boxes in a controlled environment under strict conditions of biosecurity. To ensure that the experimental birds remained free of naturally-occurring Salmonella infection, faeces were collected periodically and screened for Salmonella by an enrichment step in modified Rappaport-Vassiliadis Soya peptone broth (RVS, CM0669, Oxoid), followed by streaking onto Brilliant Green Agar (BG, CM0329, Oxoid). Faecal (and later, caecal) samples were also collected to determine if any pre-existing Salmonella phages were present using the enrichment method for the environmental samples described previously [24]. On the day of arrival the birds were orally inoculated with 0.3 ml of an 8.0 log₁₀ CFU ml^− 1 suspension of either S. Enteritidis P125109 (group 1, n = 18), or S. Typhimurium 4/74 (group 2, n = 18). Both of these Salmonella strains were resistant to sodium nalidixate (20 μg ml^− 1). All of the birds were killed by cervical dislocation 7 days after Salmonella challenge. The birds were hand plucked (without scalding or washing) and then two sections of skin (each 25 cm²) were carefully excised from the breast skin using sterile scissors and a disposable template, and placed into separate sterile Petri dishes. After the skin sections were collected from the birds and removed to a separate laboratory, the abdominal cavity was opened and the caeca were removed. This procedure was followed to ensure that there would be no possibility of gut contents contaminating the surface of the skin. The contents of the caecal lumen were collected in sterile universal tubes for Salmonella and phage enumeration using the same methods described above.

A total of 36 birds were used in the experiment, half of which were orally infected with S. Enteritidis and half with S. Typhimurium (see above). Two sections of skin were collected from each bird (72 skin samples total). One section of skin from each bird was sprayed with 1 ml (0.5 ml per side) of SM buffer (control) using a hand-operated plant spray, with the nozzle positioned approximately 10 cm from the surface of the skin and delivering 0.5 ml of liquid per spray. The other section was sprayed in an identical manner with 1 ml of a 9.0 log₁₀ PFU ml^− 1 of phage Eϕ151 (group 1), or Tϕ10 (group 2) for birds challenged with S. Enteritidis or S. Typhimurium respectively. After allowing 20 min to dry in a Class II biological safety hood, each skin section was transferred into a stomacher bag containing 50 ml of maximum recovery diluent (MRD, CM0733, Oxoid) and stomached for 1 min. Salmonellas in the stomachate were enumerated using two parallel methods. For the first method, decimal dilutions of the stomachate were prepared in MRD. Volumes (100 μl) of each dilution were then spread-plated onto BG agar containing 25 μg ml^− 1 sodium nalidixate (BG nal) and incubated at 37 °C for 24 h before examining for typical Salmonella colonies. The second method used the Most Probable Number technique (MPN) [37] using Rappaport Vassiliadis broth as the enrichment medium and BG nal agar for the plating medium.

In order to determine the rate of reduction of Salmonella numbers on the skin surface, a series of experiments were performed using a bioluminescent Salmonella host strain. The strain used was S. Typhimurium DT104 transformed with the pBBR1MCS5-LITE lux plasmid as described previously [38]. A linear relationship between photon and viable counts has previously been demonstrated for the plasmid construct when expressed in Pseudomonas aeruginosa (r2 = 0.982) [39] and in the S. Typhimurium DT104 used in this study (r2 = 0.9856) [40]. S. Typhimurium DT104 pBBR1MCS-5-LITE has been used successfully to assess the impact of in situ rapid heating and cooling on food surfaces, showing a strong correlation between bioluminescence and cell numbers (r² = 0.97) [41, 42].

Sections of chicken breast skin were inoculated with 100 μl of a 10⁶ CFU ml^− 1 suspension of an overnight NB culture of the bioluminescent Salmonella strain, which had been washed twice and resuspended in MRD. The inoculum was spread evenly over the surface of the skin and then left to dry at 20 °C for 1 h in a Class II biological safety hood. After this time, two 25 cm² sections were excised from the same piece of skin for each trial. The first section of skin (control) was sprayed with 0.5 ml of MRD; the second skin section (treated) was sprayed with 0.5 ml of a 10⁹ PFU ml^− 1 suspension of phage Tϕ11. The skin surfaces were viewed within 1 min of spraying, in a dark room with an ICCD 225 photon counting camera (Photek Ltd., St Lenards-on-Sea, East Sussex, TN38 9NS, UK). Photons were counted for 1 min at set intervals for up to 24 h at 20 °C. Each trial was repeated five times.

All statistical tests were performed on log₁₀ transformed data. The D’Agostino & Pearson omnibus normality test was used to determine if the data had an underlying normal distribution as this is a prerequisite for the use of parametric statistical tests. Where a normal distribution could not be determined, the median counts of Salmonella recovered from control and phage-treated skin sections were compared using the non-parametric Mann-Whitney U test. The proportion of control and phage-treated skin sections that were culture positive for Salmonella were compared using Fisher’s Exact Test. All statistical calculations were performed using SPSS® 14.0 for Microsoft Windows, SPSS Inc., Chicago, USA.

Data showing the recovery of S. Enteritidis and S. Typhimurium from chicken skin following treatment with phage is presented in Table 1. Salmonella and their phage were not recovered from any of the faecal samples taken from birds prior to experimental inoculation. No Salmonella phage could be isolated from the caecal content samples taken from the birds of the control group after slaughter. However, Salmonella spp. was recovered from the caeca of all birds used for the skin disinfection trials. The mean level of colonisation for birds infected with S. Enteritidis and S. Typhimurium (as determined from colony counts) was 4.0 ± 1.3 and 3.5 ± 1.0 log₁₀ CFU g^− 1 caecal contents respectively. When recovering Salmonella from some of the skin sections, all of the tubes inoculated for the MPN enumeration method tested positive for Salmonella. These samples were given a value of ≥3.04 log₁₀ MPN per skin section (the upper detection limit of the MPN method used here). In addition, as the exact number of Salmonella in these samples was unknown, the mean and standard deviation could not be calculated accurately. As such, the median values and range are presented here. All of the control skin sections treated with buffer tested positive for Salmonella using the MPN method. The phage treatment of S. Enteritidis-contaminated skin sections resulted in a significant (p < 0.0001, Mann-Whitney U Test) median reduction of 1.38 log₁₀ MPN per skin section compared with the control. Moreover, S. Enteritidis could not be recovered from 13/18 of the phage-treated skin sections, which was also significantly lower than the control group (p < 0.0001, Fisher’s Exact Test). Phage treatment of S. Typhimurium-contaminated skin sections led to similar results. The median recovery of S. Typhimurium was reduced significantly (p < 0.0001) by 1.83 log₁₀ MPN per skin section compared with the control group. The proportion of Salmonella-positive skin sections following phage treatment (11/18, 61.1%) was also significantly lower than the control group (p = 0.0076).

Photographs showing the photon counts from chicken skin sections experimentally inoculated with a bioluminescent strain of S. Typhimurium before and after spray treatment with a buffer or phage suspension are presented in Fig. 1. The mean photon count from the skin sections before buffer or phage treatment was 4.3 log₁₀ RLU per skin section. The photon count for the skin treated with buffer increased slightly to 4.4 log₁₀ RLU per skin section after spraying but this increase was not significant (p > 0.1). The mean photon count from the skin sections treated with phage fell from 4.3 to 4.1 log₁₀ RLU which was a modest but statistically significant reduction compared with the control skin sections (p < 0.01). Following incubation for 24 h at room temperature (20 °C), the mean photon counts increased for the control skin sections (6.1 log₁₀ RLU). The mean photon counts also increased for the phage-treated skin sections but reached a level significantly lower (p < 0.05) than the control skin sections (5.2 log₁₀ RLU).