Development of a Rapid Real-Time PCR Method as a Tool To Quantify Viable Photobacterium phosphoreum Bacteria in Salmon (Salmo salar) Steaks

Abstract

A specific real-time PCR quantification method combined with a propidium monoazide sample treatment step was developed to determine quantitatively the viable population of the Photobacterium phosphoreum species group in raw modified-atmosphere-packed salmon. Primers were designed to amplify a 350-bp fragment of the gyrase subunit B gene (gyrB) of P. phosphoreum. The specificity of the two primers was demonstrated by using purified DNA from 81 strains of 52 different bacterial species. When these primers were used for real-time PCR in pure culture, a good correlation (R² of 0.99) was obtained between this method and conventional enumeration on marine agar (MA). Quantification was linear over 5 log units as confirmed by using inoculated salmon samples. On naturally contaminated fresh salmon, the new real-time PCR method performed successfully with a quantification limit of 3 log CFU/g. A correlation coefficient (R²) of 0.963 was obtained between the PCR method and classic enumeration on MA, followed by identification of colonies (290 isolates identified by real-time PCR or by 16S rRNA gene sequencing). A good correlation with an R² of 0.940 was found between the new PCR method and an available specific conductance method for P. phosphoreum. This study presents a rapid tool for producing reliable quantitative data on viable P. phosphoreum bacteria in fresh salmon in 6 h. This new culture-independent method will be valuable for future fish inspection, the assessment of raw material quality in fish processing plants, and studies on the ecology of this important specific spoilage microorganism.

INTRODUCTION

Modified-atmosphere-packed (MAP) fresh fish is increasingly popular in Europe and widely sold in supermarkets as a chilled product. Compared to aerobically stored fresh fish, this kind of packaging, with typical headspace CO₂ concentrations of 20 to 60%, modifies the dominating microbiota mainly by reducing growth of aerobic Gram-negative bacteria like Pseudomonas (1, 2). This results in an extended shelf life which facilitates the chilled distribution and marketing of fresh MAP fish. However, this packaging allows the growth of CO₂-resistant bacteria, including Gram-positive lactic acid bacteria and the Gram-negative bacterium Photobacterium phosphoreum (2–4). The latter has been identified as the specific spoilage organism (SSO) responsible for trimethylamine production and sensory spoilage of MAP cod (5, 6) and as the main spoilage bacterial species of several chilled marine fish, including cod, garfish, halibut, saithe, salmon, and shrimp (7–15).

Multilocus analysis, based on 16S rRNA and gyrB and luxABFE genes, divided strains originally isolated and identified as P. phosphoreum into three distinct clades of P. phosphoreum, Photobacterium iliopiscarium, and Photobacterium kishitanii (16, 17). The members of the P. phosphoreum species group have been reported as important for spoilage of raw MAP fish, and they include both luminous and nonluminous variants (5, 18). It is therefore relevant to detect and enumerate this species group in fish products.

As no specific medium is available for enumeration of either the P. phosphoreum species group or the individual species in seafood, the control of these important spoilage bacteria is problematic. There is a specific conductance method for quantifying the P. phosphoreum species group in fish samples (19) which is quantitative, sensitive, and selective. However, it is not very rapid, with analyses taking 12 to 50 h depending on the bacterial concentration in the sample. Instruments required to run this conductance method are available, e.g., from SY-LAB in Austria, but they are far less widespread than real-time PCR thermocyclers. Therefore, the development of a rapid and specific real-time PCR quantification method for viable cells of the P. phosphoreum species group in fresh fish is clearly of interest. One of the limitations of PCR and real-time PCR methods for the quantification of bacteria in food samples is that both live and dead cells are detected (20). To overcome this problem, propidium monoazide (PMA) can be used before DNA extraction. PMA is an intercalating DNA agent which enters dead cells and binds to DNA, inhibiting subsequent PCR amplification and thereby ensuring quantification of viable bacteria (21). PMA has been used in combination with several real-time PCR methods for quantification of viable bacteria in food, e.g., Campylobacter (22), Listeria monocytogenes (23), Brochothrix thermosphacta (24, 25), Vibrio parahaemolyticus (26, 27), and Escherichia coli O157:H7 (28).

In the present study, we describe a new rapid assay to quantify and identify the P. phosphoreum species group in raw salmon using real-time PCR. This new method has been validated for artificially inoculated and naturally contaminated fresh MAP salmon and compared to both the classic enumeration method on culture media followed by colony identification and the conductance method specifically developed for these bacteria. The term P. phosphoreum will be used below to refer collectively to the closely related members of the P. phosphoreum species group.

MATERIALS AND METHODS

Raw salmon samples.

To establish a standard curve on artificially contaminated fish, 24 samples of fresh salmon were inoculated with different concentrations of P. phosphoreum and analyzed. In order to evaluate the real-time PCR quantification method using naturally contaminated samples, it was compared with viable plate counting with 13 samples of MAP salmon steaks. These were obtained from a local French processing plant or purchased from a local supermarket and were packed with a headspace gas composition of approximately 50% CO₂ and 50% N₂. With the aim of comparing the real-time PCR method and a conductance method for quantification of P. phosphoreum, 26 naturally contaminated MAP salmon steaks were assessed. These samples were obtained from a Danish processing plant and were packed with a headspace gas composition of approximately 20% CO₂ and 80% N₂.

Bacterial strains and DNA extraction.

Bacterial strains used in this study are listed in Table 1. P. phosphoreum strains were grown in brain heart infusion broth (BHI; Biokar Diagnostics, Beauvais, France) supplemented with 2.0% NaCl and incubated at 15°C for 24 h.

Table 1.

Bacterial isolates used to develop and evaluate the specificity of a real-time PCR assay for detection of the P. phosphoreum species group

Bacterial species	Strain	Source	No. of strains	Real-time PCR resulta
Photobacterium phosphoreum group
P. phosphoreum	CIP 102511T	Fish	1	+
P. phosphoreum	DSM 2167	Hake (Merluccius capensis)	1	+
P. iliopiscarium	DSM 9896	Pyloric ceca of herring	1	+
P. kishitanii	LMG 23890	Light organ of Physiculus japonicus	1	+
P. phosphoreum	Oniris/Ifremer Collection	Spoiled salmon steak or fillets under MAP	23	+
Other Photobacterium species
P. leiognathi	CIP 66.5T	Light organ of teleostean fish (Leiognathus equula)	1	−
P. angustum	CIP 75.10T	Seawater at depth of 750 m	1	−
P. damselae subsp. piscicida	CIP 103910	White perch	1	−
P. damselae subsp. damselae	DSM 7482	Fish ulcer	1	−
P. indicum	LMG 22857	Marine mud	1	−
P. lypolyticum	LMG 23071	Light organ of teleostean fish (family Leiognathidae)	1	−
P. rosenbergii	LMG 22223	Tissue extract of bleached coral	1	−
Non-Photobacterium strains
Aeromonas hydrophila	Oniris/Ifremer Collection	Spoiled shrimp under MAP	1	−
Aeromonas salmonicida subsp. salmonicida	Oniris/Ifremer Collection	Spoiled shrimp under MAP	1	−
Buttiauxela gaviniae	Oniris/Ifremer Collection	Spoiled shrimp under MAP	1	−
Brochothrix campestris	DSM 4712	Soil	1	−
Brochothrix thermosphacta	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Carnobacterium maltaromaticum	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
C. maltaromaticum	NCDO 2762	Kidney of adult cutthroat trout	1	−
Carnobacterium divergens	NCDO 2763	Vacuum-packed minced beef	1	−
C. divergens	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Escherichia coli	ATCC 25922 CIP 76.24	Clinical isolate	1	−
Enterococcus faecalis	CIP 105042	Poultry cecum	1	−
Enterococcus faecium	CIP 106742	Dairy products	1	−
Hafnia alvei	Oniris/Ifremer Collection	Spoiled salmon fillet under MAP	1	−
H. alvei	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Listeria monocytogenes	CIP 78.35	Spinal fluid	1	−
Lactococcus lactis	CNRZ 1075	Origin not available	1	−
Lactococcus piscium	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Lactobacillus raffinolactis	Oniris/Ifremer Collection	Spoiled salmon fillet under MAP	1	−
Lactobacillus curvatus	Oniris/Ifremer Collection	Cold-smoked salmon	1	−
Lactobacillus fuchuensis	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Lactobacillus sakei	NBIMCC 3453	NBIMCC	1	−
Leuconostoc gasicomitatum	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Listeria innocua	CIP 80.11T	Bovine brain	1	−
Moellerella wisconsensis	Oniris/Ifremer Collection	Spoiled salmon fillet under MAP	1	−
Morganella morganii subsp. morganii	CIP A231	A case of summer diarrhea	1	−
Morganella psychrotolerans	Oniris/Ifremer Collection	Spoiled salmon fillet under MAP	1	−
M. psychrotolerans	CIP 109403	Cold-smoked tuna	1	−
Proteus mirabilis	CIP A235	Origin not available	1	−
Pseudomonas fluorescens	CIP 69.13T	Prefilter tanks	1	−
Pseudoalteromonas sp.	Oniris/Ifremer Collection	Spoiled shrimp under MAP	1	−
Psychrobacter sp.	Oniris/Ifremer Collection	Spoiled shrimp under MAP	1	−
Salmonella enterica	CIP 81.3	Feces	1	−
Serratia proteamaculans	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Serratia quinivorans	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Serratia sp.	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−
Shewanella putrefaciens	ATCC 8071	Surface of tainted butter	1	−
Shewanella baltica	Oniris/Ifremer Collection	Spoiled salmon fillet under MAP	1	−
S. baltica	Oniris/Ifremer Collection	Spoiled shrimp under MAP	1	−
Staphylococcus aureus	ATCC 25923 CIP 76.25	Clinical isolate	1	−
Vibrio anguillarum	CIP 63.36T	Ulcerous lesion of cod (Gadus morhua)	1	−
Vibrio fischeri	DSM 2168	Fish	1	−
Vibrio orientalis	CIP102891T	Seawater	1	−
Vibrio parahaemolyticus	CIP 75.2T	Shirasu food poisoning	1	−
Vibrio sp.	Oniris/Ifremer Collection	Spoiled shrimp under MAP	1	−
Vagococcus fluvialis	CIP 102976T	Chicken feces	1	−
Vagococcus fluvialis/carniphilus	Oniris/Ifremer Collection	Trout	1	−
Yersinia intermedia	Oniris/Ifremer Collection	Spoiled salmon steak under MAP	1	−

^a+, amplification product with a T_m of 84°C obtained in the method's range of use; −, no PCR product obtained in the method's range of use.

Photobacterium isolates from culture collections were grown in BHI (Biokar) with 2.0% NaCl between 15°C and 25°C depending on the strain and for 24 to 48 h. Other strains were grown in BHI (Biokar) at 20°C for 24 to 48 h except Escherichia coli ATCC 25922/CIP 76.24, Enterococcus faecalis CIP 105042, Enterococcus faecium CIP 106742, Staphylococcus aureus ATCC 25923/CIP 76.25, and Salmonella enterica CIP 81.3, which were grown at 37°C, and Listeria innocua CIP 80.11^T, Listeria monocytogenes CIP 78.35, and Pseudomonas fluorescens CIP 69.13^T, which were grown at 30°C. Vibrio isolates from culture collections were grown in marine broth (Difco Laboratories, Detroit, MI) at 20°C for 24 to 48 h except for Vibrio parahaemolyticus, which was grown at 37°C. To validate primer specificity, a total of 27 P. phosphoreum isolates were tested as targets for inclusivity tests while 54 non-P. phosphoreum isolates belonging to 49 different species were used for exclusivity testing (Table 1). These 81 strains were tested in vitro using genomic DNA extracted from 1.5 ml of bacterial cultures at a concentration of at least 7 log CFU/ml (data not shown). The chromosomal DNA of all bacterial isolates was extracted using the Qiagen DNeasy blood and tissue kit (Qiagen, S.A., Courtaboeuf, France).

Extraction of bacterial DNA from salmon flesh.

For each salmon sample, a 30-g portion was aseptically weighed in a sterile stomacher plastic bag and 120 ml sterile peptone water (0.85% NaCl, 0.1% peptone) was added to obtain a 5-fold dilution which was then homogenized for 2 min using a stomacher 400 (Seward Medical, London, United Kingdom). In order to separate the eukaryotic cells and DNA of salmon from the bacterial cells, 9 ml of homogenized suspension (previously treated with PMA for naturally contaminated samples) was filtered on a NucleoSpin Plant L column (Macherey-Nagel, Hoerdt, France) by centrifugation at 11,700 × g for 10 min at 4°C. The supernatant was carefully removed, and the bacterial cell pellet was resuspended in 400 μl of enzymatic lysis solution (20 mM Tris-HCl [pH 8.0], 2 mM sodium EDTA, 1.2% Triton X-100, 20 mg/ml lysozyme, 11.6 U mutanolysin) in a 2-ml microtube and then incubated at 37°C for 1 h. Reagents used in this step were purchased from Sigma-Aldrich. A mechanical lysis step was then performed by adding about 0.2 g of glass beads (150 to 200 μm diameter) and shaking twice for 2 min in a bead beater MM200 (30 Hz) (BioSpec Products, Bartlesville, OK). Proteins were digested by proteinase K (20 mg/ml) during incubation at 56°C for 30 min and with 200 μl of AL buffer (DNeasy blood and tissue kit; Qiagen). To pellet the glass beads, centrifugation at 9,500 × g for 3 min was carried out and the supernatant was transferred into a 2-ml microtube to perform nucleic acid precipitation by addition of 200 μl of ice-cold absolute ethanol. DNA purification was carried out using a DNeasy blood and tissue kit as described in the Qiagen instruction manual.

PMA treatment of samples.

Prior to DNA extraction, a preliminary treatment with PMA (Biotium, Hayward, CA) was carried out in order to reduce amplification of DNA from dead cells. PMA was dissolved in 20% dimethyl sulfoxide to provide a stock solution of 2.5 mM. To obtain a final PMA concentration of 50 μM in samples, 20 μl of the PMA stock solution was added to 980 μl of P. phosphoreum pure culture or 180 μl of the PMA stock solution was added to 8,820 μl of homogenized and naturally contaminated salmon. The tubes were then shaken and placed in the dark at room temperature for 5 min to allow the PMA to penetrate the dead cells and bind to DNA. Finally, to photoactivate the PMA, the tubes were placed in crushed ice with their lids removed and exposed for 5 min to light from a 500-W halogen lamp positioned at a distance of 20 cm. They were shaken occasionally to ensure complete cross-linking of the available DNA (24).

Artificially contaminated pieces of salmon flesh.

Salmon flesh was prepared and ionized as previously described by Macé et al. (29). A 60-g portion of ionized raw salmon cubes was aseptically weighed into a sterile plastic bag, and 240 ml of sterile physiological saline solution (0.85% NaCl, 0.1% peptone) was added to obtain a 5-fold dilution. The suspension was homogenized for 2 min in a stomacher 400. Eight sterile plastic tubes were filled with 25 ml of the sterile salmon homogenate. Seven of them were inoculated with 1 ml of 10-fold serially diluted suspensions of P. phosphoreum CIP 102511, previously treated with PMA, and one noninoculated tube was used as a control. The concentrations of P. phosphoreum in the salmon samples were prepared in order to range from approximately 7.7 to 1.7 log CFU/g. DNA was extracted from each homogenate as described above. Concentrations of bacteria in the homogenates were determined by real-time PCR and viable counting in duplicate using marine agar (MA; Difco Laboratories, Detroit, MI) incubated at 15°C 7 days. These conditions were used for P. phosphoreum growth in a previous study (30). The viable counts were expressed as the arithmetic mean of the two plate counts in log CFU/g. This experiment was repeated three times independently.

Primer design and real-time PCR assay.

Alignment of the gyrase subunit B (gyrB) gene DNA sequence from P. phosphoreum and closely related species was performed using the CLC DNA Workbench 6.5 (CLC bio, Aarhus, Denmark) and BioEdit sequence alignment software (31) (Fig. 1). Primers specific to the P. phosphoreum species group were designed in a conserved region so that they excluded other Photobacterium spp. and the closely related Vibrio fischeri (Fig. 1). The forward primer MO627 (5′-TACTGTTGAAGTGGCGAT-3′) matched positions 461 to 478, and the reverse primer MO628 (5′-TCTGCTGGGCTTTCTAAT-3′) matched positions 794 to 811 of the P. phosphoreum type strain gyrB gene (ATCC 11040^T/CIP 102111^T, accession number AY455875.1). The primers amplified a 350-bp specific fragment. Primer specificity was tested in silico using the Basic Local Alignment Search Tool (BLAST) program (Fig. 1) and Primer BLAST (32). The hybridization temperature of the primers was optimized using the temperature gradient test on a Chromo4 real-time PCR detection system (Bio-Rad Laboratories, Marnes-La-Coquette, France). Real-time PCR amplification was performed using the IQ SYBR green Supermix (Bio-Rad Laboratories, Marnes-La-Coquette, France) in 15-μl reaction volumes containing 7.5 μl of 2× IQ SYBR green Supermix; 2.5 μM each primer (Fisher Bioblock Scientific, Illkirch, France) and 2 μl of template DNA or 1 ng for the specificity test on the pure strains, which are listed in Table 1. The amplification reaction was conducted using the Chromo4 real-time PCR detection system (Bio-Rad Laboratories). The optimum cycling parameters were as follows: 94°C hold temperature for 3 min for initial denaturation, 35 cycles of amplification (94°C for 30 s, 58°C for 1 min, and 72°C for 1 min), followed by a final extension at 72°C for 5 min. A melting curve between 70°C and 90°C was determined after the last amplification cycle and at a temperature transition rate of 0.5°C/s. The background limit was set at 0.02 when the cycle threshold (C_T) values were determined. All amplification reactions were run in triplicate in three independent assays. The P. phosphoreum-specific PCR product had a melting temperature (T_m) of 84°C, and a response was considered positive only when this was observed. Real-time PCR assays were interpreted as positive for P. phosphoreum if the determined C_T values were ≤27.5 for DNA from pure cultures or ≤31 for DNA from inoculated fish samples. Higher C_T values were considered out of the range for this assay. These cutoff values were determined as the lowest observed C_T for the non-P. phosphoreum isolates when evaluating specificity (Table 1).

Fig 1.

Sequence alignment of the gyrase subunit B gene (gyrB) from Photobacterium species and Vibrio fischeri. The accession numbers of the sequences used for the alignment are shown. Sequences used for primer design are boxed.

Standard curves and amplification efficiency.

Two different calibration curves were constructed to determine the sensitivity and efficiency of the real-time PCR method. A standard curve was obtained by using purified genomic DNA from P. phosphoreum CIP 102511^T, extracted from serial dilutions of a bacterial suspension. The culture was diluted in sterile peptone water (0.85% NaCl, 0.1% peptone) to obtain concentrations ranging from 1 to 8 log CFU/ml.

The other standard curve was obtained from bacterial DNA extracted from independent triplicate pieces of salmon flesh previously inoculated with 10-fold serially diluted concentrations of the P. phosphoreum type strain. The different bacterial concentrations (log CFU/g) were plotted against the corresponding C_T values. The slope of the linear relation of this curve was used to determine the amplification efficiency (E) by applying the following equation: E = 10^−1/slope − 1 (33).

Comparison of real-time PCR and plate counting for quantification of P. phosphoreum in naturally contaminated salmon.

A total of 13 samples of MAP salmon steaks with a headspace gas composition of approximately 50% CO₂ and 50% N₂was obtained from a local French processing plant or purchased from a local supermarket. The samples were stored at 4°C for 3 days and then at 8°C for 7 days after a break of 2 h at 20°C, according to a French shelf life protocol (34). At different times during the storage period, bacterial DNA was extracted from salmon samples and P. phosphoreum was quantified using the real-time PCR method described above. The C_T values obtained by real-time PCR were used to calculate the concentration (log CFU/g) present in each sample, based on the standard curve established with artificially contaminated pieces of salmon flesh. All the real-time PCRs were carried out in triplicate, and the mean values were used. Results were compared to viable counts on MA (Difco), followed by determination of the ratio of P. phosphoreum bacteria in the samples. This was done by testing about 20 colonies per countable plate using real-time PCR (see below). The concentration of P. phosphoreum was then calculated by multiplying the viable counts on MA and the proportion of P. phosphoreum isolates for each tested sample. The correlation between the concentrations obtained by plate counts and real-time PCR quantification of P. phosphoreum was determined by linear regression.

Identification of bacterial isolates from MA plates.

MA is a nonselective medium used for total viable counting so identification was required to determine the proportion of P. phosphoreum colonies on this medium for each naturally contaminated sample. From each of the 13 MAP salmon samples, about 20 colonies were randomly selected from MA plates of the highest dilution showing growth. The 290 isolates collected were purified twice on BHI agar (Biokar) supplemented with 2.0% NaCl. In order to determine whether or not they belonged to the P. phosphoreum species group, all 290 isolates were analyzed by the real-time PCR assay described above and following extraction of chromosomal DNA using the Qiagen DNeasy blood and tissue kit (Qiagen). They were thus classified as P. phosphoreum or non-P. phosphoreum isolates. To confirm these results, from the 290 isolates, 50 from 4 different salmon samples were identified by partial 16S rRNA gene sequencing as described in a previous study (11). In brief, the partial nucleotide sequence (about 600 bp) of the amplified 16S rRNA gene was determined with an automated sequencer (Beckman Coulter Genomics, Takeley, United Kingdom) using the primer SP1 (35) for the 50 strains. The resulting sequences were then submitted to BLAST, available at the National Center for Biotechnology Information (NCBI; Bethesda, MD) (http://www.ncbi.nlm.nih.gov/), for representation of sequence and similarity searches in the GenBank database.

Comparison of real-time PCR with a conductance method for quantification of P. phosphoreum in naturally contaminated salmon.

A total of 26 MAP salmon steaks with a headspace gas composition of approximately 20% CO₂ and 80% N₂ was obtained from a Danish processing plant. The samples were divided into two batches which were stored at 2°C or 4°C for 9 days. Each day, the concentration of P. phosphoreum was determined by using both an available specific conductance method (19) and the new real-time PCR method described above. An Mx3000P thermocycler (Stratagene, AH Diagnostics, Aarhus, Denmark) was applied, and a standard curve was developed for the real-time PCR method when used with this instrument. The correlation between the results obtained by the two methods was determined by linear regression.

Linear regression and statistical analysis.

Linear regression was performed using Fig.P software (version 2.98; Biosoft, Cambridge, United Kingdom). Analysis of variance (ANOVA), the Fisher least significant difference (LSD) test (95.0%), the Kruskal-Wallis test, and the Mood test on medians were all performed on the nonpairwise P. phosphoreum concentration data obtained from the real-time PCR and viable count methods on naturally contaminated samples. ANOVA was used to compare variances within and between results obtained from the two methods for 13 samples. The Fisher LSD test (95.0%) was used to measure variability in samples by displaying the standard deviation of means, the Kruskal-Wallis test was used to compare medians, and the Mood test was used to analyze the distribution of data around medians. These tests were carried out using STATGRAPHICS Centurion XVI version 16.1.11 software (Statistical Graphics Corp.).

RESULTS

Specificity of the real-time PCR assay.

Using Primer BLAST (32), the selected primers, MO627 and MO628, targeted only P. phosphoreum-related species, and by using the BLASTn program, they showed 100% maximum identity with P. phosphoreum-related species. The primers developed were specific to P. phosphoreum-related species, and all 27 culture collection strains studied were correctly identified (Table 1). The threshold of the assay was set at a fluorescent background value of 0.02, resulting in C_T values ranging from 9 to 14 for the 27 target isolates, with a C_T mean value of 10.3 ± 2.3. The melting temperature calculated at the end of each real-time PCR assay was 84°C. C_T values above 27.5 were obtained for all 54 non-P. phosphoreum isolates, and these values were considered out of the specificity range. The C_T threshold obtained was 27.5 for DNA from pure cultures and 31.0 for bacterial DNA extracted from salmon flesh.

The specificity of these primers, when tested against a large collection of nontarget bacteria (Table 1) including other species of Photobacterium and other bacterial species frequently encountered in seafood products, shows that they can be used for specific detection of P. phosphoreum.

Sensitivity and quantification limits of the assay.

The detection and quantification limits of the PCR assay when used for pure cultures were determined by analyzing the genomic DNA of 10-fold serial dilutions of a P. phosphoreum CIP 102511^T overnight culture at 15°C. The linear relation determined over a 5-log CFU/ml range, from 2.2 to 7.2 log CFU/ml, was C_T = −3.15 · viable count (log CFU/ml) + 34.6 (R² = 0.999), which indicates an efficiency of 108% (Fig. 2). The quantification was considered specific up to a C_T value of 27.5 (specificity limit described above), corresponding to 2.3 log CFU/ml.

Fig 2.

Standard curves for real-time PCR quantification of P. phosphoreum diluted in physiological saline solution (□) or in artificially contaminated salmon (●). Viable counts were determined using MA. The average efficiencies of the real-time amplification were 108% with physiological saline and 112% with artificially contaminated salmon.

Sensitivity was then investigated by using sterile salmon flesh artificially contaminated with decreasing amounts of P. phosphoreum CIP 102511^T. The linear quantification (R² of 0.998) was determined from 3 to 8 log CFU/g, achieving about 5 logs of magnitude. The equation was C_T = −3.06 · viable count (log CFU/g) + 40.73, corresponding to an efficiency of about 112% (Fig. 2).

On this salmon model, the quantification was specific with a limiting C_T value of 31 enabling P. phosphoreum to be enumerated in concentrations between 3 and 8 log CFU/g.

Bacterial counts in noninoculated ionized salmon flesh were below the enumeration threshold (0.7 log CFU/g) (data not shown). Thus, ionized flesh was validated as not contaminated.

Enumeration of P. phosphoreum in naturally contaminated MAP salmon steaks.

A total of 13 MAP salmon steaks was analyzed to confirm the accuracy of real-time PCR for P. phosphoreum quantification. P. phosphoreum was detected in the range of 3.3 to 7.7 log CFU/g by real-time PCR (Table 2). Using MA plate counting and colony identification, P. phosphoreum viable counts varied from 2.9 to 7.6 log CFU/g for the 13 samples analyzed (Table 2). A good correlation was obtained (R² = 0.963) between real-time PCR and enumeration of P. phosphoreum by viable counting on MA followed by identification of colonies (Fig. 3). The root mean square error (RMSE) was 0.34 log CFU/g. Compared to the accuracy of plate counting on MA, the relative accuracy of P. phosphoreum detection by the real-time PCR method varied from 87.7 to 115.3% (Table 2). The ANOVA, Fisher test, Kruskal-Wallis test, and Mood test statistical analyses revealed that the differences between plating and real-time PCR results were not significant (P > 0.05). The suitability of the real-time PCR assay for identifying P. phosphoreum isolates was also confirmed by partial sequencing of the 16S rRNA gene for 50 isolates from 4 of the 13 salmon samples. The identification results were exactly the same for the two methods, confirming the specificity of real-time PCR for P. phosphoreum detection (Table 3).

Table 2.

Viable counts on MA and real-time PCR enumeration of P. phosphoreum in naturally contaminated MAP salmon steaks (log CFU/g) and their relative accuracy

Sample	Real-time PCR (log CFU/g)	CT valuesa	Viable count (log CFU/g)b	Relative accuracy (%)c
1	3.32	30.57 ± 0.56	2.88	115.3
2	5.37	24.29 ± 0.77	5.49	97.8
3	7.13	18.89 ± 0.14	7.03	101.4
4	7.34	18.27 ± 0.17	7.38	99.5
5	3.47	30.11 ± 0.10	3.57	97.2
6	3.85	28.93 ± 0.47	4.42	87.1
7	7.06	19.11 ± 0.30	7.06	100
8	7.66	17.28 ± 0.20	7.61	100.7
9	7.47	17.85 ± 0.21	7.31	102.2
10	3.84	28.96 ± 0.19	4.54	84.7
11	5.36	24.32 ± 0.16	5.60	95.6
12	7.40	18.06 ± 0.17	7.20	102.8
13	6.76	20.02 ± 0.32	7.29	92.8

^aMean and standard deviation of three C_T values.

^bViable counts were determined using MA and identification of colonies.

^cDegree of correspondence between the results obtained with a conventional plating technique and with a real-time PCR assay. The relative accuracy is expressed as the percentage of numbers of log (CFU/g) calculated by the real-time PCR assay versus the conventional method (45).

Fig 3.

Correlation between concentrations of P. phosphoreum determined by real-time PCR and viable counting in 13 samples of naturally contaminated fresh MAP salmon steaks.

Table 3.

Identification of 50 isolates from MA plates using partial 16S rRNA gene sequencing and P. phosphoreum identification by real-time PCR method

Sample	Isolate	Species or genus identification by partial 16S rRNA gene sequencing	P. phosphoreum identification using real-time PCR methoda
1	1	Serratia fonticola	−
	2	Pseudomonas sp.	−
	3	Pseudomonas sp.	−
	4	Pseudomonas sp.	−
	5	Pseudomonas sp.	−
	6	Acinetobacter sp.	−
	7	Pseudomonas sp.	−
	8	Pseudomonas sp.	−
	9	Pseudomonas sp.	−
	10	Pseudomonas sp.	−
	11	Shewanella sp.	−
	12	Pseudomonas sp.	−
	13	Myroides odoratimimus	−
	14	Acinetobacter sp.	−
	15	Nonworkable sequence	−
	16	Shewanella sp.	−
	17	P. phosphoreum	+
	18	Nonworkable sequence	−
	19	Pseudomonas sp.	−
	20	Pseudomonas sp.	−
2	1	P. phosphoreum	+
	2	P. phosphoreum	+
	3	P. phosphoreum	+
	4	P. phosphoreum	+
	5	P. phosphoreum	+
	6	P. phosphoreum	+
	7	P. phosphoreum	+
	8	P. phosphoreum	+
	9	P. phosphoreum	+
	10	Shewanella sp.	−
3	1	P. phosphoreum	+
	2	P. phosphoreum	+
	3	Nonworkable sequence	+
	4	P. phosphoreum	+
	5	P. phosphoreum	+
	6	Shewanella sp.	−
	7	P. phosphoreum	+
	8	Shewanella sp.	−
	9	P. phosphoreum	+
	10	P. phosphoreum	+
4	1	Shewanella sp.	−
	2	Nonworkable sequence	+
	3	Nonworkable sequence	+
	4	P. phosphoreum	+
	5	P. phosphoreum	+
	6	P. phosphoreum	+
	7	P. phosphoreum	+
	8	P. phosphoreum	+
	9	P. phosphoreum	+
	10	Photobacterium sp.	+

^a+, amplification product with a T_m of 84°C obtained in the method's range of use; −, no PCR product obtained in the method's range of use.

Comparison of real-time PCR and a conductance method for quantification of P. phosphoreum in fresh MAP salmon.

P. phosphoreum was detected in the range of 2.5 to 8.5 log CFU/g by real-time PCR and of 1 to 8 log CFU/g by the conductance method. In one of the 26 samples tested, P. phosphoreum was not detected by either of these two methods. Three samples presenting P. phosphoreum concentrations from 1 to 2 log CFU/g were quantified by the conductance method only (data not shown). One sample was quantified by the real-time PCR method, but the level of bacteria was out of range, with a concentration of about 2.5 log CFU/g. However, this count was well correlated with that of the conductance method (2.7 log CFU/g), so this result was used for the correlation curve. Based on data for the remaining 22 samples, a high correlation with an R² of 0.940 and an RMSE of 0.49 log CFU/g was obtained (Fig. 4).

Fig 4.

Correlation curve between P. phosphoreum enumeration by the specific conductance method and real-time PCR on 22 samples of naturally contaminated MAP salmon steaks.

DISCUSSION

Microbiological quality control of fresh seafood products remains a challenge due to their high sensitivity to microbial contamination and growth. By improving good hygiene practices in processing plants and by using MAP, the shelf life of fresh fish can be increased by several days (2, 4). The detection and quantification of SSO can support and document efforts with good hygiene practices and thereby contribute to increasing quality and reducing product losses. In recent years, culture-independent molecular methods like real-time PCR have been increasingly applied to detect and quantify target microorganisms in food (36, 37). Several real-time PCR methods have been developed to enumerate or detect pathogenic bacteria in seafood products, including Vibrio parahaemolyticus (26, 38) and mesophilic Gram-negative histamine-producing bacteria (39, 40). Spoilage bacteria like Pseudomonas (41) and Brochothrix thermosphacta (24) have also been enumerated by these methods. In this work, real-time PCR was chosen for our objective of quantifying the principal SSO of MAP fish, P. phosphoreum, since the specific enumeration of this bacterium is difficult using classic methods. To our knowledge, this is the first report to present a rapid molecular quantification tool for P. phosphoreum using real-time PCR. Moreover, the use of PMA allowed exclusive quantification of viable cells (21, 42) and, consequently, only the P. phosphoreum cells active in spoilage.

When real-time PCR was used previously for the enumeration of bacteria in a meat or fish matrix, linear quantifications were reported over a range of at least 5 logs, and quantification limits without enrichment varied from about 2 to 3 log CFU/g (22, 24, 25, 37, 43, 44), as obtained in the present study.

When enumeration by viable counting and that by real-time PCR are compared, a relative accuracy of 100% indicates total agreement between the two methods. The relative accuracy values obtained in the present study for the naturally contaminated salmon samples (88 to 115%) are better than those reported in several earlier studies for other bacteria (24, 25) and similar to those obtained by Rodríguez-Lázaro et al. (45) for the quantification of Listeria monocytogenes (89 to 116%). Martin et al. (43) presented better relative accuracy and higher R² values for Lactobacillus sakei with less variability (91 to 98%), but these results were based on only 7 tested samples, compared to 13 in the present study. The correlation curve obtained here on 13 samples gave an R² value of 0.963, which is higher than the correlation coefficients reported for several real-time PCR methods applied to more than 10 samples of meat or seafood products (R² values ranging from 0.798 to 0.911) (22, 24, 25, 41). This confirms the reliability of our new real-time PCR method dedicated to the quantification of P. phosphoreum.

The accuracy of this method was validated by comparison with the available specific conductance method. In fact, the correlation curve obtained by plotting the quantification results of the 22 tested MAP salmon samples showed a high R² of 0.940. The RMSE values obtained by comparing real-time PCR with the viable count (0.34 log CFU/g) or conductance method (0.49 log CFU/g) are lower than the currently accepted error value of 0.5 to 0.8 log for plate counting on solid matrices (46). With a quantification limit of 3 log CFU/g, the new real-time PCR method is slightly less sensitive than the available conductance method, which detects less than 1 log CFU/g (19). However, among the naturally contaminated samples, only a few (3 out of the 26 samples tested) were out of range of the real-time PCR method and could be quantified only by the conductance method. The major advantage of the real-time PCR method is its rapidity; it provides results in less than 6 h (including DNA extraction and real-time PCR analysis), which is much faster than the available conductance method, requiring 12 to 50 h (19), or the viable counting method, requiring more than a week, as colonies must be identified.

Reducing the detection level is not required for P. phosphoreum as much as for pathogenic bacteria, since this organism induces spoilage rejection of products at a high population density, e.g., above 6.5 to 7 log CFU/g (5–7, 29). This species is also known as a histamine producer and is involved in histamine food poisoning (47, 48) due to the consumption of cold-smoked tuna, dried sardines, and fresh tuna. Nevertheless, histamine poisoning has never been reported following salmon consumption, as the histidine content of salmon flesh is too low for toxic concentrations of histamine to be formed (8, 49). The real-time PCR method could thus be adapted for P. phosphoreum quantification in other fish, including transformed products such as cooked tuna, with an appropriate DNA extraction method. Additionally, detection from 3 log CFU/g is sufficient because histamine is produced when P. phosphoreum concentrations are above about 6 log CFU/g (8, 48, 49). Consequently, this real-time PCR method should provide a powerful tool to evaluate the contamination of seafood products by P. phosphoreum before spoilage appears or histamine is formed.

However, in the case of its application in combination with predictive microbiology models, a lower detection level could be useful, for example, by employing the Seafood Spoilage and Safety Predictor (http://sssp.dtuaqua.dk), which includes models for growth of P. phosphoreum in fresh MAP cod, plaice, and salmon (50). To improve the sensitivity of the new real-time PCR method, an enrichment step in broth, prior to the PCR assay, could be used, e.g., in the same way as applied for the available conductance method for sensitive enumeration of P. phosphoreum (19). Enrichment steps in the broth or food samples have already been used to enhance the sensitivity of real-time PCR methods, for both qualitative and quantitative detection of the target bacteria (27, 44, 51). In fact, we tested this approach by using quantitative enrichment in Photobacterium phosphoreum differential medium (PPDM) at 15°C (19). Unfortunately, this enrichment medium seemed to inhibit the real-time PCR. Thus, on slightly contaminated samples with less than 3 log CFU/g of P. phosphoreum, this real-time PCR method is currently suitable only for qualitative purposes.

In conclusion, this method is now applicable to enumerating P. phosphoreum in salmon and, according to preliminary tests in progress on cod, it could also be used for other fish.

Thanks to the PMA treatment step, it could be used by industry to detect and quantify viable P. phosphoreum contamination in production plants and thus target processing steps which need to be controlled. This method could also be used to test the shelf life of newly packaged products, to quantify the growth of this very important fish spoiler, and to determine shelf life, as discussed previously for the available conductance method (4, 13).

Moreover, the development of new methods to quantify food spoilage organisms, like P. phosphoreum or other bacterial species (24, 25, 41, 51), could assist in the establishment of more accurate microbial hygienic criteria than those usually applied (e.g., total flora count). This would help to reduce food losses and to improve sustainable development in food production. Photobacterium spp. have recently been reported as part of beef and pork spoilage microbiota (52, 53). Thus, this new quantitative method could clearly be used by the scientific community as a complementary tool to study P. phosphoreum distribution and abundance in bacterial populations.