© INRA, EDP Sciences, 2010

1. INTRODUCTION

Cheese production is an efficient branch of the food industry in France; in 2008, 1.85 million tons of cheeses were produced for a turnover of more than 7 billion euros. The economic importance of raw milk cheeses, which accounted for 14% of French cheese production in 2008, is connected with their gastronomic reputation, their high added value and their role in rural development and land use.

One of the most important features of raw milks is the diversity of their microbial communities. More than 150 microbial species of lactic acid bacteria, non-lactic bacteria, Gram-negative bacteria, yeasts and fungi have been identified in raw milks, with the qualitative and quantitative balances in terms of species and strains varying from one milk to another. The microbes can generate a variety of flavouring compounds, generating all the diversity and richness of sensorial properties of raw milk cheeses, quite different than those of cheeses made from pasteurised or microfiltered milk [2, 6]. It cannot be denied that some pathogenic bacteria can contaminate milk or cheese. Nevertheless, the health risk with raw milk production is low, as witnessed by the low number of toxi-infections associated with the consumption of raw milk cheeses [9, 10]. The microbial safety of these cheeses depends on hygiene measures, rigorous control and application of procedures adapted from Hazard Analysis and Critical Control Point.

Microbial diversity may help to improve the safety of raw milk cheeses. In fact microbial life, through a large number of interactions between populations, is at the heart of “hurdle technology”. This technology involves microbes creating a succession of barriers such as low pH, inhibitory metabolites (organic acids, ethanol, etc.), peptides (bacteriocins) and nutritional competition. Hurdle technology is particularly effective for controlling pathogenic bacteria when combined with environmental barriers (temperature, water content of the cheese, etc.). Listeria monocytogenes, for example, may be inhibited on the surface of washed rind cheeses [13, 25]. Some microbial populations from raw milk cheeses (Saint-Nectaire [25] or Camembert [15]) have been shown to inhibit L. monocytogenes. The growth of L. monocytogenes has been shown to be highly dependent on the raw milk flora in soft cheeses and was also lower in a raw milk than in a sterilised one [5]. Lactic acid bacteria [29], Corynebacteria such as Brevibacterium linens, yeasts and fungi can inhibit L. monocytogenes [8, 16, 19, 30], mainly by secreting bactericidal and bacteriostatic substances. However, none of these studies clearly demonstrated how microbial populations interact in the inhibition. The aims of the present study were (i) to evaluate the antilisterial activities of complex microbial consortia from the surface of PDO Saint-Nectaire raw milk cheeses and (ii) to try to constitute microbial consortia with similar antilisterial activities to the complex one selected in a first step.

2. MATERIALS AND METHODS

2.1. Preparation of undefined complex consortia from cheese surfaces

Thirty-four Saint-Nectaire raw milk cheeses named TR1–TR20 and TR22–TR35, prepared in different farms and ripened in different cellars for at least 18 days, were obtained from ripeners in the Saint-Nectaire area. A quarter of each cheese’s rind (cheese surface to a thickness of 3 mm) was sampled then blended in phosphate buffer (20 mmol·L⁻¹ KH₂PO₄, 0.01 mol·L⁻¹ K₂HPO_4, Gomri, 1946) pH 7.5 (diluted 1 volume of sample to 10 volumes of buffer), two times for 2 min, using a stomacher blender (Interscience, Saint-Nom-la-Bretèche, France). This homogenised rind suspension, referred to below as a “complex consortium”, was aliquoted and frozen at −20 °C.

2.2. Bacterial strains and microbial community isolates

2.2.1. Listeria monocytogenes

The purity of L. monocytogenes strain S1 [25] was checked on ALOA agar plate medium then precultured in TS-YE broth medium. After 18 h of incubation at 37 °C, the culture was centrifuged for 15 min at 5000 rpm and 4 °C and the supernatant discarded. The pellet was homogenised in sterile reconstituted skim milk supplemented with 15% glycerol and stored at −20 °C until its inoculation onto the cheese surfaces.

2.2.2. Characterisation of microbial isolates from complex cheese surface consortium TR15 cultivated on cheese

2.2.2.1. Collection of bacterial and yeast isolates from TR15 cultivated on cheese

In order to constitute simplified defined consortia that are able to inhibit L. monocytogenes, a collection of bacterial and yeast isolates was created from the complex consortium TR15, cultivated on cheese. TR15 was inoculated onto the surfaces of cheeses that were made from pasteurised milk. The inoculated cheeses were ripened for 28 days at 9 °C and 95% relative humidity (RH) in the ripening room at INRA. The micoorganisms able to establish themselves on the cheese surface were enumerated on various agar media at seven different ripening times (1, 8, 11, 18, 21, 25 and 28 days). The agar media presenting the highest number of colonies with different morphotypes were observed for samples taken at 11, 18, 21 and 28 days of ripening. The media presenting the highest diversity in terms of morphotypes were selected from these ripening times. From these agar media, 111 colonies, representative of each of the morphotypes observed, were isolated and purified before identification at species level (Tab. I).

2.2.2.2. Total DNA extraction

Total DNA from each isolate cultivated in an appropriate broth medium was extracted using Easy-DNA kit with phenol/chloroform (Invitrogen, Cergy-Pontoise, France) according to the manufacturer’s instructions.

The ribosomal 16S rRNA genes (1450 bp) from all isolates were amplified by polymerase chain reaction (PCR) using the universal primers W02 and W18 (complete gene for bacterial strains – Tab. II) as described by Callon et al. [7].

2.2.2.3. Identification of bacterial isolates

The 16S rRNA genes of the isolates were screened by restriction fragment length polymorphism (RFLP) as described by Callon et al. [7]. The 16S rRNA RFLP profiles of the 111 isolates were compared using the BioNumerics software and the unweighted pair group method with arithmetic averages to obtain a dendrogram. Patterns showing > 90% similarity (calculated by Pearson’s correlation) were grouped in the same cluster (G1–G12, see Tab. I). Each cluster was assigned to a bacterial genus by comparing the profiles with those of a reference database. The reference database was composed of 109 different species from the laboratory collection, including Gram-positive lactic acid bacteria (27 strains), non-lactic Gram-positive bacteria (56 bacteria including 19 Staphylococci) and Gram-negative bacteria (26). The 16S rDNA of 1–7 isolates representing each restriction pattern (G1–G12) was sequenced or amplified with specific primers.

Specific sequences of 16S rRNA genes from isolates which were identified as lactic acid bacteria according to their RFLP profiles were amplified using specific PCR primers as described in Table II.

Twenty-five microlitres of bacterial 16S rRNA gene PCR amplification products were sequenced using W34 primer by GeneCust (Évry, France). The sequences were compared to the sequences available in the GenBank database using the Blast program. In this study, a 99% similarity was taken to assign an isolate to a species.

2.2.2.4. Identification of yeast isolates

Yeasts were identified by a combination of phenotypic tests and sequencing of the D1/D2 domain of the 26S rRNA gene as described by Callon et al. [7].

2.2.2.5. Antilisterial effect of microbial isolates

The antilisterial activity of each isolate in the collection was tested using the agar-well diffusion method as described by Saubusse et al. [28].

2.2.2.6. Creating defined microbial consortia

From the collection of strains from complex consortium TR15 that were able to grow on cheese surfaces, four different defined consortia were made up. The most complex, called ABCD, was composed of 19 strains randomly selected for each microbial species identified (Tab. I): six species of Gram-positive lactic acid bacteria (Group A), seven species of Gram-positive and catalase-positive bacteria (Group B), three species of Gram-negative bacteria (Group C) and three species of yeasts (Group D). Each group was composed of one strain of each species as follows: Group A, Lactobacillus casei/paracasei, Lactobacillus curvatus, Leuconostoc mesenteroides, Enterococcus faecalis, Marinilactibacillus psychrotolerans and Carnobacterium mobile; Group B, Arthrobacter ardleyensis or bergerei, Arthrobacter nicotianae or arilaitensis, Staphylococcus xylosus, Staphylococcus pulvereri, Brevibacterium antiquum or casei, Brevibacterium linens or casei and Brachybacterium sp.; Group C, Pseudomonas fluorescens or syringae, Serratia proteomaculans or liquefaciens and Proteus vulgaris; Group D, Candida sake, Debaryomyces hansenii and Yarrowia lipolytica. The strains could not be selected for their antilisterial properties as none of the isolates identified was able to inhibit L. monocytogenes by agar-well diffusion test in vitro. The three other defined consortia, ABC, ABD and ACD, were defined by omitting one group: either B, C or D.

Each isolate was purified on the appropriate agar plate medium and cultivated in 10 mL of the appropriate broth medium. The culture was centrifuged for 15 min at 8500 rpm and 4 °C. The pellet was homogenised in sterile reconstituted milk supplemented with 15% glycerol and with or without 5% ascorbate. Suspensions were transferred to 2 mL microtubes and stored at −20 °C until their inoculation onto the cheese surfaces. Before inoculation, suspensions were thawed in a water bath for 5 min at 25 °C, counted on agar media and diluted to inoculate the appropriate concentrations on the cheese surfaces.

2.3. Challenge test with L. monocytogenes on the surface of an uncooked pressed type cheese

2.3.1. Cheese manufacturing

Small cheeses (600 g) were made from pasteurised milk collected at a school farm (ENILV, Aurillac, France), using an uncooked pressed cheese method similar to that of Saint-Nectaire. Milk was pasteurised at 72 °C for 30 s, cooled to 33 °C then inoculated with 0.6% of a commercial starter culture (MY800, Streptococcus thermophilus; Lactobacillus delbrueckii spp. bulgaricus, Danisco, Sassenage, France) and with a commercial mould culture of Penicillium nalgiovensis (2.5 mL·200 L⁻¹, Laboratoire Interprofessionnel de Production – LIP, Aurillac, France). Forty-five minutes after adding 40 mL of calf rennet (Beaugel 520-Ets Coquard, Villefranche-sur-Saône, France), the curd was cut and gently stirred to eliminate whey. The curds were moulded and pressed for 24 h at 2.1 bar and 22 °C. After pressing, the cheeses (pH 5.14) were vacuum-packed and frozen at −20 °C until inoculation.

2.3.2. Inoculation of experimental cheese surfaces

The experimental cheeses made from pasteurised milk were thawed for 5 h at room temperature under a laminar hood and turned over each hour to remove excess water. In all trials, their surfaces were inoculated by depositing and brushing with a sterile toothbrush 1 mL of different microbial consortia containing strain 167 of L. monocytogenes. L. monocytogenes was inoculated at 2–5 colony-forming units (CFU)·cm⁻². The microbial consortia were inoculated in three sets of experiments.

In the first set of experiments, 34 different microbial consortia were inoculated; corresponding to each of the complex cheese microbial consortia named TR1–TR20 and TR22–TR35, prepared as described in Section 2.1 and thawed in a 25 °C water bath before inoculation.

In the second set of experiments, consortium TR15, the most inhibitory of the 34, was inoculated after 3 and 24 months of storage at −20 °C in order to verify the stability of the antilisterial activity.

In the third set of experiments, five different microbial consortia were inoculated: one TR15 consortium (selected in the first set of experiments and stored at −20 °C) and four defined microbial consortia (ABCD, ABC, ACD and ABD). The inoculum level of each species was based on counting results obtained after growing TR15 on cheese. Each species of Groups A, B and D was inoculated at 2.7 log CFU·cm⁻². The Gram-negative bacteria (Group C) were inoculated at only 1 log CFU·cm⁻².

Each set included a control, inoculated with 1 mL of a commercial ripening culture frequently used for manufacturing Saint-Nectaire cheese (Penicillium nalgiovensis 3.10⁴ CFU·cm⁻², LIP, Aurillac, France).

The inoculated cheeses were ripened for 28 days in sterile stainless steel boxes in INRA’s cellars at 8–9 °C, 98% RH (in the cellar). After 8 and 18 days of ripening, the cheeses were washed with sterile salt water (20% NaCl).

2.4. Microbial analysis

Rind samples 3 mm thick were taken from inoculated cheeses after 1, 8, 18 and 28 days of ripening, for microbial analysis.

L. monocytogenes was counted as prescribed in ISO 11 2090-2, by an accredited laboratory (LIAL, Aurillac, France).

The rind samples were blended in phosphate buffer (20 mmol·L⁻¹ KH₂PO₄, 0.01 mol·L⁻¹ K₂HPO₄) pH 7.5 using a stomacher blender (Interscience, Saint-Nom-la-Bretèche, France). Each suspension was diluted with Ringer’s solution and appropriate dilutions were spread on agar plate media using a spiral system (DS+, Interscience, Saint-Nom-la-Bretèche, France).

As described by Millet et al. [25], the microbial populations were enumerated on the following agar plate media, according to morphotype: coagulase positive and negative Staphylococci on rabbit plasma fibrinogen agar (RPF); facultatively heterofermentative lactobacilli on facultatively heterofermentative agar (FH); dextran-producing leuconostocs on Mayeux Sandine and Elliker agar medium (MSE); enterococci on Slanetz and Bartley agar (SB); Pseudomonas on cephalosporin fucidin cetrimide (CFC) agar; Gram-positive and catalase-positive bacteria on cheese ripening bacteria medium (CRBM); Gram-negative bacteria on plate count agar (PCA) supplemented with milk and crystal violet (PCA + M + CV); yeasts and moulds on oxytetracycline glucose agar (OGA) medium. All media were purchased from Biokar.

On CRBM, white colonies corresponding to S. pulvereri, yellow or cream colonies corresponding to A. ardleyensis or bergerei, A. nicotianae or arilaitensis and Brachybacterium sp. and orange colonies corresponding to S. xylosus or B. antiquum or casei and B. linens or casei were counted specifically. On OGA medium, white colonies corresponding to D. hansenii or C. sake and cream colonies corresponding to Y. lipolytica.

Colony counts were expressed as log CFU·cm⁻² of cheese rind. Counts < 10 CFU·g⁻¹ were shown as < 1 log·cm⁻².

2.4.1. Calculation of area of inhibition

In order to easily compare the L. monocytogenes populations of the different cheeses during ripening (second set of experiments), the area of inhibition (AI) between two ripening days (t1 and t2) was calculated using the following formula as described by Saubusse et al. [28]$$\mathrm{AI}=(t2-t1)/2\times \left[\right(\mathrm{Ct}1+\mathrm{Ct}2)-(\mathrm{Tt}1+\mathrm{Tt}2\left)\right],$$ where C is the count of L. monocytogenes in control cheeses whose surfaces were inoculated with a commercial ripening culture and T is the count of L. monocytogenes in cheeses whose surfaces were inoculated with other microbial consortia.

2.5. Single strand conformation polymorphism analysis of the cheese rinds

Total genomic DNA was extracted from 1 g cheese rind sample at 1, 8, 18 and 28 days of ripening, using the phenol-chloroform method described by Delbès et al. [11]. Partial 16S rRNA gene sequences were amplified using Gram1F–Gram2R primers with Gram1F labelled with hexachloro derivative of fluorescein HEX and Gram2R labelled with fluorescein phosphoramidite NED [28]. The PCR-single strand conformation polymorphism (PCR-SSCP) products were analysed by SSCP capillary electrophoresis on an ABI Prism 310 genetic analyser. The fluorescence signal was analysed using the Genescan analysis software. The SSCP peak patterns were aligned using internal standard genescan 400 Rox. The dominant peaks in the SSCP patterns obtained were assigned by comparing their migration with that of the 16 bacterial strains used to make up the microbial community. To analyse the different profiles, relative peak proportion P _i is determined as P _i = a _i × 100/Σa _i, where P _i is the proportion and Σa _i is the sum of the peak areas of the whole SSCP patterns.

2.6. Chemical analysis

2.6.1. pH measurements

The pH of cheese surface was measured at 1, 8, 18 and 28 days of ripening at three locations on each cheese with a 926 VTV pH-meter with Ingold electrode 406 MX (Mettler-Toledo S.A., Viroflay, France). The results are the means of the three measurements.

2.6.2. Analysis of sugar and organic acid contents of cheese surfaces by high pressure liquid chromatography

Suspensions of finely ground rind were prepared (10 g for organic acids and 10 g for sugar compounds) in 20 mL distilled water. After incubating for 1 h at 50 °C, the rind suspensions were homogenised using an Ultra-Turrax device at 9500 rpm for 2 min. Organic acids and sugar compounds were extracted as described by Leclercq-Perlat et al. [20]. Analysis was performed by high-performance liquid chromatography (HPLC). The HPLC system (Waters, Guyancourt, France) consisted of an automatic injector, two serial detectors (a UV detector 210 nm for organic acids and a differential refractometer for sugar compounds) and two integrators. The cation-exchange column (Aminex HPX-87H – Bio-Rad, Marnes-la-Coquette, France) was maintained at 35 °C. Sulphuric acid (0.1 mol·L⁻¹) was used as the mobile phase at a flow rate of 0.6 mL·min⁻¹. Propionic acid (1%) was used as an internal standard.

The results were expressed in grams per kilogram of dry rind content.

3. RESULTS

3.1. Inhibition of L. monocytogenes on the surfaces of uncooked pressed type cheeses

The L. monocytogenes counts on the surfaces of cheeses inoculated with the 34 complex microbial consortia from Saint-Nectaire cheeses (TR1–TR20 and TR22–TR34) and with a commercial ripening culture as a control, after different ripening times at 9 °C, are given in Table III. Results obtained for TR15 after several months of storage at −20 °C are also included.

The level of L. monocytogenes was below the detection threshold of 1 log CFU·cm⁻² after 1 and 8 days of ripening in most cheeses, whichever inoculation was used (TR1–TR20 and TR22–TR34). L. monocytogenes grew mainly between day 8 and day 18 and the L. monocytogenes count at days 8, 18 and 28 varied according to the consortium inoculated, allowing us to arbitrarily classify the cheeses in three groups. Group 1 (10 consortia) corresponded to cheeses with the lowest L. monocytogenes counts (below 5 log CFU·cm⁻²). Interestingly, one microbial consortium (TR15) from this group, tested several times, can be stored at −20 °C in phosphate buffer for at least 24 months without losing its capacity to inhibit Listeria compared to a control. In Group 2 (15 consortia), L. monocytogenes populations reached between 5.26 and 5.96 log CFU·cm⁻². Group 3 (9 consortia) had the highest L. monocytogenes counts – 6 log CFU·cm⁻² or more. The control cheeses inoculated with only the commercial ripening culture were in this group.

There was no correlation between the count of L. monocytogenes at a given ripening time and the pH values; r = 0.10 at day 18 (pH varying between 6.20 and 7.52, n = 35) and r = −0.14 at day 28 (pH varying between 5.80 and 8.40, n = 35). However, the increase in pH during ripening was associated with an increase in L. monocytogenes.

Due to its high antilisterial properties and stability in storage at −20 °C, microbial consortium TR15 was selected for the subsequent studies of antilisterial activity.

3.2. Identification of strains from culture of cheese complex consortium TR15

Ninety-five bacterial isolates were distinguished in 12 RFLP patterns (clusters G1–G12). First, by reference to a database of reference strains most commonly isolated from milks or cheeses, nine RFLP profiles (G1–G3, G6–G8 and G10–G12) were first assigned to different genera. Profiles of three clusters (G4, G5 and G9) presented no similarity with those present in the database. The 46 isolates belonging to clusters G1–G5 corresponding to lactic acid bacteria were identified according to their amplification with specific primers and the identifications were confirmed by the closest match of their 16S RNA in the GenBank database (Tab. I). Thus, 14 isolates of G1 were assigned to Lb. casei or paracasei, the other two isolates in this group to Lb. curvatus, the 12 isolates of G2 to Ln. mesenteroides or pseudomesenteroides and the 11 isolates of G3 to E. faecalis. Some representative isolates of the other groups were identified only by 16S rDNA sequencing (Tab. I). Their 16S rDNA sequences presented more than 99% homology with those of the following species: Cb. mobile (G4, two isolates), M. psychrotolerans (G5, five isolates), A. nicotianae or arilaitensis (G6, one isolate), A. ardleyensis or bergerei (G6, six isolates), B. linens or casei, B. antiquum or casei (G7, three isolates), Brachybacterium sp. (G8, one isolate), S. pulvereri, S. xylosus (G10, four isolates), P. fluorescens or syringae (G11, three isolates), Ser. proteomaculans or liquefaciens (G12, two isolates) and Prot. vulgaris (G9, four isolates).

The 16 yeast isolates were identified as C. sake (11 isolates), Y. lipolytica (three isolates) and D. hansenii (three isolates) on the basis of their phenotypic characteristics and 26S DNA sequencing (Tab. I).

3.3. Antilisterial properties on cheese surfaces of complex consortium TR15 and defined consortia

L. monocytogenes developed differently during ripening depending on the microbial consortium inoculated − complex consortium TR15 and defined consortia as shown in Figure 1. The lowest count was observed on cheese surfaces inoculated with complex consortium TR15 and the highest on those with only the commercial ripening culture.

Figure 1. L. monocytogenes growth on the surfaces of cheeses inoculated with complex consortium TR15 and defined consortia. L. monocytogenes counts expressed in log CFU·cm−2; cheese surfaces inoculated with control = commercial ripening culture, TR15 = complex consortium (rind suspension), ABCD, ABC, ABD and ACD = defined consortia from identified isolates (A = six lactic acid bacteria, B = seven Gram-positive and catalase-positive bacteria, C = three Gram-negative bacteria and D = three yeasts).

The inhibition of L. monocytogenes was then evaluated by calculating the AI as described in the Section 2. The AI data and pH values are shown in Table IV. The higher the AI values, the stronger the inhibition of L. monocytogenes. Throughout the ripening, AI values were highest in cheeses inoculated with complex consortium TR15 and lowest in those inoculated with ABCD and ABD. Between days 18 and 28, L. monocytogenes was also inhibited (positive values of AI) in the presence of ACD and to a lesser extent with ABC.

3.4. Evolution of pH

At days 1 and 2, pH values ranged from 5.52 to 6.17 (Tab. V), but nevertheless L. monocytogenes counts were below the detection limit on the surfaces of all cheeses. In all cheeses, pH dropped between days 1 and 8, ranging between 5.56 and 5.92 at day 8. The lowest pH, at day 8 in cheese ABCD (pH 5.56), was not associated with the lowest count of L. monocytogenes. Between days 8 and 18, the pH increased dramatically, reaching 7.04–7.58 at day 18 in all cheeses (Tab. V). This was matched by an increase of close to 2 log CFU·cm⁻² in L. monocytogenes counts. At day 28, the control cheese had the lowest pH value (7.04), but the highest count of L. monocytogenes.

3.5. Dynamics of microbial populations by culture-dependent methods

At all stage of ripening, it was difficult to control Gram-negative bacteria. They were present in abundance on the surfaces of all the cheeses (9.1 log CFU·cm⁻² in TR15 to 9.7 log CFU·cm⁻² in ABCD), even the ABD cheeses (8.9 log CFU·cm⁻² at 28 days) on which they had not been inoculated (Tab. VI).

At days 1 and 8, the highest lactobacilli count was on the surface of cheeses inoculated with consortium TR15, at least 1 log CFU·cm⁻² higher than in the other cheeses, except for the ACD cheeses where the counts at day 8 were similar to those for TR15 (5.6 log CFU·cm⁻²; Tab. VI).

At day 1, dextran-producing leuconostocs were detected in all cheeses inoculated with any of the five microbial consortia. As with the lactobacilli, leuconostoc counts were at least 1 log CFU·cm⁻² lower in the defined consortia than in complex consortium TR15. However, leuconostocs were counted at every ripening time on the surfaces of cheeses inoculated with TR15, where they reached 7.7 log CFU·cm⁻² at day 28. In cheeses prepared with defined consortia, they were not found on MSE medium up to day 8.

Until day 18, the trend in E. faecalis counts was similar in cheeses inoculated with complex consortium TR15 and defined consortia ABD and ACD. From day 18 to 28, they increased to 6.9 log CFU·cm⁻² in cheeses inoculated with ABD and ACD, but remained stable at 5.6 log CFU·cm⁻² in cheeses inoculated with complex consortium TR15. There was no growth of E. faecalis in cheeses inoculated with ABC (consortium without yeasts).

At day 8, S. pulvereri was the dominant population within the Gram-positive and catalase positive population in cheeses prepared with defined microbial consortia. Its level increased until day 18 and then stabilised around 8 log CFU·cm⁻². In these cheeses, the growth of Arthrobacter sp. was similar. In cheeses inoculated with complex consortium TR15 the increase in Arthrobacter sp. was > 1–1.5 log CFU·cm⁻² higher than for S. pulvereri throughout ripening. At day 28 it was difficult to count each species due to the great diversity of morphotypes on the plate (Tab. VI).

The yeast counts were similar in all cheeses and increased mainly before day 8, stabilising at 28 days at 8 log units (CFU·cm⁻²) with dominance of C. sake and D. hansenii. These two species were still the dominant populations at the end of ripening and their levels with the TR15 and ABCD consortia were similar. At 8 and 18 days counts of Y. lipolytica were higher in consortia ABD and ACD than in TR15 and ABCD.

3.6. Evaluation of bacterial balance on cheese rinds by a culture-independent method (direct SSCP) and by plate counting

Table VII shows the percentage of each bacterial population at 18 and 28 days of ripening, as assessed by the culture-independent method SSCP analysis targeting the V2 region of 16S rDNA and the culture-dependent plate counting method. Gram-negative bacteria have not been taken into account as they could not be detected by PCR-SSCP analysis of the V2 region of 16S rDNA. Moreover, data from days 1 and 8 of ripening are not considered because the SSCP V2 profiles were dominated by a single peak corresponding to St. thermophilus. Both methods, PCR-SSCP and counting, led to the conclusion that S. xylosus, B. linens or casei and B. antiquum or casei were subdominant populations as they were not detected on the cheese surfaces by either of the two methods.

PCR-SSCP analysis of V2 region was more accurate than plate counting for detecting the diversity of the bacterial populations present on the cheese surfaces. Leuconostocs were detected in all SSCP cheese profiles from day 8 to 28, but were only quantified in TR15 cheeses by the culture-dependent method. This discrepancy may be due to the leuconostoc strain selected losing its ability to produce dextrans on MSE medium, whereas strains from the natural flora were still able to produce them. Then, the two methods gave quite different pictures of bacterial community. The leuconostoc peak can represent > 70% of the peaks on SSCP profiles for cheeses inoculated with defined consortia ABC at 18 days and ACD at 18 and 28 days. With plate counting the dominant populations in cheeses with defined consortium ACD were lactobacilli (99%) at 18 days and lactobacilli (74.4%) and E. faecalis (25.2%) at 28 days. The SSCP patterns of the cheese surfaces prepared with consortium TR15 were the most diverse, with 10 peaks in the profile at 18 days and 12 peaks in the profile at 28 days. They were characterised by the presence of Cb. mobile, M. psychrotolerans, A. nicotianae or arilaitensis, A. ardleyensis or bergerei and Brachybacterium sp. These species were not specifically quantifiable by plate counting. S. pulvereri represented 30% of the profiles of complex consortium TR15, whereas it was subdominant in cheeses with defined consortia (ABC and ABD). The plate counting method gave a quite different picture, as S. pulvereri was the dominant population in these cheeses. According to the PCR-SSCP analysis, A. ardleyensis or bergerei was dominant in cheeses with defined consortia (ABCD, ABC and ABD) at day 28. Such discrepancies may either be due to S. pulvereri growing better than A. ardleyensis or bergerei on CRBM, or to better extraction and amplification of A. ardleyensis or A. bergerei-DNA for the SSCP analysis. Moreover, the SSCP method detects both dead and live cells.

3.7. Sugar and organic acid contents on cheese surfaces

The dynamics of lactate consumption was quite similar in all cheeses, with a decrease from day 8 to 28 (Fig. 2). In the same way, galactose content decreased rapidly from day 2 to 28 in all cheeses. A peak eluting with formate was not produced in the control and ABC cheeses. This molecule increased until day 18 in TR15, ABD and ACD cheeses and decreased after in cheeses TR15, ACD, whereas it remained stable in ABD cheeses. Maximum production was reached at 8 days for ABCD cheeses and was not detected at day 18. The acetate production curves also differed from one cheese to another. Acetate increased more or less rapidly between days 8 and 18 in all cheeses (Fig. 2). After day 18, acetate content decreased slightly in ABCD cheeses but tended to stabilise in ABC, ABD and ACD cheeses. In the control and TR15 cheeses, acetate content increased sharply between days 18 and 28.

Figure 2. Galactose and organic acid contents on the surfaces of cheeses inoculated with complex consortium TR15 or defined consortia. All contents measured by HPLC at 1, 2, 18 and 28 days of ripening and expressed as g·kg−1 of dry matter content of rind. Control = commercial ripening culture; TR15 = complex consortium (rind suspension); ABCD, ABC, ABD and ACD = defined consortia from identified isolates (A = six lactic acid bacteria, B = seven Gram-positive and catalase-positive bacteria, C = three Gram-negative bacteria and D = three yeasts).

4. DISCUSSION

Complex microbial consortia from the surface of raw milk Saint-Nectaire cheeses showed wide diversity in their ability to inhibit L. monocytogenes. The growth of L. monocytogenes was reduced for 10 out of the 34 cheeses inoculated with complex microbial consortia. This result shows that microbial consortia from raw milk cheeses can self-protect against some pathogens despite high pH values on their surfaces, which favours L. monocytogenes growth. This provides an argument for the safety of raw milk cheeses, in agreement with other published studies. For example, Eppert et al. [13] have described complex consortia from red smear cheeses that inhibited L. monocytogenes. Similarly, according to Guillier et al. [17], biofilm recovered from wooden shelves during ripening of soft and smear cheeses was able to inhibit L. monocytogenes growth.

Our study confirms the conclusions of studies [11, 14] showing that the picture of microbial community of cheeses varies according to the method used to assess them. The plate counting method and a culture-independent method (SSCP) provided complementary information about the dynamics of microbial populations on cheese surfaces inoculated with different microbial consortia. By the plate counting method, the most inhibitory consortium, TR15, was distinguished from the others by its high level of lactic acid bacteria, including lactobacilli and leuconostocs, at the beginning of ripening, which may play a role in the inhibition. During ripening, on MSE medium a leuconostoc population was detected in complex consortium TR15 but not in the defined consortia, whereas PCR-SSCP analysis detected this population in all cheeses. This may mean that consortium TR15 had a different leuconostoc population to that in the defined consortia. The leuconostoc populations in the defined consortia may have lost their ability to produce dextran, and the SSCP analysis may have detected dead cells as well as live. The SSCP profiles of cheese surfaces with the complex cheese consortium TR15 were more diverse than the others. They were characterised by the presence of Cb. mobile, M. psychrotolerans, A. nicotianae or A. arilaitensis, A. ardleyensis or bergerei and Brachybacterium sp., whereas the profiles of defined consortia were often dominated by only two or three species (A. ardleyensis or bergerei, S. pulvereri and Ln. mesenteroides). M. psychrotolerans and Cb. mobile are halophilic, alkalophilic lactic acid bacteria that have rarely been described in cheeses. They have been found in Saint-Nectaire cheeses [11], in red smear cheeses [14, 24] and Egyptian soft Domiati cheeses [12]. In smear cheeses, Corynebacterium casei, A. nicotianae and B. linens constituted more than 70% of the microbial population [3, 8, 24]. S. pulvereri has not often been described in cheeses other than Saint-Nectaire [11]. The population of Gram-negative bacteria present in our cheese was commonly described on the surfaces of cheeses [11, 14] not manufactured and ripened in sterile conditions. And in spite of the difficulty of controlling the proliferation of Gram-negative bacteria, differences in L. monocytogenes growth were still observed.

It can be hypothesised that these differences in the balance between groups of microorganisms may have a preponderant role in the inhibition, but it was not possible to identify one population as being responsible.

In the literature, inhibition in cheeses has been usually attributed to bacteriocin-producing strains even in the cases where the molecules have not been quantified. For example, Eppert et al. [13] reported that the inhibition of L. monocytogenes in red smear cheeses was due to a strain of B. linens producing linocin. Loessner et al. [22] showed that a pediocin produced by Lactobacillus plantarum could have an efficient antilisterial activity. In our study, none of the isolates used in the defined consortia were able to inhibit L. monocytogenes in the conditions of the agar-well diffusion method, suggesting that bacteriocins were not involved in the inhibition.

A possible link between the inhibition of L. monocytogenes with galactose consumption and acid production has not yet been established. L. monocytogenes can be inhibited by formate and by acetate [26, 28]. A wide range of lactic acid bacteria including Lactobacillus sp., Enterococcus sp. or Ln. mesenteroides [18] are able to produce acetate from lactose or galactose catabolism by the Leloir pathway or the T-tagatose pathway [1]. Moreover, some strains of lactic acid bacteria such as E. faecalis were shown to be able to co-metabolise lactate and citrate to produce acetate, formate and ethanol in a milk medium [27]. If most lactic acid bacteria may have the potentialities to catabolise lactate into acetate or formate [21], further investigations will be needed to characterise the capacity of production by the strains of our inhibitory complex consortium.

Lactate, an antilisterial substance, was consumed at the cheese surfaces. The catabolism of lactate is frequently attributed to yeast strains such as D. hansenii [4, 20] and Y. lipolytica [23]. In this study, the results did not enable us to link a specific microbial population and organic acid output to the inhibition of L. monocytogenes.

5. CONCLUSION

A natural cheese surface consortium having an antagonistic effect against L. monocytogenes was identified, but the strategy applied to reconstitute it failed to obtain a defined consortium with similar properties. New strategies need to be established to reconstitute this microbial consortium, taking better account of diversity at strain level, cell preparation and the physiological state of the cells. Further investigations will be needed to gain a better understanding of the microbial interactions involved in L. monocytogenes inhibition. Some hypotheses made about the role of leuconostocs and lactobacilli should be checked.

Acknowledgments

This work is part of the TRUEFOOD project: TRUEFOOD – “Traditional United Europe Food” is an Integrated Project financed by the European Commission under the sixth Framework Programme for RTD – Contract no. FOOD-CT-2006-016264. The authors would like to thank René Lavigne for the cheese production and Béatrice Dessere for her helpful technical support. The authors are also grateful to Daniel Picque and especially Armelle Delile (INRA, UMR782 Génie et microbiologie des procédés alimentaires, Thiverval-Grignon, France) for HLPC analysis. English proofreading was done by Harriet Coleman.

References

All Tables

All Figures

Figure 1. L. monocytogenes growth on the surfaces of cheeses inoculated with complex consortium TR15 and defined consortia. L. monocytogenes counts expressed in log CFU·cm−2; cheese surfaces inoculated with control = commercial ripening culture, TR15 = complex consortium (rind suspension), ABCD, ABC, ABD and ACD = defined consortia from identified isolates (A = six lactic acid bacteria, B = seven Gram-positive and catalase-positive bacteria, C = three Gram-negative bacteria and D = three yeasts).

In the text

Figure 2. Galactose and organic acid contents on the surfaces of cheeses inoculated with complex consortium TR15 or defined consortia. All contents measured by HPLC at 1, 2, 18 and 28 days of ripening and expressed as g·kg−1 of dry matter content of rind. Control = commercial ripening culture; TR15 = complex consortium (rind suspension); ABCD, ABC, ABD and ACD = defined consortia from identified isolates (A = six lactic acid bacteria, B = seven Gram-positive and catalase-positive bacteria, C = three Gram-negative bacteria and D = three yeasts).

In the text