Genetic diversity among dairy lactococcal strains investigated by polymerase chain reaction with three arbitrary primers


Dr M. Gueguen, Laboratoire de Microbiologie Alimentaire, Université de CAEN, Esplanade de la Paix, 14032 Caen Cedex, France (e-mail:


Randomly amplified polymorphic DNA (RAPD) has been used for the rapid typing of Lactococcus lactis strains isolated from raw milk from the Camembert region of Normandy. It is thought that the diversity and perhaps the area strain specificity due to climatic and geographical factors of such wild-type lactococcal strains could contribute to the flavour differences and specific features detected for the same product in different areas. The patterns from 58 isolates were analysed by UPGMA dendrograms. At a similarity level of 50%, four RAPD clusters were distinguished. Clusters 1 and 2 contained strains of subspecies lactis and cluster 3 contained strains related to the C2 strain which is genetically cremoris but phenotypically lactis. The type strain of cremoris subspecies was significantly differentiated from these strains with primers P2 and P3. Thus, there was a real genetic diversity in pattern, making it possible to detect potential typical RAPD fragments.

Lactococcus bacteria are widely used as starter cultures for the production of fermented milks and cheeses. Among the species of this genus, only L. lactis subsp. lactis, L. lactis subsp. lactis biovar diacetylactis and L. lactis subsp. cremoris are involved in dairy processing. Strains are blended empirically to accelerate the lowering of milk pH by fermentation of lactose to lactate, and to obtain the specific flavour and aroma required for the taste, quality and acceptance of the fermented food. The stable composition of starter cultures is thus of the utmost importance for cheese manufacturers, who need efficient ways of rapidly discriminating between these subspecies.

In various regions of Europe, industrial cheese starter cultures are mixed with the wild-type lactococcal microflora of raw milk. Thus, the diversity and perhaps the area strain specificity of such wild-type strains may also contribute to the flavour differences and specific features detected for the same product in different areas. The existence of area typical strains would account for the recognized area typicity of RDO (Registred Designation of Origin) cheeses. Therefore, techniques for identifying L. lactis subspecies at the strain level are clearly required.

For a long time, the methods used to differentiate between the two industrially important L. lactis subspecies and the biovar have been essentially based on biochemical and physiological characteristics ( Schleifer et al. 1985 ), or on plasmid DNA digestion patterns ( Davies et al. 1981 ). However, these methods have limitations. Properties typical of the L. lactis lactis subspecies, such as growth at high temperatures and high NaCl concentration, have recently been reported in strains genetically identified as L. lactis subsp. cremoris ( Salama et al. 1995 ). Similarly, many strains originally classified as lactis subspecies were identified as cremoris with reliable molecular based methods ( Klijn et al. 1991 ; Salama et al. 1991 ; Godon et al. 1992 ; Salama et al. 1993 ). Indeed, species-specific oligonucleotide probes selected on highly variable rRNA gene regions unambiguously differentiate the two subspecies ( Betzl et al. 1990 ; Salama et al. 1991, 1993 ). However, these probes cannot distinguish strains below the subspecies level. A partial solution to this problem was found involving the use of rRNA gene restriction analysis which enabled strains to be quickly and definitively identified and clustered in hybridization banding pattern types ( Köhler et al. 1991 ; Rodrigues et al. 1991 ; Desmasures et al. 1998 ). However, strains that differ in total genomic DNA RFLP patterns ( Ramos & Harlander 1990) may have the same ribosomal fragments. With randomly amplified polymorphic DNA (RAPD), larger proportions of the genomes studied are used in the generation of the banding patterns, allowing finer discrimination among strains ( Welsh & McClelland 1990; Williams et al. 1990 ). This method is now used to differentiate between L. lactis strains ( Cancilla et al. 1992 ; Cocconcelli et al. 1995 ; Erlandson & Batt 1997; Desmasures et al. 1998 ).

The aim of this study was to evaluate the typing potential of RAPD, using three primers, to discriminate between L. lactis strains isolated from raw milk from a defined area, in this case, the Camembert region of Normandy (France).

Materials and methods

Origin of the bacterial strains

High quality raw milk (100 ml) was collected from a farm in the Bocage Falaisien region (Normandy, France) and screened for lactococci. Serial dilutions were directly plated onto plate count agar (PCA) supplemented with 10% pasteurized skimmed milk and bromcresol purple; 40 μg ml−1 nalidixic acid and 10 μg ml−1 natamycin were also added to inhibit the growth of undesirable microflora. The plates were incubated at 30 °C for 48 h. Colonies able to ferment lactose, with a characteristic yellow acidifying halo, were selected and purified on M17 agar supplemented with 0·5% lactose ( Terzaghi & Sandine 1975) for further analysis.

The culture collection strains used as reference strains were L. lactis subsp. lactis NCDO604T(CNRZ142T), L. lactis subsp. lactis IL1403 (CNRZ157 plasmid-free), L. lactis subsp. cremoris HP (CNRZ105T) and L. lactis subsp. cremoris C2 (commercial starter strain derived from NCDO712). CNRZ and NCDO strains were obtained from the Unité de Recherches laitières of INRA (Jouy-en-Josas, France), and IL strain from the Laboratoire de Génétique Microbienne of INRA (Jouy-en-Josas, France).

Identification and phenotypic properties of isolates

The strains were initially identified using the following morphological and cultural characteristics: resistance to 4% and 6·5% NaCl; growth in litmus milk; growth at 10, 40 and 45 °C; maltose and sorbitol fermentation; and acetoin production (characteristic of L. lactis subsp. lactis diacetylactis biovar). Further characterization included proteolysis on fast slow differential agar (FSDA) medium ( Huggins & Sandine 1984) and hydrolysis of arginine. These tests distinguished lactococci from other related micro-organisms and differentiated between L. lactis subsp. lactis and L. lactis subsp. cremoris.

DNA preparation

DNA was prepared by the Marmur procedure ( Marmur 1961) using a higher lysozyme concentration (20 mg ml−1) and incubation time (2 h) than the original method.

Oligonucleotide primers

Three single 10-mer primers of arbitrary nucleotide sequence were used in this study: P1: 5′TGCTCTGCCC3′ P2: 5′GGTGACGCAG3′ and P3: 5′CTGCTGGGAC3′.

PCR amplification

The PCR reaction mixture contained 500 ng genomic DNA; 0·5 μmol l−1 of the selected primer (Isoprim, Toulouse, France); 200 μmol l−1 (each) of dATP, dGTP, dCTP and dTTP (Boehringer Mannheim); 2·5 U of AmpliTaq DNA Polymerase (Appligène, Illkirch, France); and 1·5 mmol l−1 MgCl2. The final volume was 100 μl. Amplification was performed in a thermal cycler Amplitron II Thermolyne (Bioblock Scientific, Illkirch, France) as follows: initial denaturation at 94 °C for 5 min, followed by 30 cycles of 1 min denaturation at 94 °C, primer annealing for 2 min at 36 °C and primer extension for 2 min at 72 °C. Samples were then cooled to 4 °C. Aliquots (20 μl) of the amplified products were subjected to electrophoresis in 1% agarose gels (Seakem GTG, Tebu, Le Perray-en-Yvelines, France) in TBE buffer. DNA molecular weight markers (123 bp polymer, Gibco Life Technologies, BRL, Cergy Pontoise, France) were used to determine the size of the bands.

Data analysis

The band patterns for amplified products were scanned from photographic negatives and digitized using Desk Scan II software (Hewlett Packard, Evry, France). Data were then normalized and further processed by the Gel Compare 3·1 Program (Applied Maths, Kortrijk, Belgium) ( Vauterin & Vauterin 1992). Patterns were analysed by Pearson’s product moment correlation coefficient and clustered using the unweighted pair group method with arithmetic averages (UPGMA) ( Sokal & Michener 1958).

The reproducibility of the RAPD technique (separate cell preparation, PCR trial and gel electrophoresis) was evaluated in previous studies ( Tailliez et al. 1996 ; Desmasures et al. 1998 ) and estimated to be at a similarity coefficient of 80%.


According to the microbiological tests, only 58 of the 99 isolates initially selected belonged to the Lactococcus genus. Moreover, phenotypic characteristics classified all the 58 isolates as Lactococcus lactis subsp. lactis.

These 58 isolates were then genetically identified at subspecies level by analysis of RAPD patterns. The relatedness of the strains studied was evaluated by computer-based comparisons of their P1-, P2- and P3-RAPD fingerprints. The patterns of the various strains differed in fragment number, size and intensity depending on the primer used. UPGMA dendrograms were derived from the patterns generated by each primer. Resolution and accuracy were improved by combining the genomic patterns obtained with the three primers and using the combined patterns to generate the dendrogram which is presented with corresponding digital patterns in Fig. 1. RAPD patterns for L. lactis subsp. lactis and L. lactis subsp. cremoris reference strains were added to the analysis, leading to the rapid identification of sample isolates at the subspecies level. The resulting structure cluster was then analysed.

Figure 1&.

emsp; upgma dendrogram derived from the combined patterns generated by P1, P2 and P3 primers. i, Isolated pattern

The strains tested fell into four distinct clusters at a similarity level of 50%. Most strains belonged to clusters 1 and 3 ( Table 1). The Pearson product moment correlation coefficient was used so therefore, the similarity level reflects the overall similarity of the patterns rather than just that of the bands. The 28 environmental strains included in cluster 1 had very similar patterns and were 54% similar to the reference strain L. lactis subsp. lactis IL1403. These strains were therefore identified as L. lactis subsp. lactis. The diversity of RAPD products generated by the P2 and P3 primers was higher than that generated by the P1 primer, and this subdivided cluster 1 into 8 subclusters, each containing strains that were at least 80% similar. Strains with profiles more than 80% similar are considered to be extremely close genetically, and perhaps identical ( Tailliez et al. 1996 ). Therefore, a similar conclusion could be drawn for the strains of each subcluster. The percentage similarity between the outgroup strain, DZ16 A, and the reference strain L. lactis subsp. lactis IL1403, was about 48%. Electrophoretic analysis of PCR products showed one lower mobility band with the P1 primer in comparison with some cluster 1 strains, apparently due to gel distortion. PCR products amplified with the P1 primer from the DZ26B strain, subjected to electrophoresis in the same gel, showed the same low mobility band, which was not detected in a previous preparative gel electrophoresis for these two strains. Thus, strain DZ16 A should also belong to L. lactis subsp. lactis.

Table 1.  RAPD-types (primers P1/P2/P3 or combined) of Lactococcus lactis strains isolated from raw milks from the Camembert region of Normandy
RAPD-types with:  
Strains3 primersP1P2P3GenotypePhenotype
IL14031   lactislactis
NCDO6042   lactislactis
HPi   cremoriscremoris
Table 1a.  (Continued)
RAPD-types with:  
Strains3 primersP1P2P3GenotypePhenotype
C23   cremorislactis

Cluster 2 contained only two strains, which were 53% similar. The first was the L. lactis subsp. lactis type strain NCDO604T and the second was the DZ37 isolate. Both were 48% similar to the outgroup strain DZ18B. Therefore, both DZ37 and DZ18B belong to L. lactis subsp. lactis. There were some conserved fragments between these strains and the other L. lactis subsp. lactis strains, supporting this conclusion.

The three strains of cluster 4 gave very similar patterns, consistent with the close relationship observed between them (similarity level of 85%). However, they had no significant similarity to any reference strain and should therefore probably not be classified as L. lactis. The type strain L. lactis subsp. cremoris HP did not belong to any of the clusters formed by the strains studied.

The percentage similarity between the 25 strains of cluster 3 was about 60%. There were many common bands, regardless of the primer used. Most of these strains were 64% similar (60% for DZ28 and DZ48) to the genetically identified L. lactis subsp. cremoris C2 strain ( Salama et al. 1991 ). Despite this genetic identification, this strain behaves phenotypically more like the lactis subspecies, as do all the strains of cluster 3; the cluster 3 strains should therefore also be genetically identified as L. lactis subsp. cremoris. Eleven subclusters were distinguished between strains that were at least 80% similar. It should be noted that the L. lactis subsp. cremoris type strain HP was differentiated from the strains that were phenotypically lactis but genetically cremoris with primers P2 and P3, but not significantly with P1.

Similar results were obtained for the analysis of each primer-dendrogram (data not shown). At a similarity level of 50%, the same clusters were defined for the same strains ( Table 1). Only the topology of the subclusters identified by analysis of the various combinations of genomic patterns changed as expected, as every primer amplified different parts of the genome and these regions were not equally similar to each other.


Many attempts have been made in recent years to screen L. lactis strains isolated from natural sources for their potential application in the manufacture of cheese. New L. lactis subsp. cremoris strains, in particular, were needed for use as starter cultures. However, there appear to be very few L. lactis subsp. cremoris strains and their natural habitat is unclear. Recently, differences between phenotypic and genetic identification have been reported for this subspecies. Based on phenotypic characteristics, NCDO172- and NCDO172-derived strains as C2 were formerly classified as L. lactis subsp. lactis. However, data from 16S rRNA sequencing ( Klijn et al. 1991 ; Salama et al. 1991 ) and Southern hybridizations with probes specific for the two subspecies ( Klijn et al. 1991 ; Godon et al. 1992 ), or detection of ribosomal gene polymorphism ( Rodrigues et al. 1991 ), have shown that these strains are genetically more related to L. lactis subsp. cremoris than to L. lactis subsp. lactis. According to Salama et al. (1991) , the subspecies cremoris phenotype may have evolved naturally from subspecies lactis in association with dairy practices by the loss of certain phenotypic traits. Another possibility is that the strains of the C2 group originally had the phenotype of subspecies cremoris but have acquired certain traits of subspecies lactis. Thus, the high level of phenotypic variability among the novel L. lactis subsp. cremoris isolates might lead to ambiguous taxonomic designations by standard methods, and may account for the very small number of strains with the cremoris phenotype isolated over many decades. In this study, the RAPD data were analysed by UPGMA dendrograms, enabling the simultaneous comparison of all patterns. All environmental strains used were previously identified by traditional microbiological methods as L. lactis subsp. lactis. At a similarity level of 50%, the L. lactis subsp. lactis group consisted of two distinct clusters (1 and 2, see Fig. 1) with only a few RAPD products in common. Only a 1000 bp fragment seemed to be common to all strains in patterns generated with the P2 primer. It was associated in all strains, except DZ18B, with a 600 bp fragment of various intensities. At least eight significantly distinct banding patterns were distinguished for the 28 strains classified as cluster 1 with data for all the primers combined. Clusters 1 and 2 are only 30% similar. All the patterns generated with the P1 primer were compared and a 1650 bp band was identified which seems to be specific for cluster 1 strains. This fragment was absent from the banding patterns of reference strain IL1403 cluster 2 strains, including the type strain NCDO604T, and of 61 other collection or industrial strains classified as lactis subspecies ( Tailliez et al. 1998 ). This 1650 bp fragment may be specific to some of the strains isolated in the Normandy region. Another distinctive feature of some of the environmental strains identified as L. lactis subsp. lactis by the P2 primer was a common 800 bp AP-PCR product which was absent from the other strains of the lactis subspecies. Thus, there was a major intra-subspecific diversity in patterns, making it possible to detect potential typical bands.

There was similar genetic diversity for strains of subspecies cremoris. Cluster 3 strains were only 10% similar to the L. lactis subsp. cremoris type strain HP, but they were 60% similar to strain C2, which has the rRNA genotype of L. lactis subsp. cremoris ( Salama et al. 1991 ). Cluster 3 strains behaved like L. lactis subsp. lactis, so they resemble the strains of the L. lactis subsp. ‘lactis’ C2 group and should also belong to the genetic cremoris subspecies. Previous studies have also reported the isolation of this type of strain from naturally fermented milk products ( Salama et al. 1993, 1995 ; Weerkamp et al. 1996 ). At least 11 significantly different patterns were distinguished for the 24 cluster 3 strains when the primers were combined. The patterns generated with the P1 primer were compared, showing that two RAPD products (1400 bp and 150 bp) were shared by the strains of cluster 3 and L. lactis subsp. cremoris type strain HP. In contrast, the banding patterns generated with P2 and P3 primers showed no conserved products between the subspecies cremoris type strain HP and the strains of cluster 3. With P2 used as the primer, a 1700 bp band specific to the C2 group was detected and with P3, there was a specific fragment at 250 bp, but with extremely low intensity for strain DZ28.

The strains of the two subspecies formed very distinct RAPD clusters. This is not surprising because the DNA sequence divergence between the two subspecies has been estimated to be between 20 and 30% ( Godon et al. 1992 ). A previous study showed that the RAPD subspecies classification was consistent with that of ribosomal gene restriction analysis ( Desmasures et al. 1998 ). It has been shown in this study that primers P2 and P3 discriminated more effectively between the L. lactis strains studied than did primer P1, especially for the subspecies cremoris strains.

In conclusion, some of the L. lactis strains isolated from raw milk of the Camembert region of Normandy may be specific to this area. Indeed, climatic (precipitation, temperature) and geographical (altitude, distance from the sea) factors may affect the types of strain found. Further investigations should be carried out to confirm that these strains are typical of the area and to determine the properties that make them useful to cheese manufacturers.


The authors are grateful to the Conseil Régional de Basse Normandie, the Syndicat Normand des Fabricants de Camembert (SNFC) and the European Funds for Regional Development (FEDER) for financial support.


  • *

    Present address: Division des Procédés Biotechnologiques, Université Technologique de Compiègne, BP20529, 60205 Compiègne cedex, France.