Genotypic and phenotypic heterogeneity of Streptococcus ...

Journal of Applied Microbiology 2001, 91, 937±943

Genotypic and phenotypic heterogeneity of Streptococcus thermophilus strains isolated from dairy products G. Giraffa, A. Paris, L. Valcavi, M. Gatti and E. Neviani Istituto Sperimentale Lattiero Caseario, 26900 Lodi, Italy 2001/52: received 9 April 2001, revised 29 May 2001 and accepted 31 May 2001

G . G I R A F F A , A . P A R I S , L . V A L C A V I , M . G A T T I A N D E . N E V I A N I . 2001.

Aims: Forty strains of Streptococcus thermophilus isolated from dairy products were identi®ed and typed by a polyphasic approach which included phenotypic and genotypic criteria. Methods and Results: Strains were identi®ed by sugar fermentation pattern and speciesspeci®c PCR. Phenotypic diversity was evaluated by a chemometric model taking into account some biochemical characteristics (e.g. acidifying and peptidase activities) of technological interest. Genotypic diversity was evidenced by PCR ®ngerprinting. The overall phenotypic and genotypic information was elaborated on a multivariate statistical basis by principal components analysis and cluster analysis, respectively. When acidifying and peptidase activities were considered, PCA indicated that most of the strains isolated from Pecorino Toscano cheese were separable from the others. Similarly, most of the starter culture strains tended to separate from the cheese isolates. Conclusions: A wide strain heterogeneity among Strep. thermophilus strains isolated from dairy products was observed. Signi®cance and Impact of the Study: A computerized analysis of genotypic and phenotypic information could be applied successfully to differentiate and characterize reliably and rapidly isolates occurring in different dairy products and to comprehend the technological role of speci®c Strep. thermophilus strains in dairy technology. INTRODUCTION Streptococcus thermophilus is one of the most important lactic acid bacteria in the dairy industry, well known as a starter culture component in yoghurt fermentation and cheesemaking. Starters for yoghurt production contain Lactobacillus delbrueckii subsp. bulgaricus and Strep. thermophilus, growing in a symbiotic association. Strep. thermophilus is also present as major microbial component in combination with other LAB species in natural thermophilic starters used for the production of Swiss, French and Italian cheeses such as GruyeÁre-type, Emmental, Grana-types, Provolone and Gorgonzola (Auclair and Accolas 1983). It has been suggested that bacterial strains isolated from natural habitats can show phenotypic differences compared with the reference strain or show interstrain phenotypic variability, due to the different environmental pressures Correspondence to: Giorgio Giraffa, Department of Microbiology and Enzymology, Istituto Sperimentale Lattiero Caseario, Via A. Lombardo 11, 26900 Lodi, Italy (e-mail: [email protected]). ã 2001 The Society for Applied Microbiology

(Fortina et al. 1998). Using genotypic methods it is also often problematic to characterize and identify species that occupy similar ecological niches and are believed to play similar functional roles. Different phenotypic and genotypic characteristics revealed at the molecular level have been applied with success for identifying and typing Strep. thermophilus (Salzano et al. 1994; Moschetti et al. 1998; O'Sullivan and Fitzgerald 1998; Morea et al. 1999; Bizzarro et al. 2000). Most of these techniques, however, present advantages and limits; they should therefore be used in combination by polyphasic taxonomy, which takes account of overall molecular biology, biochemical, morphological, physiological, genetics and ecological information. Using polyphasic taxonomy it is possible to obtain a successful differentiation of species which are phenotypically very similar but genotypically quite different, and vice-versa, and to assure a reliable intraspecies differentiation (Vandamme et al. 1996). A source of confusion in the identi®cation and typing of Strep. thermophilus is the differentiation from other

938 G . G I R A F F A ET AL.

Streptococcus spp. (such as Streptococcus salivarius) and from Enterococcus spp. Species-speci®c PCR, ampli®ed rDNA spacer polymorphism and rDNA sequencing have in part solved taxonomic problems in this species (Lick et al. 1996; Moschetti et al. 1998; Morea et al. 1999). On the other hand, Strep. thermophilus strains showing phenotypic anomalies are not easy to locate within Strep. thermophilus clusters using random ampli®ed polymorphic DNA (RAPD) analysis (Moschetti et al. 1998). This study was therefore aimed at characterizing several strains of Strep. thermophilus isolated from different dairy products by phenotypic and genotypic methods. The overall data were elaborated on a statistical basis by multivariate approaches. M A T E R I A LS A N D M E T H O D S Strains, media, and cultivation conditions Thirty-nine Strep. thermophilus strains were used in this study. They included strains isolated from raw milk, yoghurt and cheeses (both cheese starter cultures and cheese curd) as reported in Fig. 3. Five type strains, Strep. thermophilus DSM 20617T, Strep. macedonicus LMG 15061T, Strep. uberis LMG 9465T, Strep. salivarius LMG 11489T and Strep. bovis LMG 8518T were also included as reference strains. Strains were maintained as frozen stocks at ±80°C in the presence of 15% glycerol as a cryoprotective agent and, unless otherwise speci®ed, grown routinely overnight at 42°C in M17 broth medium (Scharlau Chemie SA, Barcelona, Spain). Strain identi®cation Strains, which had been presumptively identi®ed as Strep. thermophilus by simple physiological tests (Hardie 1986), i.e. morphology, growth at different temperatures (15 and 45°C), production of CO2 from glucose, resistance to NaCl (2á5 and 4á0%) and heat treatment (60°C for 30 min), were identi®ed by: (i) carbohydrate fermentation pro®les, which were obtained the API 50 CHL galleries and elaborated by APILAB Plus software (API BiomeÂrieux, France) according to the manufacturer's instructions and (ii) species-speci®c PCR. The PCR identi®cation was performed by using primers with sequence: 5¢ CACTATGCTCAGAATACA 3¢ and 5¢ CGAACAGCATTGATGTTA 3¢, which were shown to be speci®c to amplify an intragenic fragment of 968 bp from the lacZ gene of Strep. thermophilus (Lick et al. 1996). Strep. thermophilus DSM 20617T was used as positive control; Strep. macedonicus LMG 15061T, Strep. uberis LMG 9465T, Strep. salivarius LMG 11489T and Strep. bovis LMG 8518T were included as negative controls. Ampli®cation reactions were carried out in a total volume of 20 ll containing 200 lmol l±1 each of dNTP, 2á0 ll of 10 ´ Taq

reaction buffer, 0á5 lmol l±1 each of the two primers, 1á5 mmol l±1 of MgCl2, 0á5 U of AmpliTaq DNA polymerase (Applied Biosystems, Monza, Italy), and between 2á5 and 10 ll of template DNA extracted by the Instagene matrix (Bio-Rad Laboratories, Milano, Italy) according to the manufacturer's instructions. DNA ampli®cations were performed in a Perkin Elmer (model 9700) thermal cycler (Applied Biosystems). The cycling programme consisted of an initial denaturation step at 94°C for 180 s and then 30 cycles of 90°C for 30 s, 54°C for 70 s and 72°C for 30 s. The ®nal elongation was performed at 72°C for 10 min. PCR products were analysed by electrophoresis through 1á0% agarose gels at 70 V for 3 h in 1 ´ Tris-acetate EDTA (TAE) buffer (1 ´ TAE: 40 mmol l±1 Tris acetate, 1 mmol l±1 EDTA, pH 8á0). Strain characterization PCR ®ngerprinting ampli®cation. Total DNA from different strains was used as a template for PCR ®ngerprinting using as a primer the M13 minisatellite core sequence (Huey and Hall 1989) with sequence 5¢-GAGGGTGGCGGTTCT-3¢. Ampli®cation conditions consisted of an initial denaturation step of 94°C for 120 s, followed by 40 cycles of 94°C for 60 s, 45°C for 20 s and 72°C for 120 s; a ®nal elongation step of 72°C for 10 min was performed. Reagents and conditions for PCR and electrophoresis were similar to above except for the primer and MgCl2 concentrations which were, respectively, set at 2á0 lmol l±1 and 3á0 mmol l±1 1-kbp plus DNA Ladder (Life Technologies Italia, Milan, Italy) was used as a DNA molecular weight marker. Acidifying activity. The cultures, revitalized in M17 broth, were used as inoculum (2% v/v) in 10 ml skim milk powder (Oxoid Ltd, Basingstoke, UK), reconstituted to 10% (w/v), adjusted to 6á8 pH with 0á1 mol l±1 NaOH, and sterilized at 110°C for 30 min. The cultures were then incubated at 42°C. Once coagulated, the cultures obtained were inoculated at 1% (v/v) ratio in 50 ml of (i) sterile skim milk (SSM) and (ii) SSM with 6 g l±1 yeast extract (YE). In addition, revitalized M17 broth cultures were directly inoculated (1% v/v) in sterile galactose broth (GAL cultures). The galactose broth composition consisted of galactose (20 g l±1); peptone (10 g l±1); and yeast extract (4á5 g l±1); the ®nal pH was adjusted at 6á8. Inoculated media were then incubated at 42°C. Five-ml cultures were sampled at four successive times: immediately after inoculation and after 3, 6 and 24 h. pH was measured in the samples (pHmeter Metrohm 654, Metrohm Ltd, Herisau, Switzerland) and the values were expressed as pH decrease, calculated as a difference between the value immediately after inoculation and values at the successive times.

ã 2001 The Society for Applied Microbiology, Journal of Applied Microbiology, 91, 937±943

HETEROGENEITY OF STREP. THERMOPHILUS

Peptidase activity. The revitalized M17 broth cultures were centrifuged (3000 g, 10 min, 4°C) and washed twice with sterile water. The ®nal pellets were resuspended in 0á05 mol l±1 sterile buffered phosphate pH 6á5 to obtain an O.D.550nm 1á25. The enzyme activity was evaluated for the 40 Strep. thermophilus strains with 0á656 mmol l±1 solutions of Phe-Pro-bNa (Phe-Pro), Arg-bNa (Arg) and Leu-bNa (Leu) (Bachem Feinchemikalien AG, Bubendorf, Switzerland) substrates at pH 6á5 after 1 h incubation at 37°C. Aminopeptidase activity was evaluated by reading spectrophotometrically at 580 nm using the method of Boquien et al. (1988). The values were expressed as O.D. at 580 nm. Statistical analysis The photographs of the gels after PCR ®ngerprinting ampli®cations were scanned (Scanjet 6100 C/T, Hewlett Packard Italia, Milan, Italy), and the resulting densitometric traces of the band pro®les were analysed with the pattern analysis software package GelCompar Version 4á1 (Applied Maths, Kortrjik, Belgium). Calculation of similarity of the PCR ®ngerprinting pro®les was based on the Pearson product-moment correlation coef®cient. The Pearson correlation coef®cient, which provides similarity based upon densitometric curves, was chosen because it is generally more appropriate to evaluate similarities between PCR ®ngerprinting pro®les. To the matrices of similarities we applied two grouping techniques: (i) the cluster analysis, by using the unweighted pair group method using arithmetic average (UPGMA) clustering algorithm (Vauterin and Vauterin 1992), which provided a dendrogram and (ii) Multi-Dimensional Scaling (MDS), which provided a tri-dimensional plot of the entries. The minimum level of repeatability of the ampli®cation conditions was calculated by running DNA samples from duplicated ampli®cations of each DNA extract. To limit problems of repeatability (and reproducibility) all the samples to be compared were processed at the same time. Data for acidifying activity at 3, 6 and 24 h in SSM (with or without yeast extract) and in galactose broth, and data for peptidase activities were subjected to multivariate statistical analysis by use of the PARVUS package (Forina et al. 1988). Thirty-nine strains (objects) and 12 variables, i.e. the SSM, YE, GAL, Phe-pro, Arg and Leu cultures sampled at 3, 6 and 24 h incubation (designated SSM3, SSM6, SSM24, YE3, YE6, YE24, Gal3, Gal6, Gal24, Phe-Pro, Arg and Leu, respectively) were considered. This dataset was subjected to data standardization by normalization (autoscaling) and to the principal component analysis (PCA) as dimensioning technique (Massart et al. 1988) according to the chemometric model described previously (Gatti et al. 1999).

939

RESULTS Fermentation pro®les of carbohydrates indicated that only ®ve sugars, i.e. glucose, lactose, sucrose, galactose and fructose were variously utilized by the 39 strains. Only 7% of the strains were identi®ed as Strep. thermophilus with a high degree of reliability (identi®cation%, ID 99á2%); they fermented glucose, lactose and saccharose. For most of the other strains, a clear species assignment was not possible because of the doubtful pro®le of sugar fermentation, evidenced by the shared ability to ferment galactose and, to a lesser extent, mannose; for this reason, they resulted as Strep. thermophilus with a lower ID, which ranged from 21á2 to 81á1%. It is worth noting that the ID information came from the API database. After species-speci®c PCR, the expected amplicon of 968 bp was obtained from total DNA of Strep. thermophilus DSM 20617T (positive control) and the other strains; no ampli®cation was observed from DNA of Strep. macedonicus LMG 15061T, Strep. uberis LMG 9465T, Strep. salivarius LMG 11489T and Strep. bovis LMG 8518T used as negative controls. Table 1 shows the means and standard deviations for the acidi®cation (SSM3, SSM6, SSM24, YE3, YE6, YE24, Gal3, Gal6 and Gal24) and peptidase (Phe-Pro, Arg and Leu) activities of the different Strep. thermophilus strains. The considerable variation of the 12 phenotypic variables prompted us to apply a dimensioning technique, i.e. the PCA analysis, for data processing. In Fig. 1 are shown the Table 1 Acidi®cation and peptidase activities of 39 strains of S. thermophilus Activity

Variable

Max

Min

Means

S.D.

Acidifying

SSM3* SSM6* SSM24* YE3 YE6 YE24 Gal3à Gal6à Gal24à Phe-Pro§ Arg§ Leu§

1á48 1á99 2á55 1á70 2á27 2á57 0á44 1á17 1á46 2á37 0á78 2á09

0á04 0á44 1á34 0á20 1á21 2á08 0á01 0á04 0á06 1á04 0á06 0á10

0á58 1á27 2á04 1á13 1á87 2á34 0á09 0á23 0á65 1á62 0á37 1á12

0á37 0á45 0á33 0á44 0á23 0á13 0á08 0á24 0á40 0á39 0á17 0á39

Peptidase

*SSM3, SSM6, and SSM24: acidifying activity in sterile skim milk after 3, 6 and 24 h of incubation. Y3, Y6, and Y24: acidifying activity in sterile skim milk with yeast extract after 3, 6 and 24 h of incubation. àGal3, Gal6 and Gal24: acidifying activity in galactose broth after 3, 6 and 24 h of incubation. §Phe-Pro, Arg and Leu: aminopeptidase activity evaluated by hydrolysis of Phe-Pro-ûNa, Arg-b-Na, and Leu-b-Na substrates.


Eigenvector 2

Eigenvector 2


Eigenvector 1

Eigenvector 1

Fig. 1 Plot of loadings and scores for the ®rst and second eigenvector of principal component analysis (PCA) carried out on 12 variables and 40 objects. Symbols represent strains isolated from: s yoghurt; n raw milk; h Valle Trompia cheese; d Gorgonzola cheese; m Provolone cheese; j Pecorino Toscano cheese; ´ strain Streptococcus thermophilus DSM 20617T

Fig. 2 Plot of loadings and scores for the ®rst and second eigenvector of principal component analysis (PCA) carried out on 12 variables and 40 objects. Symbols represent strains isolated from: s yoghurt; n raw milk; h cheese; d starter cultures; ´ strain Streptococcus thermophilus DSM 20617T

biplot on the ®rst and second eigenvectors of both the loadings of the 12 variables and the scores of 39 objects (strains) divided into six categories, each corresponding to the different sources of isolation. In this analysis, both cheese isolates and corresponding starter culture isolates were included within each cheese typology. Most strains isolated from Pecorino Toscano cheese (indicated by the symbol (j), tended to separate on the right bottom area of the plot, where the variables corresponding to the acidifying activities in milk (SSM3, SSM6, and SSM24) were located. The separation occurred principally on the ®rst component axis. The strains from Provolone and Valle Trompia cheeses (indicated by the symbols m and h, respectively) tended to group near to the loadings corresponding to the peptidase (Phe-Pro, Leu, and Arg) and acidifying (YE3, YE6, and YE24) activities, respectively. Strains isolated from yoghurt and raw milk remained dispersed along the plot (Fig. 1). Figure 2 shows the biplot on the ®rst and second eigenvectors of both the loadings of the 12 variables and the scores of 39 objects (strains) divided into four categories, i.e. cheese (without distinction into cheese typologies), starter cultures, milk and yoghurt. Most of the strains deriving from starter cultures tended to separate, in the left region of the plot, from the strains originating in cheeses. The separation between the two categories principally occurred on the ®rst component axis, toward the area where were located the peptidase activity variables (Fig. 2). PCR ®ngerprinting pro®les using the M13 minisatellite sequence allowed us to calculate a minimum repeatability of about 78% for our PCR ®ngerprinting experiments, which was similar to the minimum level of repeatability veri®ed in previous investigation (Giraffa et al. 2000). Hence, it was

deduced that only clusters with values of the correlation coef®cient (expressed as a percentage value) above 78% were considered identical. This allowed different groups of strains inside the Strep. thermophilus species to be distinguished. At a similarity level of 78%, however, the dendrogram showed that too many clusters were formed with little or no information about their meaning in relation to the different sources of isolation. At a similarity level of about 30%, three main clusters were delineated (A, B and C; Fig. 3). The corresponding groups of strains were partly correlated with the sources of isolation. In cluster A were grouped the type strain DSM 20617T and strains isolated from Provolone whey starter. In cluster B were grouped most of the strains isolated from yoghurt, Gorgonzola cheese and Pecorino Toscano cheese. In cluster C, which was a heterogeneous cluster, were put together all the strains of Valle Trompia cheese. Strains 1, 17 and 71 remained unclustered (Fig. 3). MDS produced a tri-dimensional plot in which the strains were spread according to their relatedness. It was shown that the strains tended to group into three separated areas, which were in good agreement with the strain composition of clusters A, B, and C; strains 1 and 17 were still ungrouped (Fig. 4). DISCUSSION In the present work, a wide phenotypic and genotypic heterogeneity within strains of Strep. thermophilus isolated from different dairy products was observed. Most strains were designated atypical because of their ability to ferment galactose. In this regard, species-speci®c PCR based on the



941

Fig. 3 RAPD-PCR ®ngerprinting patterns of all the Streptococcus thermophilus strains obtained using the primer M13 and corresponding dendrogram based on the unweighted pair group method with arithmetic average (UPGMA) of Pearson correlation coef®cient (expressed as a percentage value). The similarity value of 78% (minimum level of repeatability) is indicated by a vertical dotted line. On the right-hand side of the ®gure are indicated the strain numbers, the source of strain isolation and the cluster delineation at a similarity value of 30%

ampli®cation of an infragenic fragment of the lacZ gene of Strep. thermophilus (Lick et al. 1996) resolved some of the problems. The strains that can metabolize the galactose, which may be used by other spoilage or pathogenic micro-organisms or may cause browning defects (Hutkins et al. 1986), are of technological importance. Hence it is interesting that we found a large number of galactose

positive strains, although this characteristic is considered the most discriminating for the identi®cation of Strep. thermophilus according to traditional taxonomy. This ®nding led us to investigate the acidifying activity of the different strains in galactose broth, in order to use this character as an additional biochemical parameter to group strains on the basis of PCA analysis.



Fig. 4 Tri-dimensional plot obtained by multi-dimensional scaling grouping technique in which the entries, corresponding to the numbers of the Streptococcus thermophilus strains studied, are spread according to their relatedness. Relatedness is calculated by analysis of the matrix of similarities of the M13 ®ngerprinting pro®les, obtained using the Pearson correlation coef®cient. The three groups of strains A, B and C, which correspond to the clusters obtained by cluster analysis (see Fig. 3), are evidenced

The high degree of variation of the 12 phenotypic variables did not allow discrimination or grouping of the strains on the basis of the mean values of both acidi®cation and peptidase activities. This prompted us to use a dimensioning technique, i.e. the PCA, which allowed grouping of strains by a simultaneous evaluation of the 12 variables. According to the chemometric model previously described for L. helveticus (Gatti et al. 1999), each eigenvector does not represent a single variable but contains a contribution from each of the 12 variables. The total percentage of variance explained by the two eigenvectors obtained for Strep. thermophilus was not high (54á6% of the total variance) as a consequence of the wide phenotypic heterogeneity of the strains. By PCA, strains isolated from Pecorino Toscano cheese were separable from strains isolated from Provolone/Valle Trompia cheeses; in this regard it is worth noting that some variables almost coincided in the bidimensional plot (e.g. SSM6 and SSM24), i.e. they had the same loading on both components and provided the same information. In addition, few strains tended to be grouped near to the acidifying activity in galactose in the right top area of the plot indicating that the three variables Gal3, Gal6 and Gal24 were poorly discriminative.

PCA analysis was also conducted on the same objects considering different categories, e.g. starter strains and cheese isolates. In general, the starter strains seemed more peptidolytic than cheese isolates. Such an observation can be explained easily considering that starter cultures where Strep. thermophilus is used are generally made using milk as culture substrate. Differently from cheese, milk is a buffered medium with a poor low molecular weight nitrogen content; it is conceivable therefore that peptidase activity may play a signi®cant technological role in selecting strains able to grow, acidify and dominate the Strep. thermophilus population of the milk starters. Indeed, peptidase activities against low molecular weight peptides are particularly important for Strep. thermophilus to provide free essential amino acids which are not present in suf®cient amounts in milk (Thomas and Mills 1981; Desmazeaud 1994). Peptidase activities of Strep. thermophilus were shown previously to be related to the individual amino acid content of casein and the non-protein nitrogen fraction present in milk (Neviani et al. 1995). Similarly, by examining genotypic data, the distribution of strains into different PCR ®ngerprinting groups matched with the type of cheese. Most strains isolated from Provolone, Pecorino Toscano and Valle Trompia cheeses were grouped into three distinct clusters. In addition, in a single cluster were grouped strains isolated from yoghurt and strains isolated from Gorgonzola starter culture; this observation, which was not possible by the analysis of phenotypes, does not seem surprising assuming that starter cultures for Gorgonzola cheese-making are generally prepared by using a yoghurt culture association (Auclair and Accolas 1983; Gobbetti et al. 1997). However, an easy interpretation of the strain hierarchies at the level of reproducibility veri®ed in this work was not possible. This left the choice of the value of similarity coef®cient to obtain an acceptable level of comparison. MDS allowed us to simplify the interpretation of the data and to group the strains into three well-separated areas, which matched with the cluster analysis but were easier to visualize. Indeed, MDS does not analyse the original character set, but the matrix of similarities obtained using a similarity coef®cient. Rather than being a separate grouping technique, MDS simply replaces the clustering step. However, it is a valuable alternative to the dendrogram methods, which often oversimplify the data available in a similarity matrix and tend to produce overestimated gerarchies. Therefore, the overall data suggest that phenotypic and genotypic characterization, examined on a multivariate statistical basis, can provide a complementary view of the microbial diversity of Strep. thermophilus. It seems reasonable, therefore, to suppose that by a polyphasic approach strain differences, which are not always observed by a single type of information, would be better appreciated. This combined polyphasic information package can have import-



ant practical implications in dairy technology. The possibility to obtain strain-speci®c molecular markers by genotypic ®ngerprinting could ®nd useful application in dairy microbial ecology, e.g. to monitor dynamics of speci®c strains in mixed microbial populations of Strep. thermophilus. Phenotypic information based on metabolic and biochemical traits of technological interest (such as acidifying and peptidase activity) could be applied successfully to comprehend the technological role of these speci®c strains in dairy technology. REFERENCES Auclair, J. and Accolas, J.P. (1983) Use of thermophilic lactic starters in the dairy industry. Antonie Van Leeuwenhoek 49, 312±326. Bizzarro, R., Torri Tarelli, G., Giraffa, G. and Neviani, E. (2000) Phenotypic and genotypic characterization of lactic acid bacteria isolated from Pecorino Toscano cheese. Italian Journal of Food Science 12, 303±316. Boquien, C.Y., Corrieu, G. and Desmazeaud, M.J. (1988) Enzymatic methods for determining populations of Streptococcus cremoris AM2 and Leuconostoc lactis CNRZ 1091 in pure and mixed cultures. Applied Microbiology and Biotechnology 30, 402±407. Desmazeaud, M.J. (1994) Le lait milieu de culture. In BacteÂrie Lactique Vol. 2 eds De Roissart, H. and Luquet, F.M. pp. 25±36. Uriage, France: Lorica. Forina, M., Leardi, R., Armanino, C. and Lanteri, S. (1988) PARVUS: an Extendable Package of Programs for Data Exploration, Classi®cation, and Correlation. Amsterdam, The Netherlands: Elsevier Scienti®c Software. Fortina, M.G., Nicastro, G., Carminati, D., Neviani, E. and Manachini, P.L. (1998) Lactobacillus helveticus heterogeneity in natural cheese starters: the diversity in phenotypic characteristics. Journal of Applied Microbiology 84, 72±80. Gatti, M., Contarini, G. and Neviani, E. (1999) Effectiveness of chemometric techniques in discrimination of Lactobacillus helveticus biotypes from natural dairy starter cultures on the basis of phenotypic characteristics. Applied and Environmental Microbiology 65, 1450±1454. Giraffa, G., Rossetti, L. and Neviani, E. (2000) An evaluation of chelex-based DNA puri®cation protocols for the typing of lactic acid bacteria. Journal of Microbiological Methods 42, 175±184. Gobbetti, M., Burzigotti, R., Smacchi, E., Corsetti, A. and De Angelis, M. (1997) Microbiology and biochemistry of Gorgonzola cheese during ripening. International Dairy Journal 7, 519±529.

943

Hardie, J.M. (1986) Genus Streptococcus. In Bergey's Manual of Systematic Bacteriology Vol. II eds Sneath, P.H.A., Mair, N.S., Sharpe, M.E. and Holt, J.G. pp. 1043±1071. Baltimore: Williams & Wilkins. Huey, B. and Hall, J. (1989) Hypervariable DNA ®ngerprinting in E. coli. Minisatellite probe from bacteriophage M13. Journal of Bacteriology 171, 2528±2532. Hutkins, R., Halambeck, S.M. and Morris, H.A. (1986) Use of galactose-fermenting Streptococcus thermophilus in the manufacture of Swiss, Mozzarella, and short-method Cheddar cheese. Journal of Dairy Science 69, 1±8. Lick, S., Keller, M., Bockelmann, W. and Jochem Heller, K. (1996) Rapid identi®cation of Streptococcus thermophilus by primer-speci®c PCR ampli®cation based on its lacZ gene. Systematic and Applied Microbiology 19, 74±77. Massart, D.L., Vandegiste, B.G.M., Deming, S.N., Michotte, Y. and Kaufmann, L. (1988) Chemometrics: a Textbook. Amsterdam, The Netherlands: Elsevier Scienti®c Software. Morea, M., Baruzzi, F. and Cocconcelli, P.S. (1999) Molecular and physiological characterisation of dominant bacteria populations in traditional Mozzarella cheese processing. Journal of Applied Microbiology 87, 574±582. Moschetti, G., Blaiotta, G., Aponte, M. et al. (1998) Random ampli®ed polymorphic DNA and ampli®ed ribosomal DNA spacer polymorphism: powerful methods to differentiate Streptococcus thermophilus strains. Journal of Applied Microbiology 85, 25±36. Neviani, E., Giraffa, G., Brizzi, A. and Carminati, D. (1995) Amino acid requirements and peptidase activities of Streptococcus salivarius ssp. thermophilus. Journal of Applied Bacteriology 79, 302±307. O'Sullivan, T.F. and Fitzgerald, G.F. (1998) Comparison of Streptococcus thermophilus strains by pulsed ®eld gel electrophoresis of genomic DNA. FEMS Microbiology Letters 168, 213±219. Salzano, G., Moschetti, G., Villani, F., Pepe, O., Mauriello, G. and Coppola, S. (1994) Genotyping of Streptococcus thermophilus evidenced by restriction analysis of ribosomal DNA. Research in Microbiology 145, 651±658. Thomas, T.D. and Mills, O.E. (1981) Proteolytic enzymes of starter bacteria. Netherlands Milk and Dairy Journal 35, 255±273. Vandamme, P., Pot, B., Gillis, M., De Vos, P., Kersters, K. and Swings, J. (1996) Polyphasic taxonomy, a consensus approach to bacterial systematics. Microbiological Reviews 60, 407±438. Vauterin, L. and Vauterin, P. (1992) Computer aided objective comparison of electrophoresis patterns for grouping and identi®cation of micro-organisms. European Microbiology 1, 37±41.
