International Dairy Journal 11 (2001) 293305

! Origin and diversity of mesophilic lactobacilli in Comte cheese, as revealed by PCR with repetitive and species-specic primers
! ! Francoise Berthier*, Eric Beuvier, Andre Dasen, Remy Grappin
" INRA, Station de Recherches en Technologie et Analyses Laitieres, B.P. 89, 39801 Poligny Cedex, France

Abstract The objectives of this work were to describe the diversity of mesophilic lactobacilli in Comt! cheese at the strain and species levels, e to determine the origin(s) of this non-starter microora, and to get a collection of well characterised strains from Comt! cheeses. e Strains were isolated from milks, starter cultures and eight cheeses from two factories, with four cheeses made from the same vat in each factory. Strain and species assignations were performed with a combination of two PCR-based methods, amplication with the pairs of repetitive primers ERIC1/ERIC2 and REP1R-Dt/REP2-D, and amplication with specic primers for Lactobacillus zeae, Lactobacillus paracasei and Lactobacillus rhamnosus. The reliability and reproducibility of these methods were assessed using 49 collection strains of mesophilic lactobacilli commonly detected in cheeses. A total of 488 isolates of mesophilic lactobacilli was collected and was assigned to 44 dierent strains and three dierent species. Lactobacillus paracasei and Lactobacillus rhamnosus were the predominant species in milks, starter cultures and cheeses, and constituted 98.7% of the isolates. Strain diversity was found at both individual cheese and factory levels. Thirteen and fteen dierent strains were detected throughout cheesemaking and ripening in two individual cheeses made in dierent factories; only 11 dierent strains were detected in the two corresponding mature cheeses. The data strongly suggest that most mesophilic lactobacilli strains originate from raw milk. r 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Species-specic PCR; Strain typing; Lb. paracasei; Lb. rhamnosus; Lb. zeae; REP-PCR; ERIC-PCR; Comt! cheese; Raw milk; Mesophilic e lactobacilli

1. Introduction Comt! cheese is a hard-cooked ripened cheese variety e manufactured from raw cows milk in a limited region in the East of France, and labelled Appellation dOrigine Prot! g! e (AOP) (Beuvier, 1996). Thermophilic and e e mesophilic whey starter cultures, including selected strains of Lactobacillus helveticus, Streptococcus thermophilus and Lactococcus lactis, are added during the cheesemaking process. Mesophilic lactobacilli are detected as a dominant non-starter microora in Comt! e cheese, where their viable numbers increased from 103 to 104 cfu g1 cheese at the beginning of ripening to 108 cfu g1 after four weeks of ripening, and remain at this level throughout a ripening period of at least ve months (Grappin, Beuvier, Bouton, & Pochet, 1999). Mature Comt! cheeses exhibit complex and varied e sensory properties (St" venot, B! rodier, & Schlich, 1997), e e
*Corresponding author. Fax: +33-3-84-37-37-81. E-mail address: berthier@poligny.inra.fr (F. Berthier).

which could originate from various mechanisms, including the activities of the microbial ecosystem. This was demonstrated with experimental mini Comt! -type e cheeses, where changes in the level or origin of the milk microora were shown to aect notably the sensory properties of the mature cheeses (Beuvier, Berthaud, Cegarra, Dasen, Pochet, & Duboz, 1997; Demarigny, Beuvier, Dasen, & Duboz, 1996). The mesophilic lactobacilli could participate to the elaboration of the sensory properties of mature Comt! cheese because of e their abundance and time of presence during ripening, as suggested and investigated in other cheese varieties (Fox, McSweeney, & Lynch, 1998; Sollberger, 1990). To investigate this aspect, and especially to explain and understand the diversity of the sensory properties in mature Comt! cheese, there is a need to know the origin e and to characterise the microora of Comt! cheese at e the strain level, as dierent strains of a same species often have dierent enzymatic potentialities in terms of avour compound production (Williams, Felipe, & Banks, 1998). A collection of well-characterised strains

0958-6946/01/$ - see front matter r 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 8 - 6 9 4 6 ( 0 1 ) 0 0 0 5 9 - 0

294

F. Berthier et al. / International Dairy Journal 11 (2001) 293305 Table 1 Isolation of mesophilic lactobacilli Source of isolates No. of isolates Factory 1 Milks Starter cultures: Lb. culture Lc. culture St. culture Cheese curd b Ripening cheese b 1 day 7 day 21 day 49 day 63 day 91 day 122 day Mature cheese 1a 1b 1c 1d 2a 2b 2c 2d
a

from Comt! cheese is also essential to conduct experie ments in cheesemaking. The objectives of this study were to evaluate the diversity of mesophilic lactobacilli in Comt! cheese e according to cheesemaking and ripening conditions, and to investigate the origin(s) of the strains found in cheese to elucidate at least partially the role of the raw milk microora on the sensory properties of mature cheese. In these respects, a new and reliable approach which allowed a rapid and easy assignment of isolates at the strain and species levels was developed and applied to isolates of mesophilic lactobacilli from milks, starter cultures and Comt! cheeses collected in two cheese e factories.

Factory 2 16a

20

15 16a 1a 20

0 F 1a 4a

2. Materials and methods 2.1. Samples Bacterial strains were isolated at the same time in two dierent factories, 1 and 2, equipped with four cheese vats. These factories were known to produce cheeses with dierent sensory properties. Cheeses were ripened between 5.6 and 9.3 months according to four dierent schemes (ad) used in Comt! technology. According to e Scheme a, cheeses were ripened at 131C for 2 weeks, at 171C for 5 weeks, and at 61C until their optimal ripening time. According to Scheme b, cheeses were ripened at 131C for 7 weeks, at 171C for 6 weeks, and at 61C until their optimal ripening time. According to Scheme c, cheeses were ripened at 131C for 2 weeks, at 101C for 5 weeks, at 171C for 4 weeks, and at 61C until their optimal ripening time. According to Scheme d, cheeses were ripened at 131C for 7 weeks, at 191C for 5 weeks, and at 61C until their optimal ripening time. Cheeses 1ad and 2ad were graded by the same sensory analyst to determine their ripening endpoint. Milk 1 and milk 2 were from factories 1 and 2, respectively. Bacterial strains from the two cheeses 1b and 2b, which were ripened under the same conditions, were isolated at 1, 7, 21, 49, 63, 91, 122 days, and at their optimal ripening time. Strains from the six other cheeses were isolated only at their optimal ripening time. In addition, strains were isolated from the two raw milks, from the ve starter cultures and from curds before pressing. Cheese samples of 10 g without rind were taken at the mid-radius of each Comt! wheel. The isolate e numbers in each sample are given in Table 1. 2.2. Isolation of mesophilic lactobacilli Milks, starter cultures, curds aseptically sampled. Samples were 2% (wt/vol) trisodium citrate (pH on MRS agar pH 6.5 (De Man, and cheeses were emulsied in sterile 8.5), diluted, plated Rogosa, & Sharpe,

17 19 20 20 20 20 20

0 0 20 20 20 20 20

20 20 20 19 19 20 20 19

Isolated on FH medium.

1960) and on FH agar ( Isolini, Grand, & Gl. ttli, 1990) a and incubated anaerobically for 5 days at 201C and 3 days at 371C, respectively. About 20 dierent colonies were randomly picked up from the MRS plates, and in a few cases also from FH plates, and puried twice on MRS agar plates. All isolates were checked for growth at 151C and were examined microscopically prior to storing. They were maintained at 201C in a 1 : 1 glycerolMRS mixture and routinely streaked on MRS plates before use. 2.3. Collection strains A selection of 49 type strains or well-characterised strains of mesophilic lactobacilli were obtained from dierent culture collections, DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen, Braunschweig, Germany), LMG (Laboratorium voor Microbiologie, Universiteit Gent, Gent, Belgium), ATCC (American Type Culture Collection, Rockville, Md., USA), CNRZ (Centre National de Recherches Zootechniques, INRA Jouy-en-Josas, France), NCFB (National Collection of industrial Bacteria, Shineld, Reading, Berkshire, UK), NCDO (National Collection of Dairy Organisms, Shineld, Reading, Berkshire, UK)

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

295

or from dierent laboratories, Station de Recherches sur la Viande, INRA Theix, France, Laboratoire de Recherches sur la viande, INRA Jouy-enJosas, France and IRTA Meat Technology Centre, Monells, Spain (Berthier & Ehrlich, 1999). Details of the species and strain numbers are given in Fig. 2. 2.4. DNA isolation from lactobacilli isolates DNA was extracted from 1 mL samples of fresh MRS cultures in the exponential growth phase. Total DNA was either phenol extracted as described previously (Berthier et al., 1999), or rapidly extracted with the Instagene matrix as described by the manufacturer (Biorad, Ivry sur Seine, France). Instagene isolated DNA was ethanol precipitated and resuspended in 10 mL 10 mm Tris (pH 8.0), 1 mm EDTA. The quantity of DNA obtained by the rst method was estimated by comparison with known standards in ethidium bromidestained 0.7% agarose gels. 2.5. Rep-PCR Primer sets ERIC1R/ERIC2 and REP1R-Dt/REP2D (Versalovic, Koeuth, & Lupski, 1991) were used for ERIC- and REP-PCR amplications, respectively. They were synthesised by Genosys Biotechnologies (Cambridge, UK). PCR amplication was performed in a nal volume of 20 mL containing 1x PCR buer (Applig" ne), 420 ng phenol-extracted DNA, or 5 mL e Instagene-extracted DNA, 1.0 mm MgCl2, 0.25 mm each primer, 200 mm each dNTP, and one unit Taq DNA polymerase (Applig" ne Oncor, Illkirch, France). PCR e reactions were carried out in a thermal cycler Gene Amp PCR system 9600 apparatus (PerkinElmer Applied Biosystems) programmed for 30 cycles of amplication of 1 min at 941C, 1 min at 401C, 6 min ramping to 721C, and 1 min at 721C, preceded by 5 min at 941C. Electrophoresis and computer analysis were performed as previously described (Berthier et al., 1999), except that a GS 670 Molecular Imager System (Biorad, Ivry sur Seine, France) and the version 4.0 instead of 3.1 of the software package GelCompar were used. 2.6. Species-specic PCR The oligonucleotide primers were obtained from Genosys Biotechnologies (Cambridge, UK) and are listed in Table 2, together with the references of their description. The primers zeae16S and zeaeITS were designed from the nucleotide sequences listed in Table 2. The primer 16reverse was paired with primers paracasei16S, rhamnosus16S or zeae16S. The primer 16 was paired with primers paracaseiITS, rhamnosusITS or zeaeITS. PCR reactions were performed in 10 mL 1x PCR buer (Applig" ne) supplemented with 1.0 mm e

Primer specicity

Lb. rhamnosus

Lb. rhamnosus CGATGCGAATTTCTATTATT rhamnosusITS

Universal Universal Lb. zeae

Lb. paracasei

(Berthier & Ehrlich, 1999)

Reference

Oligonucleotide sequence (50 -30 )

GCTGGATCACCTCCTTTC GAAAGGAGGTGATCCAGC GCATCGTGATTCAACTTAA

TTGCATCTTGATTTAATTTTG

(Ward & Timmins, 1999), with an additional T at 50 end (Ward et al., 1999)

CGATGCGAATTTCTAAATT

Table 2 Sequences of the oligonucleotide primers used for species-specic PCR amplication

rhamnosus16S

paracasei16S

16 16reverse zeae16S

zeaeITS

paracaseiITS

Primer

16S rRNA gene, 50 end, forward 16S rRNA gene, 50 end, reversed primer 16S 16S rRNA gene of Lb. zeae type strain, 50 end, forward/d86516 16S rRNA gene of Lb. rhamnosus type strain, 50 end, forward/d16552 16S rRNA gene of Lb. casei ATCC 334, 50 end, forward/d86517 Short 16S23S intergenic spacer of Lb. casei ATCC 393, 50 end, reverse/z75479 Short 16S23S intergenic spacer of Lb. rhamnosus G1, 50 end, reverse/u32966 Short 16S23S intergenic spacer of Lb. paracasei ATCC 27092, 50 end, reverse/u32964

Location/Gene-Bank accession number

CACCGAGATTCAACATGG

CGATGCGAATTTCTTTTTC

(Tilsala-Timisjarvi & Alatossava, 1997) without G with an additional T at 50 end (Tilsala-Timisjarvi et al., 1997) without the second C at 50 end

Lb. paracasei

Lb. zeae

296

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

Fig. 1. Comparison of enumeration after FH and MRS plating of samples from identically ripened Comt! cheeses 1b and 2b. e

MgCl2, 0.3 mm each primer, 200 mm each dNTP, 0.5 unit Taq DNA polymerase (Applig" ne Oncor, Illkirch, e France), and 210 ng phenol-extracted DNA, or 1.5 mL Instagene-extracted DNA. All the ingredients, except specic primers, were rst mixed together, aliquoted and then specic primers were added. PCR reactions were carried out in a thermal cycler Gene Amp PCR system 9600 apparatus (PerkinElmer Applied Biosystems) programmed for 30 cycles of amplication of 1 min at 941C, 0 min at 551C (ITS pairs), or 0 min at 531C (16S pairs), and 1 min at 721C, preceded by 5 min at 941C; The 10 mL were electrophoresed in a 1% agarose gel and were subsequently visualised by UV illumination after ethidium bromide staining. Two PCR products of 350 and 185 bp were observed when DNA could be amplied.

MRS incubated at 201C selected almost exclusively mesophilic lactobacilli in cheeses after seven days of ripening, even if pediococci were occasionally isolated. As shown in Fig. 1, MRS and FH enumerations were systematically higher on MRS compared with FH for cheese 1b, mainly because one dominant strain in cheese 1, Lb. paracasei A12 (see below), was unable to grow at 371C, which is the incubation temperature recommended for selecting Lb. paracasei on FH (Isolini et al., 1990). Enumeration was higher on MRS compared with FH for milk 2, curd 2b and cheese 2b at 1 day because of the presence of cocci, presumably enterococci according to their phenotypic characterisation. Lactococci were never isolated on MRS medium at 201C, even in the curds or the young cheeses, although added as starter culture. The isolates of mesophilic lactobacilli collected on FH medium in factory 2 from milk, curd and young cheese were included in the study because they will not impede comments relating to strain diversity, despite the fact that two dierent media at two dierent temperatures were used. If present in older cheeses, they would indeed have grown on MRS under the conditions used, and all isolates from older cheeses of factory 2 were able to grow on both MRS and FH under the conditions used. 3.2. Strain typing by Rep-PCR MgCl2, primers and DNA concentrations, as well as the temperature prole in the PCR cycle were optimised to obtain reproducible ngerprints with a sucient number of bands. A ramping was thus introduced in the original and usually used procedure (Versalovic et al., 1991) to ensure the reliability of the method (Sobral & Honeycutt, 1993). The annealing temperature recommended for REP primers, 401C, was used with both REP- and ERIC-primers. Finally, amplications with pairs of primers were found more informative than with a single primer. (i) Type strains and collection strains. The reproducibility and the discriminatory power of Rep-PCR to strain level was assessed. Rep-PCR analysis was applied to 49 type strains or collection strains of mesophilic lactobacilli (Table 3) which are commonly found in cheese or closely related to them genetically. The strains used are listed in Fig. 2. Most of these strains were assigned to dierent species by DNA/DNA hybridisation (Bringel, Curk, & Hubert, 1996; Collins, Phillips, & Zanoni, 1989; Dellaglio, Botazzi, & Vescovo, 1975; Dellaglio, Dicks, du, & Torriani, 1991; Montel, Talon, Fournaud, & Champomier, 1991) and/or species-specic PCR amplication (Berthier & Ehrlich, 1998; Berthier et al., 1999). As shown in Fig. 2, ngerprints which were visually identical merged at the 88% similarity coecient following cluster analysis of combined REP- and ERIC-ngerprints, DNA being isolated by the phenol

3. Results 3.1. Isolates Altogether, 488 isolates of mesophilic lactobacilli were collected, 287 from factory 1 and 201 from factory 2, with twenty isolates of mesophilic lactobacilli collected from each sample, except from some of them (Table 1). As shown in Table 1, mesophilic lactobacilli were isolated from two out of the ve starter cultures, and from all the cheese and milk samples. MRS plating at 201C was eective to select mesophilic lactobacilli from most samples, but not all, selecting cocci in milk 2, curd 2 and cheese 2b at 1day, or selecting no bacteria in cheese 2b at 7 days. From the latter samples, FH plating at 371C selected mesophilic lactobacilli in milk 2, mesophilic lactobacilli together with thermophilic lactobacilli in curd 2b and only thermophilic lactobacilli from the other samples, cheese 2b at 1 and 7 days.

F. Berthier et al. / International Dairy Journal 11 (2001) 293305 Table 3 Lb. paracasei, Lb. rhamnosus and Lb. zeae aliation of collection strains Strain CNRZ 313 DSM 20178T LMG 9191T ATCC 334 CNRZ 320 CNRZ 383 CNRZ 763 DSM 4905 DSM 20006 DSM 20008 DSM 20012 DSM 20020 DSM 20207 NCDO 151T=CNRZ 62T LMG 6400T=CNRZ 212T, DSM 20247T CNRZ 205 CNRZ 442 DSM 20023 DSM 20711
a

297

Current name Lb. casei Lb. zeae Lb. paracasei subsp. paratolerans Lb. casei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. Lb. Lb. Lb. Lb. rhamnosus rhamnosus rhamnosus rhamnosus rhamnosus

Species aliation with DNA/DNA hybridisation Lb. zeae species-specic PCRc Lb. zeae species-specic PCRc DNA/DNA hybridisationb Lb. paracasei species-specic PCRc Lb. paracasei species-specic PCRc Lb. paracasei species-specic PCRc Lb. paracasei species-specic PCRc Lb. paracasei species-specic PCRc DNA/DNA hybridisationa Lb. paracasei species-specic PCRc DNA/DNA hybridisationd Lb. paracasei species-specic PCRc DNA/DNA hybridisationa,b Lb. paracasei species-specic PCRc DNA/DNA hybridisationa,b Lb. paracasei species-specic PCRc DNA/DNA hybridisationb Lb. paracasei species-specic PCRc DNA/DNA hybridisationa,b Lb. paracasei species-specic PCRc DNA/DNA hybridisationd,b Lb. paracasei species-specic PCRc Lb. rhamnosus species-specic PCRc Lb.rhamnosus species-specic PCRc Lb. rhamnosus species-specic PCRc Lb. rhamnosus species-specic PCRc DNA/DNA hybridisationa Lb. rhamnosus species-specic PCRc
a,b

Species aliation Lb. zeae Lb. zeae Lb. paracasei Lb. Lb. Lb. Lb. Lb. paracasei paracasei paracasei paracasei paracasei

Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. paracasei Lb. Lb. Lb. Lb. Lb. rhamnosus rhamnosus rhamnosus rhamnosus rhamnosus

Dellaglio et al. (1975). Collins et al. (1989). c This work. d Dellaglio et al. (1991).
b

method. Identical ngerprints with both REP- and ERIC-primers were obtained for all strains provided from dierent collections and known as identical, Lb. plantarum ATCC 14917T/CNRZ 211T, Lb. paracasei CNRZ 62T/NCDO 151T, Lb. rhamnosus CNRZ 212T/ LMG 6400T/DSM 20247T. Identical ngerprints were also obtained for Lb. plantarum CNRZ 211T/CNRZ 1228 and Lb. pentosus CNRZ 1555/CNRZ 1537 as found with RAPD ngerprints (Tailliez, Qu! n! e, & e e Chopin, 1996), and for Lb. rhamnosus CNRZ 442/ CNRZ 205, Lb. pentosus CNRZ 1570/CNRZ 1537, Lb. rhamnosus DSM 20247/DSM 20711, and Lb. paracasei subsp. tolerans DSM 20012/LMG 9191T. Dierent ngerprints were obtained from the other 33 strains with both REP- and ERIC-primers, except from three strains, Lb. curvatus CTC 448, Lb. curvatus CTC 243 and Lb. pentosus CNRZ 1547, which exhibited dierent ngerprints only with REP primers. (ii) Isolates. According to the above results, we decided to ngerprint all cheese isolates with REP primers, and then to conrm REP-based ngerprint discrimination with ERIC-based ngerprint discrimina-

tion of isolates subsets representing each putative strains. As Instagene-isolated DNA gave the same ngerprints as phenol-extracted DNA from the collection strains (data not shown) when similar DNA concentration were added in the PCR mixture, it was used in Rep-PCR analysis of isolates because of rapidity in isolating DNA. The similarity between dierent Rep ngerprints of the same strain was lower with Instageneisolated DNA because of the more variable DNA concentration in the PCR mixture, which sometimes led to dierences in band intensity between ngerprints, less or more DNA increasing or decreasing the intensity of bands with low molecular weight compared to the intensity of bands with high molecular weight, respectively. In that respect, and to avoid the erroneous assessment of two dierent ngerprints instead of one, ngerprint identity was deduced from both clustering analysis and visual inspection, and new ngerprints were generated changing the DNA concentration when doubtful ngerprint were detected. REP-based ngerprint discrimination was further conrmed with ERICbased ngerprint discrimination.

298

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

Fig. 2. REP- and ERIC-PCR ngerprints of collection strains of mesophilic lactobacilli, and generated dendrogram from combined ngerprints. Lb. paracasei CNRZ 62 and Lb. paracasei NCDO 151 are synonyms, as well as Lb. plantarum ATCC 14917 and Lb. plantarum CNRZ 211, and Lb. rhamnosus LMG 6400, Lb. rhamnosus CNRZ 212 and Lb. rhamnosus DSM 20247. Species assignation after species-specic PCR amplication was indicated in italics when there were discrepancies with the current name of strains.

Forty-four dierent ngerprints among the 448 isolates collected were identied in this manner (Fig. 3). All were dierent with REP primer amplication alone, except ve of them, A1, A31, A32, A33 A34 and A35, which were discriminated only with ERIC primer amplication.

3.3. Presumptive species assignation by Rep-PCR ngerprint Rep-PCR ngerprints could be used to give presumptive assignation of strains to species. Indeed, as can be seen in Fig. 3, combined REP and ERIC ngerprints

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

299

Fig. 3. REP- and ERIC-PCR ngerprints diversity among 488 isolates of mesophilic lactobacilli from raw milks, starter cultures and Comt! cheeses e of factories 1 and 2, and generated dendrogram from combined ngerprints. The strains isolated from milk, starter cultures and cheeses of factory 2 are preceded by a black dot. A1A35, Lb. paracasei strains; B1B3, Lb. rhamnosus strains; C1C3, presumed Lb. parabuchneri strains; D1F1, unassigned strains. The number of isolates per strain is given in bracket.

from strains assigned to the same species merged in most cases at 40% or more similarity, while those of strains assigned to dierent species merged at less than 40% similarity. Nevertheless, in some cases, ngerprints from species represented by a few strains, such as Lb. zeae, merged with those from other species, leading to the misclassication of species if the 40% similarity is taken as the criteria for species discrimination. In contrast, all ngerprints from strains assigned to the same species sometimes merged at less than 40% similarity, as those of dierent species. In this case, the presumptive assignation depended on the identication library used.

The 49 ngerprints from collection strains were compared to the 44 ngerprints from this work (results not shown). Strains B1 and B3 merged at 72.4% similarity into the cluster including Lb. rhamnosus LMG 6400T, and strains B2 merged at 58.7% similarity with Lb. rhamnosus DSM 20023; strains A1 to A34, except three, merged at 38% similarity into the cluster including Lb. casei ATCC 334. Strains C1, C2 and C3 merged at 49% similarity with Lb. parabuchneri LMG 17769. All ngerprints from these strains exhibited several common bands with the ngerprints from the collection strains they merged into, strengthening their

300

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

presumptive assignation. Fingerprints from strains A11, A27, A28, A35, D1, E1 and F1 exhibited similarity coecients of less than 40% with the ngerprints from the 49 collection strains, as well as the type strains of other mesophilic lactobacilli, Lb. farciminis, Lb. alimentarius, Lb. maltoromicus, Lb. coryneformis, Lb. amylophilus, Lb. bifermentans and Lb. sharpeae (results not shown). Strains D1, E1 and F1 were obligatory heterofermentative, whereas strains A11, A27, A28, and A35 were facultatively heterofermentative. 3.4. Species assignation by species-specic PCR We used PCR to specically aliate strains or isolates to the three closely related species Lb. paracasei (type strains: strains NCDO 151T; and ATCC 25599T; synonym: LMG 9191T), Lb. rhamnosus (type strain ATCC 7469T; synonyms: LMG 6400T and CNRZ 212T), and Lb. zeae (type strain ATCC 15820T; synonym: DSM 20178T). Newly- designed primers were used to amplify Lb. zeae DNA. The primers derived from the 16S23S intergenic spacer were paired with primer 16, which was formerly used in other speciesspecic PCR amplication of Lactobacillus (Berthier et al., 1998). (i) Type and collection strains. The specicity of the dierent PCR reactions, described in the method section, was assessed as follows. The six sets of PCR primers were used with each DNA of the Lb. casei, Lb. rhamnosus, Lb. paracasei and Lb. zeae strains listed in Table 2. The 16reverse/rhamnosus16S and 16/rhamnosusITS pairs amplied only Lb. rhamnosus DNA. The 16reverse/paracasei16S and 16/paracaseiITS pairs amplied Lb casei 334 and all Lb. paracasei DNA. The 16reverse/zeae16S and 16/zeaeITS pairs amplied Lb. casei CNRZ 313 and Lb. zeae DSM 20178T DNA. (ii) Isolates. Representative isolates of each strain discriminated by Rep-PCR were subjected to Lb. paracasei, Lb. rhamnosus and Lb. zeae specic PCR amplication. 35 strains, A1A35, and three strains, B1B3, were thus assigned to Lb. paracasei and Lb. rhamnosus, respectively. 3.5. Strain diversity in factories 1 and 2 The strains isolated in factory 1 were genetically dierent from those isolated in factory 2. As seen in Fig. 3, no identical Rep ngerprints were indeed found between the 23 and 21 ngerprints from factories 1 and 2, respectively, even if some were very similar, for example A1 and A18. As seen in Fig. 4, six and ten dierent strains were isolated in milks 1 and 2. Strains B1 and A17 were the dominant strains in milks 1 and 2, representing 55% and 25% of the milk isolates. Six and seven dierent strains were isolated from the starter cultures of Lactococcus

and thermophilic lactobacilli used in factory 1, respectively, with strain B1 being dominant in both. Thirteen and fteen dierent strains were isolated in cheeses 1b and 2b throughout cheesemaking and ripening. Three strains, B1, A12 and A31, and six strains, A14, A27, A22, A13 A15 and B3, dominated, 88% in cheese 1b and 2b, respectively. At the end of ripening, similar numbers of mesophilic lactobacilli were found between dierently ripened cheeses from the same factory (for cheeses 1b and 2b, Fig. 1), but the number of strains between cheeses varied from two to nine, with an average of ve strains (Fig. 5). One to two strains represented between 80% and 85% of the isolates in all cheeses, except one. Most of the dominant strains in cheeses 1b and 2b throughout ripening were dominant in the eight mature cheeses, but some strains were only detected at the end of ripening. In factory 1, only one strain, A12, was common to all mature cheeses and was the most dominant, representing from 50% to 85% of the isolates. Strain B1 was the second dominant strain in cheeses 1a and 1c. These two strains represented from 80% to 85% of the isolates from cheeses 1a to 1d. A more variable and complex pattern was observed at the end of ripening of cheeses manufactured in factory 2. In contrast to factory 1, three strains, A13, A14 and A27, were common to all cheeses. In cheeses 2a, 2b and 2d, strain A14 was dominant representing between 30% and 65% of cheese isolates, while in cheeses 2a, 2c and 2d A14 together with a second strain represented 8085% of the cheese isolates. In cheese 2b, six strains made up 85% of the cheese isolates. 3.6. Species diversity The same two predominant species of mesophilic lactobacilli, Lb. paracasei and Lb. rhamnosus, were identied in milks, in two starter cultures, and in cheeses from factories 1 and 2, but in very dierent proportions between factories. For instance, in cheeses, Lb. rhamnosus and Lb. paracasei represented respectively 48% and 52% in cheese 1, and 90.9% and 8.3% in cheese 2b throughout cheesemaking and ripening. Furthermore, Lb. rhamnosus was only present in two of the eight mature cheeses at the end of ripening. A minor species, Lb. parabuchneri was detected in milk 2 and in two cheeses from factory 1 at the end of ripening. 3.7. Strain origin(s) Data from this study indicated that a large number of the mesophilic lactobacilli strains in Comt! cheese could e originate from raw milk. However, the data were not conclusively proving because we did not detect all cheese strains in the raw milks. Indeed, with 20 isolates collected per sample, strains could only be detected

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

301

Fig. 4. Strain diversity among mesophilic lactobacilli isolates from milks, starter cultures, and Comt! cheeses 1b and 2b throughout cheesemaking e and ripening. Milk 1, starter cultures 1 and cheese 1b were from factory 1. Milk 2 and cheese 2b were from factory 2. , strains which were detected only in milk or curd. &, minor strains in cheese. All strains, except those from milk 2 and Lc. starter culture 1, were isolated on MRS medium. Strains from milk 2 and Lc. starter culture 1 were isolated on FH medium. Strain A12 did not grow on FH medium in the incubation conditions used. A1A35, Lb. paracasei strains; B1B3, Lb. rhamnosus strains; C1C3, presumed Lb. parabuchneri strains; D1F1, unassigned strains.

from the dierent samples if they represented 5% or more of the bacterial population enumerated after plating. As seen in Fig. 4, four dierent situations were observed among the strains of cheese 1b: (i) one minor strain, A2, was found only in the milk; (ii) two major strains, A12 and B1, and two minor strains A1 and A11, were found in both the milk and in at least one of the starter culture; (iii) the third major strain, A31, was only found in starter cultures; and (iv) the seven other minor strains were not found in either the milk or the starter cultures. These results suggest four dierent ways of contamination, depending on the strains considered: milk for A2, both milk and starter culture(s) for A12, B1, A1 and A11, starter culture(s) for A31, and the equipment and/or the factory environment for the others. The initial levels of strains in milk, the relative contribution of starter microora and raw milk microora to these levels, and the levels they reached in the curd, should be considered to choose between raw milk and starter cultures, or both, as source of mesophilic lactobacilli. Strain A2 was present at 7 102 cfu g1 in curd 1b, a level corresponding to that postulated from its level in milk, provided that no growth took place during cheesemaking. Strain A2 could then originate from milk. The strains A1, A11, A12 and B1 most

probably originated also from milk rather than from starter culture(s), as they were detected in curd 1b at levels similar with those postulated from their respective levels in milk without any growth, and as the contribution of starter cultures to the nal cell number of mesophilic lactobacilli in milk was 200 times less than that of raw milk microora. The relative contribution of starter cultures to the initial level of mesophilic lactobacilli strains favoured also a milk origin for the major strain A31. To reach the 2.7 103 cfu g1 enumerated in curd 1 and if it originated from the starter cultures, strain A31 should indeed grow with a generation time of 20 min postulating that growth was exponential from the beginning of cheesemaking (if not, generation time would be still lower), which is impossible. But, if inoculated from milk at a level below the detection limit of the method, strain A31 should grow with a minimum generation time of 7.4 h, which is possible. If so, all the minor strains which were not detected in milk could also originate from milk. The mesophilic lactobacilli detected in starter cultures probably originated from the whey which was used as a growth medium for the starter cultures. This whey was collected a day before the cheeses used in this study were manufactured. All starter strains, except A4, B2 and F1, were also detected in curd 1 (not shown). Moreover

302

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

Fig. 5. Strain diversity among mesophilic lactobacilli isolates from eight mature Comt! cheeses manufactured in factories 1 and 2 according to four e dierent ripening Schemes, ad, used in Comt! cheese technology. &, strains which were detected only in mature cheeses. All strains were isolated on e MRS medium. A5A35, Lb. paracasei strains; B1, Lb. rhamnosus strains; C1C2, presumed Lb. parabuchneri strains.

strains A4, B2 and F1 were never detected in cheese 1 throughout cheesemaking and ripening. This indicates that the same strains of mesophilic lactobacilli were present in curd of cheeses manufactured on two consecutive days, from daily collected milk. Among strains of cheese 2b, only the situations (i) and (iv) described above for cheese 1b were found for both minor and major strains. The strains A13, A15, A16, A17, A20 and B3 could thus originate from milk, while the other strains could originate either from the milk, where they were then below the detection limit of the method, or from the equipment and/or the factory environment. The small number of isolates we obtained from curd 2b did not allow to choose denitively between these two hypotheses. The four isolates we obtained in curd were one of strain A15, two of strain A16 and one of strain A17, reecting the prole of milk strains presented in Fig. 4, and favouring thus the milk origin of all the strains of cheese 2b. The level of

mesophilic lactobacilli present in milk was very low, 1020 cfu mL1 milk.

4. Discussion To describe in detail the diversity of mesophilic lactobacilli in Comt! cheeses, a new, rapid, easy and e reliable approach was developed to assign a large number of uncharacterised isolates at both the strain and the species levels. This approach was based on strain typing and presumptive species assignation of the isolates with Rep-PCR, followed by reliable species assignation with species-specic PCR of representative isolates for each strain. The present work showed that Rep-PCR analysis was very well adapted to strain discrimination of mesophilic lactobacilli. In terms of rapidity and ease of performance, the other available strain typing method is RAPD. Rep-PCR

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

303

reproducibility was the same as that for RAPD under similar conditions of ngerprint analysis (Berthier et al., 1999). However, Rep-PCR analysis presents three advantages over RAPD analysis. First, contrary to RAPD, the same primers can reliably discriminate strains of many dierent Gram-positive and Gramnegative species (Rademaker, Louws, & de Bruijn, 1998), which was not possible with RAPD primers, even among mesophilic lactobacilli (Berthier, unpublished); thus, Rep-PCR can be applied to completely uncharacterised isolates. The second advantage of RepPCR is that each REP- and ERIC-PCR ngerprints of mesophilic lactobacilli strains contained more bands than the single RAPD ngerprints previously obtained for the same strains (Berthier et al., 1999; Tailliez et al., 1996), as was previously observed for Listeria strains (Jersek, Tcherneva, Rijpens, & Herman, 1996); thus, the number of dierent PCR reactions to perform for obtaining the same discrimination level is reduced. The third advantage is that Rep-PCR amplication is less sensitive than RAPD amplication to minor changes in reaction conditions, because REP and ERIC primers are longer (Gillings & Holley, 1997). The robustness and reproducibility of Rep ngerprints was improved in this work by increasing the ramp time from the annealing to the extension steps (Sobral et al., 1993). In other respects, two sets of primers designed from two dierent sequences, the 16S rRNA coding sequence and the 16S 23S small intergenic spacer, are now available together with reliable PCR conditions to specically amplify DNA from each of the three related species Lb. paracasei , Lb. rhamnosus and Lb. zeae. Newly-designed primer pairs, which combined previously-and newlydesigned primers, were used to specically amplify with PCR these three related species. The reliability of species-specic PCR amplication was assessed using a large selection of well-studied and dierent strains of each species. Our optimised PCR conditions ensured a species assignation in accordance with all previous results (Collins et al., 1989; Dellaglio et al., 1975; Dellaglio et al., 1991). In particular, the two strains Lb. zeae DSM 20178T and Lb. casei ATCC 393, classied in Lb. zeae after DNA/DNA hybridisation, but harbouring dierent 16S rRNA coding sequences (Mori et al., 1997) were identically assigned to the same species with the two sets of primers. In this work, 0.06% of isolates were not species assigned; and 0.06% were assigned to a presumptive lactobacilli species. The successful application of Rep-PCR combined with species-specic PCR to assign directly the isolates of mesophilic lactobacilli at the species level, without ambiguities or discrepancies between the results of the two methods, avoided the use of the ambiguous phenotypic species assignation (Collins et al., 1989; Curk, Hubert, & Bringel, 1996). It is striking to note that 35 dierent strains of Lb. paracasei, but only three dierent strains of Lb.

rhamnosus were detected in this work. Lb. paracasei seems to exhibit a dierent intraspecies genomic variability from that of Lb. rhamnosus. This dierence can be observed both among the collection and Comt! e cheese strains. Lb. paracasei and Lb. rhamnosus strains can exhibit very dierent Rep ngerprints. But Lb. rhamnosus ngerprints varied discontinuously and were grouped into three distinct subclusters, while Lb. paracasei ngerprints varied more continuously and were grouped in a main cluster, when the collection and cheese strains were analysed together. This work showed that mesophilic lactobacilli from milk, starter cultures and Comt! cheese at dierent ages e could be selectively isolated and enumerated on FH medium incubated at 201C and not 371C as recommended (Isolini et al., 1990). These lactobacilli can be facultative or obligatory-heterofermentative lactobacilli. The MRS medium was insucient for selecting lactobacilli from some milks, starter cultures and cheeses in the early stages. The temperature of 201C, not 371C, allowed the growth of all mesophilic lactobacilli strains. The dynamics of mesophilic lactobacilli population enumerated on FH incubated at 371C presented in this work is typical of that found in Comt! cheese (Grappin e et al., 1999). A striking result of this work is that diversity among mesophilic lactobacilli from Comt! cheeses of two e dierent origins was found at the strain, but not at the species level. Each Comt! cheese origin could be e identied by its prole of mesophilic lactobacilli strains, with no strain overlapping between proles. Qualitatively, the strain prole in individual mature cheese of the same manufacturing batch varied little according to four dierent ripening conditions used in Comt! e technology, some minor strains appearing or disappearing; the most important changes were in the dierent predominant strains found in the mature cheeses of one cheese batch. The stability of the strain proles of mesophilic lactobacilli over dierent time periods for the same Comt! factory remains to be explored. The origine specic strain prole, but not the species prole, of mesophilic lactobacilli was recently observed in mature Irish Cheddar cheese (Fitzsimons, Cogan, Condon, & Beresford, 1999). The same mean number of dierent mesophilic lactobacilli strains were identied in Comt! (this work) e and Irish Cheddar mature cheeses. However, twice as many were identied throughout Comt! cheese ripening. e Thus, there is a need to examine precisely the strain dynamics throughout cheesemaking and ripening. That will be developed in a further report. As mentioned above, diversity among mesophilic lactobacilli from dierent Comt! cheeses was not found e in their species prole. Indeed, the same two species, Lb. paracasei and Lb. rhamnosus, were predominant, albeit at dierent proportions, in the two Comt! cheeses e

304

F. Berthier et al. / International Dairy Journal 11 (2001) 293305

monitored throughout ripening. It should be noted however that Lb. rhamnosus was absent in the dominant population at the end of ripening. Whereas Lb. paracasei was previously detected as a predominant species in many dierent cheese varieties, Lb. rhamnosus was only detected in some mature hard cheeses, Swisstype, Idia$ abal, Swedish, and Parmigiano Reggiano z cheeses (Coppola et al., 1997; Elortondo, Echobarria, Albisu, & Barcina, 1998; Jimeno, Lazaro, & Sollberger, 1995; Lindberg, Christiansson, Rukke, Eklund, & Molin, 1996). A third minor species, Lb. parabuchneri was detected in two cheeses from one factory. This last species was also mentioned in three English Cheddar mature cheeses (Williams & Banks, 1997). The species diversity of mesophilic lactobacilli in individual mature Comt! cheese was of the same magnitude order as that e reported for mature Irish Cheddar cheese (Fitzsimons et al., 1999), but lower than that reported by in mature English Cheddar cheeses (Williams et al., 1997). The mesophilic lactobacilli population in Comt! cheese was e almost exclusively composed of facultatively heterofermentative lactobacilli, as in Irish Cheddar and Parmigiano Reggiano cheeses (Coppola et al., 1997; Fitzsimons et al., 1999). This work strongly suggests that a large number of the mesophilic lactobacilli strains in Comt! cheese e originated from the raw milk, and that this source was probably more important than factory-environment, processing equipment or starter culture. This result supports the milk origin of mesophilic lactobacilli that could be previously hypothesised (Demarigny et al., 1996). The origin, associated to the specic strain prole of raw milk mesophilic lactobacilli according to milk origin, could partly explain the dierences in sensory properties of experimental Swiss-type cheeses which diered only by the origin of the raw milk microora that were present (Demarigny, Beuvier, Buchin, Pochet, & Grappin, 1997). The origin of the other non-starter populations present in Comt! cheese throughout cheee semaking and ripening remains to be elucidated. On the other hand, it would be also interesting to understand why all of the mesophilic lactobacilli strains detected in the dominant population in milk were not detected in the dominant mesophilic lactobacilli population in cheese throughout cheesemaking and ripening.

and Helen Lamprell and Jean M. Banks for revising the English language.

References
Berthier, F., & Ehrlich, S. D. (1998). Rapid species identication within two groups of closely related lactobacilli using PCR primers that target the 16S/23S rRNA spacer region. FEMS Microbiology Letters, 161, 97106. Berthier, F., & Ehrlich, S. D. (1999). Genetic diversity within Lactobacillus sakei and Lactobacillus curvatus and design of PCR primers for its detection using randomly amplied polymorphic DNA. International Journal of Systematic Bacteriology, 49, 9971007. Beuvier, E. (1996). Comt! cheese. In T. M. Cogan, & M. C. Rea (Eds.), e Artisanal European cheeses (pp. 6368). Brussels: Oce for Ocial Publications of the European Communauties. Beuvier, E., Berthaud, K., Cegarra, S., Dasen, A., Pochet, S., & Duboz, G. (1997). Ripening and quality of Swiss-type cheese made from raw, pasteurized or microltered milk. International Dairy Journal, 7, 311323. Bringel, F., Curk, M. C., & Hubert, J. C. (1996). Characterization of lactobacilli by Southern-type hybridization with a Lactobacillus plantarum pyrDFE probe. International Journal of Systematic Bacteriology, 46(2), 588594. Collins, M. D., Phillips, B. A., & Zanoni, P. (1989). Deoxyribonucleic acid homology studies of Lactobacillus casei, Lactobacillus paracasei sp. nov., subsp. paracasei and subsp. tolerans, and Lactobacillus rhamnosus sp. nov., comb. nov. International Journal of Systematic Bacteriology, 39(2), 105108. Coppola, R., Nanni, M., Iorizzo, M., Sorrentino, A., Sorrentino, E., & Grazia, L. (1997). Survey of lactic acid bacteria isolated during the advanced stages of the ripening of Parmigiano Reggiano cheese. Journal of Dairy Research, 64, 305310. Curk, M. C., Hubert, J. C., & Bringel, F. (1996). Lactobacillus paraplantarum sp. nov., a new species related to Lactobacillus plantarum. International Journal of Systematic Bacteriology, 46(2), 595598. De Man, J. C., Rogosa, M., & Sharpe, M. E. (1960). A medium for the cultivation of lactobacilli. Journal of Applied Bacteriology, 23, 130135. Dellaglio, F., Botazzi, V., & Vescovo, M. (1975). Deoxyribonucleic acid homology among Lactobacillus species of the subgenus Streptobacterium Orla-Jensen. International Journal of Systematic Bacteriology, 25(2), 160172. Dellaglio, F., Dicks, L. T., du, T. M., & Torriani, S. (1991). Designation of ATCC 334 in place of ATCC 393 (NCDO 161) as the neotype strain of Lactobacillus casei subsp. casei and rejection of the name Lactobacillus paracasei (Collins et al., 1989). International Journal of Systematic Bacteriology, 41(2), 340342. Demarigny, Y., Beuvier, E., Buchin, S., Pochet, S., & Grappin, R. (1997). Inuence of raw milk microora on the characteristics of Swiss-type cheese: II. Biochemical and sensory characteristics. Lait, 77, 151167. Demarigny, Y., Beuvier, E., Dasen, A., & Duboz, G. (1996). Inuence of raw milk microora on the characteristics of Swiss-type cheeses: I. Evolution of microora during ripening and characterization of facultatively heterofermentative lactobacilli. Lait, 76, 371387. Elortondo, F. J. P., Echobarria, P. A., Albisu, M., & Barcina, Y. (1998). Indigenous lactic acid bacteria in Idiazabal ewes milk cheese. International Dairy Journal, 8, 725732. Fitzsimons, N. A., Cogan, T. M., Condon, S., & Beresford, T. (1999). Phenotypic and genotypic characterization of non-starter lactic acid bacteria in mature cheddar cheese. Applied and Environmental Microbiology, 65(8), 34183426.

Acknowledgements The technical support of Franck Dufrene is greatly appreciated. This work was nancially supported by the INRA grant A.I.P. Structure et dynamique des " ecosyst" mes bact! riens, and by the region council of e e Franche-Comt! and the European Community e contract no. 96/9 R00202003. The authors would like to thank S. Pochet for critical reading of the manuscript,

F. Berthier et al. / International Dairy Journal 11 (2001) 293305 Fox, P. F., McSweeney, P. L., & Lynch, C. M. (1998). Signicance of non-starter lactic acid bacteria in cheddar cheese. Australian Journal of Dairy Technology, 53, 8389. Gillings, M., & Holley, M. (1997). Amplication of anonymous DNA fragments using pairs of long primers generates reproducible DNA ngerprints that are sensitive to genetic variation. Electrophoresis, 18(9), 15121518. Grappin, R., Beuvier, E., Bouton, Y., & Pochet, S. (1999). Advances in the biochemistry and microbiology of Swiss-type cheeses. Lait, 79, 322. Isolini, D., Grand, M., & Gl. ttli, H. (1990). Selektivmedien zum a Nachweis von obligat und fakultativ heterofermentativen Laktobazillen. Schweiz Milchwirtschaft Forschung, 19, 5759. Jersek, B., Tcherneva, E., Rijpens, N., & Herman, L. (1996). Repetitive element sequence-based PCR for species and strain discrimination in the genus Listeria. Letters in Applied Microbiology, 23(1), 5560. Jimeno, J., Lazaro, M. J., & Sollberger, H. (1995). Antagonistic interactions between propionic acid bacteria and non-stater lactic acid bacteria. Lait, 75, 401413. Lindberg, A.-M., Christiansson, A., Rukke, E.-O., Eklund, T., & Molin, G. (1996). Bacterial ora of Norwegian and Swedish semihard cheese after ripening, with special reference to Lactobacillus. Netherlands Milk and Dairy Journal, 50, 563572. Montel, M. C., Talon, R., Fournaud, J., & Champomier, M. C. (1991). A simplied key for identifying homofermentative Lactobacillus and Carnobacterium spp from meat. Journal of Applied Bacteriology, 70, 469472. Mori, K., Yamazaki, K., Ishiyama, T., Katsumata, M., Kobayashi, K., Kawai, Y., Inoue, N., & Shinano, H. (1997). Comparative sequence analyses of the genes coding for 16S rRNA of Lactobacillus casei-related taxa. International Journal of Systematic Bacteriology, 47(1), 5457. Rademaker, J. L. W., Louws, F. J., & de Bruijn, F. J. (1998). Characterization of the diversity of ecologically important microbes by rep-PCR genomic ngerprinting. In Molecular

305

microbial ecology manual (pp. 127). Dordrecht: Kluwer Academic Publishers. Sobral, B. W., & Honeycutt, R. J. (1993). High output genetic mapping of polyploids using PCR-generated markers. Theoretical and Applied Genetics, 86, 105112. Sollberger, H. (1990). Der Einuss fakultativer Milchs. urebakterien a auf die Propions. ureg. rung und die K. sequalit. t im Emmenthaler. a a a a International Report FAM (pp. 117). St" venot, C., B! rodier, F., & Schlich, P. (1997). Typologie aromatique e e des fromages de Comt! . Sciences des Aliments, 17, 547553. e Tailliez, P., Qu! n! e, P., & Chopin, A. (1996). Estimation de la diversit! e e e parmi les souches de la collection CNRZ: Application de la RAPD " a un groupe de lactobacilles. Lait, 76, 147158. Tilsala-Timisjarvi, A., & Alatossava, T. (1997). Development of oligonucleotide primers from the 16S23S rRNA intergenic sequences for identifying dierent dairy and probiotic lactic acid bacteria by PCR. International Journal of Food Microbiology, 35(1), 4956. Versalovic, J., Koeuth, T., & Lupski, J. R. (1991). Distribution of repetitive DNA sequences in eubacteria and application to ngerprinting of bacterial genomes. Nucleic Acids Research, 19(24), 68236831. Ward, L. J. H., & Timmins, M. J. (1999). Dierentiation of Lactobacillus casei, Lactobacillus paracasei and Lactobacillus rhamnosus by polymerase chain reaction. Letters of Applied Microbiology, 29, 9092. Williams, A. G., & Banks, J. M. (1997). Proteolytic and other hydrolytic enzyme activities in non-starter lactic acid bacteria (NSLAB) isolated from Cheddar cheese manufactured in the United Kingdom. International Dairy Journal, 7, 763774. Williams, A. G., Felipe, X., & Banks, J. M. (1998). Aminopeptidase and dipeptidyl peptidase activity of Lactobacillus spp. and nonstarter lactic acid bacteria (NSLAB) isolated from cheddar cheese. International Dairy Journal, 8, 255266.