Polyphasic Investigation of the Diversity within Lactobacillus plantarum Related Strains Revealed Two L. plantarum subgroups

System. Appl. Microbiol. 24, 561–571 (2001) © Urban & Fischer Verlag http://www.urbanfischer.de/journals/sam

Polyphasic Investigation of the Diversity within Lactobacillus plantarum Related Strains Revealed Two L. plantarum subgroups FRANCOISE ¸ BRINGEL1, PASCAL QUÉNÉE2 and PATRICK TAILLIEZ2 1 2

Laboratoire de Microbiologie et de Génétique, Université Louis-Pasteur Strasbourg-I, CNRS, Strasbourg, France Unité de Recherches Laitières et Génétique Appliquée, Collection CNRZ de bactéries lactiques et de bactéries propioniques, INRA, Jouy-en-Josas, France

Received August 30, 2001

Summary The diversity of 140 strains related to Lactobacillus plantarum was investigated using a polyphasic approach combining two molecular techniques: randomly amplified polymorphic DNA fingerprinting (RAPD) and Southern hybridisation with a pyr probe on BglI digests of chromosomal DNA, as well as phenotypic characterization. The RAPD technique allowed us to classify a subset of 60 representative strains into four groups. One group belonged to Lactobacillus paraplantarum, the second to Lactobacillus pentosus and the two remaining groups to L. plantarum (GLp1 and GLp2). The Southern hybridisation technique (F. BRINGEL, M.-C. CURK and J.-C. HUBERT, Int. J. Syst. Bacteriol. 46: 588–594, 1996) revealed nine groups of profiles (I to IX). Results indicated an excellent convergence between RAPD and hybridisation classifications for more than 93% (56/60) of the strains studied. When we compared the fermentation patterns of the L. plantarum strains, three differences were found. Melezitose fermentation was not fermented by the GLp2 RAPD group, unlike the GLp1 RAPD group which included L. plantarum type strain NCIMB11974T. Second, α-methyl-D-mannoside was fermented by a majority of the strains of the GLp1 RAPD group but by none of the strains in the GLp2 RAPD group. Third, dulcitol was catabolized by nearly half of the strains of the GLp2 RAPD group but by none of the strains in the GLp1 RAPD group. Molecular diversity within L. plantarum was confirmed using Southern profiles, PCR amplification and subsequent sequencing of these PCR products. A 773 bp sequence overlapping the pyrDF genes showed high homology: at least 97% identical in L. plantarum strains (V to IX) and 99.9% identical in hybridisation groups VII and VIII. The same G-T transversion which destroyed the pyrF BglI site was found in 11 strains (hybridisation groups VI, VII and VIII). DNA rearrangements were identified downstream from the pyr genes, by PCR amplification and Southern hybridisation profile analysis in three strains of hybridisation groups VIII and IX, two of which also harboured the G-T transversion. Key words: Lactobacillus – plantarum – paraplantarum – pentosus – RAPD – pyrimidine – melezitose – dulcitol – α-methyl-D-mannoside

Introduction Lactobacilli related to Lactobacillus plantarum play an important role in fermented foods: dairy products, silage, pickled vegetables, meat and fish products. They are also potential probiotics (biotherapeutic agents as opposed to antibiotics) and are found in mammal intestinal tracts (AHRNE et al. 1998; OSAWA et al. 2000). Our laboratory collection of L. plantarum strains contains Human isolates, strains isolated from various fermented foods and beverage contaminants. The purpose of this paper is to assess the genetic diversity or relatedness of these L. plantarum strains.

Until recently, the lactobacilli related to L. plantarum constituted an heterologeous group. Some L. plantarum have been regrouped into two other species: L. pentosus (ZANONI et al. 1987) and L. paraplantarum (CURK et al. 1996). L. pentosus can be distinguished from L. plantarum by its ability to produce acid from D-xylose and glycerol, but strains which do not ferment D-xylose have also been described (BRINGEL et al. 1996). L. paraplantarum is indistinguishable from L. plantarum on the basis of physiological tests (BRINGEL et al. 1996). rRNA probes or 16S RNA sequence can not be used to distinguish 0723-2020/01/24/04-561 $ 15.00/0

562

F. BRINGEL et al.

species closely related to L. plantarum since their rRNA exhibit more than 99% homology (COLLINS et al. 1991). However, on the basis of DNA relatedness as determined by DNA reassociation, L. plantarum, L. pentosus and L. paraplantarum clearly constitute distinct taxons (CURK et al. 1996; FRED et al. 1921; ZANONI et al. 1987). As a result, two other discriminating genomic methods which are easier to perform have been described: a Southern-type hybridisation method using a L. plantarum pyrDFE probe on BglI-restricted DNA (BRINGEL et al. 1996), and a PCR-based method amplifying 16S–23S ribosomal DNA (rDNA) spacer regions (BERTHIER & EHRLICH 1998). Methods which have been tested for L. plantarum might be able to distinguish these three species. These techniques include randomly amplified polymorphism DNA (RAPD) (TAILLIEZ et al. 1996; VAN REENEN & DICKS 1996; JOHANSSON et al. 1995); PCR amplification with RAPD-generated DNA fragment primers (QUERE et al. 1997); specific electrophoretic patterns of peptidoglycan hydrolases (LORTAL et al. 1997); rRNA restriction fragment length polymorphism (RFLP) (CHEVALLIER et al. 1994; ZHONG et al. 1998); restriction endonuclease analysis of total chromosomal DNA (JOHANSSON et al. 1995); analysis of total soluble proteins by polyacrilamide gel electrophoresis (VAN REENEN & DICKS 1996); Fourier-transformed infrared spectroscopy (FTIR) (CURK et al. 1994). In this paper, a combination of RAPD, Southern-type hybridisation with a L. plantarum pyrDFE probe, and carbohydrate utilisation was used. This polyphasic approach not only clearly distinguished the three species but also pointed out two subgroups of L. plantarum.

Materials and Methods Bacterial strains, cultivation and physiological tests The sources of the lactobacilli are listed in Table 1. Strains were propagated in MRS broth (Difco Laboratories) at 30 °C in a 4% CO2-enriched atmosphere using a water-jacketed CH/P incubator (Forma Scientific). Glycine was added at a concentration of 2.5% when cells were lysed for DNA extraction. Pyrimidine nutritional requirements were tested in the presence of 4% CO2 on DLA agar plates (BRINGEL et al. 1997) supplemented with arginine (50 µg/ml) with or without pyrimidine (50 µg uracil/ml). Carbohydrate fermentation profiles were studied by using API 50CH strips with API 50CHL medium (bioMérieux). Southern hybridisations Use of Southern-type hybridisation in lactobacilli taxonomy was described previously (BRINGEL et al. 1996) with modifications for DNA extraction. Total DNA was extracted from a 2 ml culture cell pellet suspended in 400 µl TE supplemented with lysozyme (20 mg/ml). After one hour of incubation, the cells were lysed by adding 50 µl of SDS 10% and if necessary to clear up the lysate, incubated at 65 °C. After addition of 100 µl of sodium perchlorate 5 M and 600 µl of chloroform, the precipitated cell debris was eliminated by centrifugation (10 min at 12000 rpm). The supernatant was depleted of proteins with phenol/chloroform extractions and the purified nucleic acids were precipitated with 1 ml of ethanol. Analysis of the Southern hybridisation patterns used PCR amplification. PCR cycle was 1 min at 95 °C followed by 30 cy-

cles of 45 sec at 94 °C, 45 sec at 54 °C, 4 min at 72 °C using the DNA engine PTC-200 Peltier Thermal Cycler (MJ Research). Primer set 1 (N35; 5′-TGATGAATACGTGGCAGTCG-3′; 1,5′CGCAGCATGAACCGTCG-3′) and set 2 (N35; 2,5′-ATCAACTGGACCCCCG-5′) were used to amplify fragments of 0.8 kb and 3.0 kb potentially containing the BglI site in pyrF and in pyrE respectively (Fig. 1). PCR fragments were purified using S-400 HR MicroSpin columns (Amersham Pharmacia Biotech) and sequenced with an Applied Biosystems 373 DNA sequencer using primer 1. The GCG package from the University of Wisconsin (DEVEREUX et al. 1984) was used for nucleotidique sequence analysis. RAPD conditions for genome fingerprinting in L. plantarum Extraction of total cellular DNA, RAPD reactions, electrophoresis and numerical comparison of RAPD profiles were performed as described elsewhere (TAILLIEZ et al., 1998). The sequences of the three RAPD primers used were P1, 5′-TGCTCTGCCC-3′; P2, 5′-GGTGACGCAG-3′ and P3, 5′-CTGCTGGGAC-3′ (BEN OMAR et al., 2000) and were not related to the pyr operon sequence. For each strain, the three RAPD profiles were combined using the GelCompar software (Applied-Maths, Sint-Martens-Latem, Belgium). The standardised RAPD method gave reliable results (TAILLIEZ et al., 1998; CIBIK et al., 2000). Partial Least Squares (PLS) regression PLS predictive models using PLS regression were established using the SIMCA software, version 7.0 (UMETRI, Umeå, Sweden). Phenotype characteristics of strains were considered as variables X. The RAPD groups (GLp1 and GLp2) were considered as variables Y. The PLS regression between variables X and variables Y gave PLS components t[1], t[2], … , t[n], which were linear combinations of variables X (t[1] = w*11X1 + w*12X2 + … + w*1nXn, t[2] = w*21X1 + w*22X2 + … + w*2nXn, t[n] = w*n1X1 + w*n2X2 + … + w*nnXn). These components described the variables X and explained the variables Y. The number of useful PLS components was determined by cross-validation (SIMCA 7.0, 1998). Regressions relating the variables Y to the PLS components t[1], t[2], … ,t[n] were then build as follows : Y1 = c11t[1] + c12t[2] + … + c1nt[n] + residuals, Y2 = c21t[1] + c22t[2] + … + c2nt[n] + residuals, Yn = cn1t[1] + cn2t[2] + … + cnnt[n] + residuals. The X-loadings and the Y-loadings were noted w* and c, respectively. Groups of strains were presented as situated on a plane defined by the PLS components. The predictive quality of a model was evaluated using the R2 coefficient which corresponded to the part of the variance of variables Y explained by the variables X.

Results Southern-type hybridisation The BglI-restricted genomic DNA of 140 lactobacilli were tested by Southern-type hybridisation with a pyrDFE probe and the results are shown in Table 1. Of the nine BglI band patterns obtained, numbered from I to IX, only four have been previously described (BRINGEL et al. 1996). Group V, the majority of the lactobacilli tested (95 strains including NCIMB11974T the L. plantarum type strain), had the 7, 4 and 1 kb BglI pattern which is characteristic of L. plantarum. Group IV, (all 15 strains of L. pentosus including the type strain NCFB363T), had a single BglI band. Group I, (8 strains including

Phylogenetic diversity among Lactobacillus plantarum

563

Table 1. Strain list and comparison between pyrDFE Southern hybridisation and RAPD classification. Species

Detected Bgl I bands (kb)

L. paraplantarum 4; 1

Southern Strains profile

RAPD group

Origins

Sourcea

I

CIP 104445 CNRZ 1885T CNRZ 1887 CNRZ 1888 H 41 H 43 H 48 KOG 16 CNRZ 745 KOG 1 KOG 3 KOG 15 NCFB 1088

Gpa Gpa Gpa Gpa Gpa Gpa Gpa NT Gpa NT NT NT Gpa

Human feces Beer, France Beer, France unknown Polish sauerkraut Polish sauerkraut Polish sauerkraut Korean cabbage kimchi silage, France pickled rice bran Korean cabbage kimchi Korean cabbage kimchi cheese

F. GASSER as strain 61D BRINGEL et al. 1996 BRINGEL et al. 1996 ATCCb P. QUÉNÉE P. QUÉNÉE P. QUÉNÉE OSAWA et al. 2000 CNRZ OSAWA et al. 2000 OSAWA et al. 2000 OSAWA et al. 2000 NCFB

4; 1.2

II

8; 4

III

L. pentosus

7.1 7.5 7.5 7.5 7.1 7.5 7.5 7.5 7.5 7.1 7.1 7.5 6.6 7.5 7.5

IV

CNRZ 1537 CNRZ 1538 CNRZ 1544 CNRZ 1546 CNRZ 1547 CNRZ 1552 CNRZ 1555 CNRZ 1558 CNRZ 1561 CNRZ 1569 CNRZ 1570 CNRZ 1573 NCFB 1059 NCFB 363T NCIMB 8531

Gpe Gpe Gpe Gpe Gpe Gpe Gpe Gpe Gpe Gpe Gpe Gpe NT Gpe NT

fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain fermented olives, Spain red cheshire cheese, UK saw dust fermentation waste sulfite liquor fermentation

CNRZ CNRZ CNRZ CNRZ CNRZ CNRZ CNRZ CNRZ CNRZ CNRZ CNRZ CNRZ NCFB NCFB NCIMB

L. plantarum

7; 4; 1

V

1N5 38AA 54 61T

NT GLp2 GLp1 NT

cow udder, Algeria Colombian cassava Colombian cassava Human feces

1188 A9 Agrano 15b ALAB20 CCM 1904 CCM 3626 CCM 4279 CIP 102021 CIP 102359 CIP 104439 CIP 104440 CIP 104441 CIP 104446 CIP 104447 CIP 104448 CIP 104449 CIP 104450 CIP 104451 CIP 104452 CIP 104453 CIP 104454 CIP 71.39 CNRZ 184 CNRZ 424 CNRZ 738

NT GLp1 NT GLp1 NT NT NT NT NT NT NT NT NT GLp1 NT NT NT NT GLp1 NT NT NT GLp1 GLp1 GLp1

cow udder, Algeria Colombian cassava bread dough sour cassava starch fermentation corn silage percolino romano cheese hard cheese unknown spinal fluid Human Human Human Human Human Human Human Human Human urine Human tooth abscess sauerkraut cantal cheese sauerkraut France bread dough, France silage, France

Lab strain C. FIGUEROA C. FIGUEROA F. GASSER, equivalent to CIP 104442 Lab strain C. FIGUEROA Lab strain C. FIGUEROA CCM CCM CCM CIP CIP CIP CIP CIP CIP CIP CIP CIP CIP CIP CIP CIP CIP CST CNRZ CNRZ CNRZ

564

F. BRINGEL et al.

Table 1. (Continued). Species

Detected Bgl I bands (kb)

RAPD group

Origins

Sourcea

CNRZ 764 CNRZ 1220 CNRZ 1228 CNRZ 1246 CNRZ 1849 CNRZ 1891 CST 10928 CST 10952 CST 10967 CST 11019 CST 11023 CST 11031 CST 12008 CST 12009 DK 15 DK 21 DK 30 DK 32 DK 38 DKO 2A DKO 7 DKO 8 DKO 9 DKO 12 DKO 18 DKO 20A DSM 2648 FB 101 FB 108 FB 115 FOEB 8402 FOEB 9106 FOEB 9113 FOEB 9532 Hd 4 Hd 17 JCL 1279 JCL 1283 KOG 4

GLp1 GLp1 GLp1 GLp1 NT GLp1 NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT GLp1

CNRZ CNRZ CNRZ CNRZ CNRZ Lab strain CST CST CST CST CST CST CST CST Lab strain Lab strain Lab strain Lab strain Lab strain Lab strain Lab strain Lab strain Lab strain Lab strain Lab strain Lab strain CST Lab strain Lab strain Lab strain QUERE et al. 1997 QUERE et al. 1997 QUERE et al. 1997 QUERE et al. 1997 BRINGEL et al. 1996 BRINGEL et al. 1996 J. JIMERO J. JIMERO OSAWA et al. 2000

KOG 24 LMAB1 LMAB2 LN 32 LP 80 LT NCFB 772 NCFB 773 NCFB 963 NCFB 965 NCFB 1042 NCFB 1088 NCFB 1193 NCFB 1204 NCFB 1206 NCIMB 1406 NCFB 2171 NCIMB 5914 NCIMB 6461 NCIMB 7220

NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT NT

cheese, France cheese, Egypt domiati cheese, Egypt domiati cheese, Egypt unknown munster, France recycled beer bootle, France unkwown beer, France beer, France beer, France beer, France milk products fresh products from Gervais fermented millet, Nigeria fermented oil bean, Nigeria fermented cereals, Nigeria fermented cow milk, Nigeria fermented cassava, Nigeria tapioca, Nigeria fermented cereals, Nigeria fermented cereals, Nigeria fermented cereals, Nigeria fermented cereals, Nigeria cucumber, Nigeria fermented cassava, Nigeria silage tortillas dough, Guatemala tortillas dough, Guatemala Hawaiian fermented taro, USA grapes Porto, Portugal White wine pinot wine unknown unknown fermented cucumber, Spain fermented cucumber, Spain Japonese cucumber pickled with rice bran cheese sauerkraut cheese cow udder, Algeria unknown cow udder, Algeria cheese, Sweden cheese, Sweden cheese, Sweden cheese, Sweden English hard cheese cheese silage dairy starter cheese dental carries cheese, New Zealand unknown unknown picked cabbage

Southern Strains profile

OSAWAet al. 2000 F. QUERE F. QUERE Lab strain C. PLATTEEUW Lab strain NCFB NCFB NCFB NCFB NCFB NCFB NCFB NCFB NCFB NCIMB NCFB NCIMB NCIMB NCIMB

Phylogenetic diversity among Lactobacillus plantarum

565

Table 1. (Continued). Species

Detected Bgl I bands (kb)

RAPD group

Origins

Sourcea

NCIMB 8825 NCIMB 8826

NT GLp1

Human saliva Human saliva

NCIMB 11974T SF2A31B SF2A33 SF2A39 SF2B37-1 A 12 CST 12007 SF2B41-1 13 57.2 A1 A4 A7 DK 19 CNRZ 1889

GLp1 NT NT GLp1 NT GLp2 GLp1 GLp1 GLp2 GLp2 GLp1 GLp2 GLp1 GLp2 GLp2

CNRZ 1890 DK 36 DKO 22 SF2A35B DK 9 LP 85-2 NCIMB 12120

GLp2 GLp2 GLp2 GLp2 GLp2 GLp2 GLp2

pickled cabbage sour cassava starch fermentation sour cassava starch fermentation sour cassava starch fermentation sour cassava starch fermentation cassava, Columbia milk product sour cassava starch fermentation cow udder, Algeria Colombian cassava Colombian cassava Colombian cassava Colombian cassava white maize, Nigeria ogi (fermented maize, millet or sorghum), Nigeria baba (fermented millet), Nigeria tapioca, Nigeria sour cassava, Nigeria sour cassava, South America fermented cucumber, Nigeria silage, France ogi (fermented maize, millet or sorghum), Nigeria

NCIMB NCIMB, equivalent to CNRZ1995 NCIMB C. FIGUEROA et al. 1995 C. FIGUEROA et al. 1995 C. FIGUEROA et al. 1995 C. FIGUEROA et al. 1995 C. FIGUEROA CST C. FIGUEROA et al. 1995 Lab strain C. FIGUEROA C. FIGUEROA C. FIGUEROA C. FIGUEROA Lab strain Lab strain DK 24Y

Southern Strains profile

L. plantarum

7; 4.7

VI

L. plantarum

6.7; 4.7

VII

L. plantarum

4.7; 1.1

VIII

L. plantarum

4; 1.1; 1

IX

Lab strain DK 28Y Lab strain Lab strain C. FIGUEROA et al. 1995 Lab strain Lab strain CST

(a) ATCC – American Type Culture Collection, Rockville, Md.; CCM – Czechoslovak Collection of Microorganisms, Brno, Czech Republic; CIP – Collection of bacterial strains of Institut Pasteur, Paris, France; CNRZ – Centre National de Recherches, Zootechniques, INRA, Jouy-en-Josas, France; CST – Collection souches Tepral des brasseries Kronenbourg, Strasbourg, France; NCFB – National Collection of Food Bacteria, Reading, UK; NCIMB – National Collection of Industrial and Marine Bacteria, Aberdeen, Scotland. (b) strain equivalent to ATCC 10776 which is not currently included in the ATCC catalog; (NT) not tested.

CNRZ1885T the type strain of L. paraplantarum), had a 4 and 1 kb BglI bands. Group VIII with 4.7 and 1.1 kb BglI bands, contained the previously characterised L. plantarum strain LP85-2 (BRINGEL et al. 1996). The other groups had new BglI patterns. The profiles for groups V to IX are shown on Fig. 1. The new BglI profiles include group II (four strains with 4 and 1 kb BglI bands); group III (strain NCFB1088 with two BglI bands of 8 and 4 kb); group VI (3 strains with 7 and 4.7 kb BglI bands); group VII (11 strains with 6.7 and 4.7 kb BglI bands) and group IX (strain NCIMB12120 with 4, 1.1 and 1 kb BglI bands ). These BglI-restricted patterns indicate genetic diversity through DNA rearrangements. In the case of strain CCM1904 from group V, the sequence of the pyrimidine biosynthetic genes has been described (ELAGÖZ et al. 1996; accession number gen bank Z54240) and the 2.2 kb pyrDFE probe which harboured two BglI restriction sites, revealed three bands (4 kb; 1 kb and 7 kb) corresponding to the expected band size (3,899 kb; 0.922 kb and a band of unknown length). On the other hand, in strains of group VI, VII and VIII, the two smaller bands

(3,899 kb and 0.922 kb) were replaced by a 4.7 kb band, suggesting that the pyrF BglI site has been lost (Fig. 1). This was confirmed by sequencing a PCR product obtained with primer set 1 (N35/1; Fig. 1). A 773bp sequence obtained with primer 1 revealed a point mutation destroying the BglI recognition site (GCTnnnnnGGC) in CST12007 (group VI), DK9 and LP85-2 (group VIII), and in strains of the group VII (DK19; CNRZ1889; DKO22; 13; 57.2; CNRZ1890 and DK36). On the other hand, the BglI restriction site (GCCnnnnnGGC) was found in NCIMB12120 (group IX) and NCIMB11974T (group V) as expected by the Southern profiles (Fig. 1). When analysing the Southern profiles corresponding to the DNA downstream of the pyr operon, the 7 kb BglI band was replaced by a 1.1 kb band in three strains (Fig. 1). When using one pyrD primer (N35) and one primer localised downstream of the pyr operon (primer 2), the expected 3 kb band was obtained with NCIMB11974T (group V) and CST12007 (group VI). However, no PCR amplification was detected in strains of group VIII and IX. Since N35 was successfully used in PCR amplification in combination with primer 1 but not with primer 2, we

deduced that primer 2 lost its hybridisation site in strains of groups VIII and IX. The PCR amplifications indicate and Southern data confirms, that in strains DK9, LP85-2 and NCIMB12120, DNA rearrangement different from a single point mutation occurred. Thus, using band size, sequencing and PCR amplifications, the BglI sites were localised in the different groups as summarised in Figure 1. The 773bp partial pyrD and pyrF sequences were compared in 13 L. plantarum and were at least 97.4% identical (data not shown). Only 3 nucleotides differed between CST12007 (or SF2B41-1) and CCM1904 (or NCIMB12120). The sequence was identical in strains from group VII (DKO22; CNRZ1889; CNRZ1890; 57.2; 13; DK36) and VIII (DK9; LP85-2) with only a nucleotide variation in strain DK19 (group VII) (data not shown). RAPD classification of 60 lactobacilli strains and comparison with the pyrDFE hybridisation classification

Fig. 1. Southern hybridisation patterns of BglI-restricted lactobacilli genomic DNA with a L. plantarum probe. The band size detected with the pyrDFE probe was specified for each group in Table 1. (a), five different BglI patterns are shown: group V (CIP104447 and CCM1904 in tracks 3 and 10 respectively); group VI (SF2B41-1 and CST12007 in tracks 2 and 5); group VII (CNRZ1889 and DK36 in tracks 1 and 4); group VIII (LP85-2 and DK9 in tracks 6 and 7) and group IX (NCIMB12120 in track 8). DNA digested with HindIII and EcoRI was used as molecular size marker and detected with a digoxigenin-labelled DNA (track 9). (b), the 12 kb region analysed by the Southern hybridisation contains pyrAbDbFE, an ORF of unknown function and downstream DNA which has not been sequenced. The 2.2 kb probe and the BglI restriction sites are defined by a grey rectangular and upwards lines respectively. Primer set 1 (N35, 1) and set 2 (N35, 2) used in PCR analysis are shown with small arrows.

We classified the combined RAPD profiles of 60 representative lactobacilli strains distributed within the nine Southern hybridisation groups reported above (see also Table 1). Four RAPD groups were identified GLp1, GLp2, Gpa and Gpe using the Pearson similarity coefficient and the UPGMA algorithm (Fig. 2). RAPD group Gpe was composed of 13 L. pentosus strains including the type strain NCFB363T. All these strains were also grouped in the hybridisation group IV. RAPD group Gpa was assigned to L. paraplantarum. It contained nine strains including the characterised L. paraplantarum strains CNRZ1885T, CNRZ1887, CNRZ1888 and CIP104445 (CURK et al. 1996). These strains were also included in hybridisation groups I and II. L. plantarum strains are distributed in RAPD groups GLp1 and GLp2 containing 24 strains including L. plantarum type strain NCIMB11974T and 14 strains including L. plantarum strain LP85.2 respectively. Four isogenic strains (CNRZ1360 to CNRZ1363) from strain CNRZ1246 were also included in this study. The Pearson’s coefficient of similarity (90100%) between the profiles of these five latter strains indicated that this technique was highly reproducible. Strains of RAPD group GLp1 were included in hybridisation groups V and VI except strains A1 and A7 which belonged to hybridisation group VII. Strains of RAPD groups GLp2 were included in hybridisation groups VII to

❿ Fig. 2. RAPD patterns of 60 lactobacilli strains, and deduced dendrogramme obtained by UPGMA of merged RAPD patterns. The scale gives the measure of the correlation values (Pearson’s coefficient ×100). The RAPD patterns shown are negative digitised images of RAPD products separated on ethidium bromide stained agarose gels. RAPD products are obtained with primers P1, P2 and P3 as indicated at the top in the figure. Groups corresponding to the different species are indicated in brackets (GLp1 and GLp2, L. plantarum; Gpa, L. paraplantarum and Gpe, L. pentosus). Dashed lines represent non-significant links between groups. Strains which have numbers followed by T correspond to type strains.

568

F. BRINGEL et al.

IX, except strains 38AA and A12 which belonged to hybridisation groups V and VI respectively. Results indicated an excellent convergence between RAPD and hybridisation classifications for more than 93% (56/60) of the strains studied. Relationship between phenotype characteristics of L. plantarum strains and their RAPD classification using PLS regression We calculated a PLS model that linked the 49 phenotype characteristics (variables X) of 34 L. plantarum

strains (training set) and the RAPD groups GLp1 and GLp2 (variables Y). The cross-validation lead to two components. The corresponding PLS model explains 92.9% of the variation of the Y-matrix, indicating an excellent separation of the strains within two groups in the projection plane proposed (Fig. 3a). Each of these groups corresponded to strains belonging to a unique RAPD group. The X-loadings (w*) corresponding to the phenotype characteristics and the Y-loadings (c) corresponding to the RAPD classification of strains are presented in Figure 3b. Acid production from melezitose and α-methyl-Dmannoside were specific to RAPD group GLp1 strains and

Fig. 3. Relationships between phenotype characteristics (variables X) and RAPD groups GLp1 and GLp2 (variables Y) of 34 L. plantarum strains (training set) using PLS regression. a) The cross-validation led to two components represented here as t[1] and t[2]. They define a projection plane allowing a clear separation (indicated by a dashed line) of the two groups of strains represented here as black triangles. The corresponding model explained 92.9% of the variation of the Y-matrix. The 95% probability region defined by the model was delimited by the ellipse in bold (Hotelling T2, (SIMCA 7.00, 1998)). b) The window shows the X-loadings (w*) of the variables X (phenotype characteristics) and the Y-loadings (c) of the variables Y (RAPD groups), and thereby shows the correlation between X and Y. The X (black squares) and Y (black circles) variables combine in the projections, and the variables X relate to the variables Y, as shown in the figure. Characters (melezitose, α-methyl-Dmannoside, dulcitol and α-methylD-glucoside) that are significant for strain classification in the two RAPD groups are indicated beside large black squares (small black squares represent less significant variables X).

Phylogenetic diversity among Lactobacillus plantarum

acid production from dulcitol and α-methyl-D-glucoside were specific to RAPD group GLp2. The other phenotype characteristics were not significantly representative of a RAPD group. We then tested the PLS model with a prediction set of four L. plantarum strains with contradictory molecular classification (A1, A7, A12 and 38AA). Strains 38AA and A12 could be classified in RAPD group GLp1 and strains A1 and A7 in RAPD groups GLp2 with a probability higher than 0.9. Thus, the phenotype classifi-

569

cation of these four strains using the PLS approach confirmed the pyrDFE hybridisation classification. Using PLS modeling, we tried to define the phenotype traits of L. pentosus or L. paraplantarum strains among the 60 strains studied. A PLS model was calculated to group L. pentosus strains (data not shown): D-xylose and glycerol fermentations were confirmed to be the specific traits of L. pentosus strains (ZANONI et al. 1987; FRED et al. 1921). Using this method, no specific phenotype traits allowed us to distinguish L. paraplantarum strains from L. plantarum strains. This confirmed that L. paraplantarum and L. plantarum could not be differentiated by phenotype (CURK et al. 1996). Polyphasic approach revealed two L. plantarum groups The three distance matrices (one for each classification: RAPD, pyrDFE hybridisation and phenotype) were combined using the BioNumerics software (Appliedmaths; Sint-Martens-Latem, Belgium). A dendrogramme was drawn using the combined matrix (Fig. 4). Only four significant phenotype characteristics were used: melezitose, α-methyl-D-mannoside, dulcitol and α-methyl-Dglucoside. Two groups of strains were determined: GA and GB. Group GA contained most of the strains including the type strain NCIMB11974T with strains isolated from fermented plants or of Human and dairy origin. Group GB contained strains exclusively isolated from fermented plant products except strain 13, isolated from cow udder. This strain may have been contaminated by vegetable products such as silage. Strains of group GB were unable to catabolise melezitose and α-methyl-D-mannoside but were often able to ferment dulcitol and α-methyl-D-glucoside.

Discussion Molecular and phenotypic classification of lactobacilli strains

Fig. 4. Dendrogramme showing the diversity of 38 L. plantarum strains by using phenotypic and molecular typing methods. Three similarity matrices were calculated independently for four relevant phenotypic traits (melezitose, α-methyl-D-mannoside, dulcitol and α-methyl-D-glucoside), pyrDFE hybridisation groups and RAPD profiles using the Gower’s correlation coefficient, the Jaccard’s correlation coefficient and the Pearson’s correlation coefficient, respectively. The three matrices were combined using the BioNumerics software (Applied-Maths, SintMartens-Latem, Belgium). The resulting dendrogramme shown is based on the combined similarity matrix. Groups of strains (GA and GB) are indicated in brackets.

Strains of the three species L. plantarum, L. paraplantarum and L. pentosus cannot be easily distinguished using either the 16S rDNA sequence nor the phenotypic traits (COLLINS et al. 1991; CURK et al. 1996; ZANONI et al. 1987). The analysis of carbohydrate fermentation revealed that L. pentosus can be distinguished from L. plantarum and L. paraplantarum using two fermentation criteria. All the L. pentosus strains studied produced acid from glycerol and a majority (88%) produced acid from D-xylose while no other species fermented D-xylose or only a few fermented glycerol (less than 13% and always weakly) (data not shown), thus confirming earlier observations (ZANONI et al. 1987; FRED et al. 1921). However, the classification based on these two characteristics did not seem reliable enough. Moreover, we confirmed that L. plantarum and L. paraplantarum strains cannot be distinguished using carbohydrate fermentation tests (Table 2) even when PLS regression was used.

570

F. BRINGEL et al.

To classify the strains, we used two molecular typing methods: the RAPD and the pyrDFE hybridisation fingerprinting. Four groups were determined using RAPD. One RAPD group (Gpe) was assigned to L. pentosus and another one (Gpa) to L. paraplantarum. The two remaining groups (GLp1 and GLp2) were assigned to L. plantarum. L. pentosus strains belonged to a single pyrDFE hybridisation group IV. L. paraplantarum strains were distributed amongst three pyrDFE hybridisation groups (I to III). Group I included the type strain CNRZ1885T and group II harboured strains KOG1, KOG3 and KOG15 which have been recently identified as L. paraplantarum using rDNA spacer regions PCR techniques (OSAWA et al. 2000). For L. plantarum, RAPD group GLp1 was composed of strains distributed in pyrDFE hybridisation groups V and VI except strains A1 and A7 (group VII). RAPD group GLp2 was composed of strains distributed in the three pyrDFE hybridisation groups VII to IX, except strains 38AA and A12 (groups V and VI respectively). Thus, using RAPD fingerprinting, we were able to classify the strains in 4 groups and to assign each group to a species. This classification has also been confirmed by pyrDFE hybridisation classification. Phenotype classification allowed us to confirm the molecular classification of L. pentosus strains, but not L. plantarum. Based on the four significant phenotypic traits, classification of strains A1, A7, A12 and 38AA was convergent with the pyrDFE hybridisation classification. However, phenotype classification of strains CNRZ1220 and CNRZ1246 was different than that obtained by the two molecular typing classifications. This indicates that a polyphasic approach may be used to classify closely related strains of the same species and define subspecies. Using a polyphasic approach, the L. plantarum strains were separated into two groups. Group GA contained most of the strains and the type strain NCIMB11974T. Group GB was composed of 12% of the tested L. plantarum strains including LP85-2. The latter has previously been shown to harbour atypical L. plantarum features. Its BclI digested genomic DNA Southern hybridisation profile was distinct from that of five L. plantarum from group GA (CHEVALLIER et al. 1994). LP85-2 was shown to form an heteroduplex with the type strain presenting a ∆Tm of 3 °C (BRINGEL et al., 1996) which is at the limit of subspecies specification (GRIMONT 1984) (strains having a ∆Tm between 2.5 and 5.5 °C may belong to different subspecies). Thus, these observations confirm the classification of L. plantarum strains within the two groups determined by our polyphasic approach.

growth was assessed on defined media in absence of pyrimidines (see methods for growth conditions). All strains were prototrophic in the tested conditions in contrast of FB331 which harboured the ∆pyrAaAb mutation (data not shown). Since the pyr genes remained functional, we predicted that the different L. plantarum Southern groups were the result of minor mutational events in the pyr genes. As demonstrated by sequencing of a 773 bp DNA fragment, of the 16 strains which lost the BglI restriction site within the pyrF gene (groups V, VI and VII; Fig. 1), 11 strains were sequenced and harboured a G-T transversion destroying the BglI restriction site with no PyrF amino acid change. It was surprising to find the same mutation in all these strains eventhough other mutations (G-C or G-A) would have been silent. The comparison of the 773bp-long sequence revealed a high degree of homology between L. plantarum strains with at least 97% nucleotide sequence identity. These differences had little effect on amino acid composition of the 238aalong PyrF (only 76aa analysed; D27G change in only one strain, DK9) and of the 278aa-long PyrD (183 aa analysed; T278S change in all tested strains from group VII (DKO22, DK36, 57.2, 13, DK19, CNRZ1889) and group VIII (DK9, LP85-2) but not in CCM1904 (group V), CST12007 and SF2B41.1 (group VI), and NCIMB12120 (group IX)). More extensive DNA rearrangements downstream of the pyr genes were demonstrated in three strains (DK9; LP85-2 and NCIMB12120). The inability to PCR amplify the rearranged region with primer set 2 (N35 and 2 on Fig. 1) was corroborated with a shorter Southern hybridisation detected band of 1.1 kb instead of 7 kb. Taken together, these data strongly suggest that an insertion or deletion event occurred in these strains. In strains from group VIII (DK9 and LP85-2) both events may have occured: the loss of the BglI site (found in strains of group VI and VII), and the DNA rearrangement found in strain NCIMB12120 (group IX). Thus, whether the 1.1 kb BglI band was generated by a DNA rearrangement which was subsequently transferred to the other strains by horizontal gene transfer or independent events occurred in the different strains is not known. Other potential DNA rearrangement loci are genes involved in dulcitol, α-methylD-mannoside and melezitose catabolism. Analysis of these genes in the two L. plantarum subgroups and in strains with atypical fermentation profiles is required to apprehend the nature of DNA heterogeneity within L. plantarum.

Diversity of a 12 kb region evaluated by DNA relatedness within 112 L. plantarum strains

Acknowledgements We thank E. LEPAGE (INRA Jouy-en-Josas, France) for her participation to analyse the data; J-C. HUBERT (University LouisPasteur of Strasbourg, France) for his useful suggestions; B. JOHNSON for reading the paper; P. HAMMANN for DNA sequencing and RO OSAWA (Kobe University, Japan), C. FIGUEROA (CIRAD, France), C. URDACI (University of Bordeaux, France), J. JIMERO (Instituto de Agroquimica y tecnologia de alimentos, Valencia, Spain) for sending strains. D. K. OLUKOYA (National Institute of Medical Research, Nigeria) collected the strains from Nigeria that were studied in this paper.

The Southern hybridisation experiment concerned a region of 12 kb with a 5.3 kb sequence involved in pyrimidine biosynthesis (part of pyrAb, pyrD, pyrF and pyrE). Five different patterns or groups (V to IX) were detected for L. plantarum strains using BglI digests with a pyrDFE probe (Fig. 1). To test if these groups were linked to pyrimidine nutritional requirements, L. plantarum

Phylogenetic diversity among Lactobacillus plantarum

References AHRNÉ, S., NOBAEK, S., JEPPSSON, B., ADLERBERTH, I., WOLD, A. E, MOLIN, G.: The normal Lactobacillus flora of healthy human rectal and oral mucosa. J. Appl. Microbiol. 85, 88–94 (1998). BEN OMAR, N., AMPE, F., RAIMBAULT, M., GUYOT, J.-P., TAILLIEZ, P.: Molecular diversity of lactic acid bacteria from cassava sour starch (Columbia). Syst. Appl. Microbiol. 23, 285–291 (2000). BERTHIER, F., EHRLICH, S. D.: Rapid species identification within two groups of closely related lactobacilli using PCR primers that target the 16S/23S rRNA spacer region. FEMS Microbiol. Lett. 161, 97–106 (1998). BRINGEL, F., CURK, M.-C., HUBERT, J.-C.: Characterization of lactobacilli by Southern-type hybridization with a Lactobacillus plantarum pyrDFE probe. Int. J. Sys. Bacteriol. 46, 588–594 (1996). BRINGEL, F., FREY, L., BOIVIN, S., HUBERT, J.-C.: Arginine biosynthesis and regulation in Lactobacillus plantarum: the carA gene and the argCJBDF cluster are divergently transcribed. J. Bacteriol. 179, 2697–2706 (1997). CHEVALLIER, B., HUBERT, J.-C., KAMMERER, B.: Determination of chromosome size and number of rrn loci in Lactobacillus plantarum by pulsed-field gel electrophoresis. FEMS Microbiol. Lett. 120, 51–56 (1994). CIBIK, R., LEPAGE, E., TAILLIEZ, P.: Molecular diversity of Leuconostoc mesenteroides and Leuconostoc citreum isolated from traditional French chesses as revealed by RAPD fingerprinting, 16S rDNA sequencing and 16S rDNA fragment amplification. Syst. Appl. Microbiol. 23, 267–278 (2000). COLLINS, M. D., RODRIGUES, U., ASH, C., AGUIRRE, M., FARROW, J. A. E., MARTINEZ-MURCIA, A., PHILLIPS, B. A., WILLIAMS, A. M., WALLBANKS, S.: Phylogenetic analysis of the genus Lactobacillus and related lactic acid bacteria as determined by reverse transcriptase sequencing of 16S rRNA. FEMS Microbiol. Lett.77, 5–12 (1991). CURK, M.-C., PELADAN, F., HUBERT, J.-C.: Fourier transform infrared (FTIR) spectroscopy for identifying Lactobacillus species. FEMS Microbiol. Lett. 123, 241–248 (1994). CURK, M.-C., HUBERT, J.-C., BRINGEL, F.: Lactobacillus paraplantarum sp. nov., a new species related to Lactobacillus plantarum. Int. J. Sys. Bacteriol. 46, 595–598 (1996). DEVEREUX, J., HAEBERLI, P., SMITHIES, O.: A comprehensive set of sequence analysis programs for the VAX. Nucleic Acid Res. 12, 387–395 (1984). ELAGÖZ, A., ABDI, A., HUBERT, J.-C., KAMMERER, B.: Structure and organisation of the pyrimidine biosynthesis pathway genes in Lactobacillus plantarum: a PCR strategy for sequencing without cloning. Gene 182, 37–43 (1996). FIGUEROA, C., DAVILA, A. M, POURQUIÉ, J.: Lactic acid bacteria of the sour cassava starch fermentation. Lett. Appl. Microbiol. 21, 126–130 (1995).

571

FRED, E. B., PETERSON, W. H., ANDERSON, J. A.: The characteristics of certain pentose-destroying bacteria, especially as concerns their action on arabinose and xylose. J. Biol. Chem. 48, 385–413 (1921). GRIMONT, P. A. D.: DNA/DNA hybridization in bacterial taxonomy, pp. 11–19. In: New horizons in Microbiology (A. SANNA, G. MORACE, eds.), Elsevier Science Publishers B. V., Amsterdam 1984. JOHANSSON, M.-L., QUEDNAU, M., AHRNÉ, S., MOLIN, G.: Classification of Lactobacillus plantarum by restriction endonuclease analysis of total chromosomal DNA using conventional agarose gel electrophoresis. Int. J. Sys. Bacteriol. 45, 670–675 (1995). JOHANSSON, M.-L., QUEDNAU, M., MOLIN, G., AHRNÉ, S.: Randomly amplified polymorphic DNA (RAPD) for rapid typing of Lactobacillus plantarum strains. Lett. Appl. Microbiol. 21, 155–159 (1995). LORTAL, S., VALENCE, F., BIZET, C., MAUBOIS, J.-L.: Electrophoretic pattern of peptidoglycan hydrolases, a new tool for bacterial species identification: application to 10 Lactobacillus species. Res. Microbiol. 148, 461–474 (1997). QUERE, F., DESCHAMPS, A., URDACI, M. C.: DNA probe and PCR-specific reaction for Lactobacillus plantarum. J. Appl. Microbiol. 82, 783–790 (1997). OSAWA, R., KUROISO, K., GOTO, S., SHIMIZU, A.: Isolation of tannin-degrading lactobacilli from humans and fermented foods. Appl. Environ. Microbiol; 66, 3093–3097 (2000). TAILLIEZ, P., TREMBLAY, J., EHRLICH, S. D., CHOPIN, A.: Molecular diversity and relationship within Lactococcus lactis, as revealed by randomly amplified polymorphic DNA (RAPD). Syst. Appl. Microbiol. 21, 530–538 (1998). TAILLIEZ, P., QUÉNÉE, P., CHOPIN, A.: Estimation de la diversité parmi les souches de la collection CNRZ: application de la RAPD à un groupe de lactobacilles. Lait 76, 147–158 (1996). VAN REENEN, C., DICKS, L. M. T.: Evaluation of numerical analysis of random amplified polymorphic DNA (RAPD)-PCR as a method to differentiate Lactobacillus plantarum and Lactobacillus pentosus. Curr. Microbiol. 32, 183–187 (1996). ZANONI, P., FARROW, J. A. E., PHILLIPA, B. A., COLLINS, M. D.: Lactobacillus pentosus (Fred, Peterson, and Anderson) sp. Nov., nov. rev. Int. J. Sys. Bacteriol. 37, 339–341 (1987). ZHONG, W., MILLSAP, K., BIALKOWSKA-HOBRZANSKA, H., REID, G.: Differentiation of Lactobacillus species by molecular typing. Appl. Environ. Microbiol. 64, 2418–2423 (1998).

Corresponding author: FRAN COISE ¸ BRINGEL, Laboratoire de Microbiologie et de Génétique, Université Louis-Pasteur Strasbourg-I, CNRS, 28 rue Goethe, 67083 Strasbourg, France Tel.: ++33-3-90 24 18 15; Fax: ++33-3-90 24 20 28; e-mail: [email protected]