INTERNATIONAL
JOURNAL OF SYSTEMATIC
BACTERIOLOGY,
Apr. 1990, p. 189-193
0020-7713/90/020189-05$02.00/0
Copyright 0 1990, International Union of Microbiological Societies
Vol. 40, No. 2
Restriction Endonuclease Patterns and Multivariate Analysis as a
Classification Tool for Lactobacillus spp.
MARIE STAHL,’ GORAN MOLIN,l* ANDERS PERSSON,l SIV AHRNE,l AND STEN STAHL2
Department of Applied Microbiology, Chemical Center, University of Lund, S-221 00 Lund, and Department of
Microbiology, University of Lund, S-223 00 Lund,2 Sweden
Three Lactobacillus plantarum and seven Lactobacillus reuteri strains were studied by using restriction
endonuclease analysis (REA) combined with principal-component analysis (PCA) and soft independent
modeling of class analogy (SIMCA). Chromosomal DNAs from the strains were extracted and cleaved with
restriction enzymes, and the DNA fragments were separated according to size by agarose gel electrophoresis.
Band patterns were read by using a laser densitometer. The procedure to obtain reproducible digestion
patterns with a suitable number of DNA fragments was optimized. Asp 718, ClaI and EcoRI were the most
suitable of the 17 restriction enzymes tested, each giving 30 to 50 DNA fragments down to a molecular weight
of 4.2 X lo6. The digestion patterns of all three enzymes, which gave a data set for 10 strains and 80 variables,
were used for classification by PCA and SIMCA. All strains were clearly separated, and the separation within
each species was in general accordance with data on DNA-DNA homology found in the literature. A class model
was created for L. reuteri 1073, 1068, 1063, 1048, and 1044. Closest to this model was L. reuteri DSM 20016T
(T = type strain), followed by L. reuteri DSM 20015; the most distant strains were the L. plantarum strains.
We concluded that restriction endonuclease analysis, combined with PCA and SIMCA, can be used for the
classification of Lactobacillus spp.
The genus Lactobacillus comprises about 50 validly described species; for historical reasons, in this genus circumscription of taxa has been based mainly on phenotypic
features (e.g., the ability to ferment carbohydrates). Furthermore, the members of the genus Lactobacillus are formally
differentiated from members of the genera Leuconostoc,
Lactococcus, and Pediococcus on morphological grounds;
in several cases this differentiation has caused systematic
problems, as it can be diacult to distinguish between rods
and cocci (23). Numerical studies have revealed the weakness in the phenotypic foundation (5,19). The systematics of
this group have been improved by chemotaxonomy and
DNA-DNA homology studies. Thus, the genus consists of
deterministically defined taxa, but the relationships among
the taxa are vague.
Examples of characteristics that indicate the weakness in
the present definition of the genus Lactobacillus are (i) the
wide range in guanine-plus-cytosine contents of DNAs from
different Lactobacillus species (32 to 53 mol%), (ii) the
different kinds of cross-linkage of the peptidoglycans in cell
walls, and (iii) the mingling in a phylogenetic tree of Lactobacillus, Leuconostoc, and Pediococcus strains as indicated
by rRNA sequencing data (25). Thus, there seems to be a
need to study the genus Lactobacillus with reclassification in
mind. What method should be applied? Sequencing of rRNA
is technically difficult and labor intensive. The accuracy of
using enzymes as genetic markers (15) can be questioned as
long as the majority of the information in the DNA remains
unknown. And DNA-DNA homology is, unfortunately, not
suitable outside the constraints of highly related species. An
alternative method is restriction endonuclease analysis
(REA), which, compared with rRNA sequencing, is technically easy to handle and may, in contrast to DNA-DNA
homology, have good resolution capacity at both the genus
and species levels.
REA has been proved to be useful in identifying serova-
* Corresponding author.
riants in different types of taxa, including leptospires (13,
17), Mycobacterium bovis (8), mycoplasmas (20, 21), Rickettsia prowazekii (22), Campylobacter jejuni (7), and Neisseria meningitidis (3). Furthermore, REA has been used for
identification of closely related species in the genera Bacteroides (6), Brucella (18), Mycobacterium (12), and Lactobacillus (16). Thus, there seems to be no doubt that REA is
a useful tool for identification.
On the other hand, the suitability of REA for classification
has been questioned (9, 21, 26), mainly on the grounds of
evaluation problems with large band spectra or, as stated by
Razin et al. (21), “The problem lies in cases of partial
similarity, as the cleavage data are not expressed by numbers, such as percent homology provided by DNA hybridization tests.” However, complex band patterns can be read
easily by using a laser densitometer, and similarities between
patterns can be expressed in numbers by using, for example,
principal-component analysis (PCA) and soft independent
modeling of class analogy (SIMCA). This method has been
tested previously for cluster analysis of REA patterns (24).
The aim of this study was to combine REA with PCA and
SIMCA to evaluate the similarities among the DNAs from
Lactobacillus strains belonging to two species, in order to
develop a tool for Lactobacillus classification. Seven Lactobacillus reuteri and three Lactobacillus plantarurn strains
were used in this study. Special efforts were made to
improve the REA method to give clear band patterns over a
wide molecular weight range.
MATERIALS AND METHODS
Strains and media. The strains which we used were L .
plantarum DSM 20174T (T = type strain), ATCC 8014, and
ATCC 10776, L . reuteri DSM 20016= and DSM 20015, and
five L . reuteri strains that were originally isolated from pig
intestines and had known DNA-DNA homology to strain
DSM 20016T (strains 1073, 1068, 1063, 1048, and 1044) (1).
The strains were cultured in Lactobacillus carrying medium containing 1%glucose (11).Stock cultures were stored
at -80°C.
189
Downloaded from www.microbiologyresearch.org
by
IP: 88.99.165.207
On: Sun, 18 Jun 2017 07:21:49
190
INT. J. SYST.BACTERIOL.
STAHL ET AL.
Preparation and purification of DNA. High-molecularweight chromosomal DNAs were prepared after the strains
were grown overnight in stationary cultures at 37°C in 100-ml
portions of Lactobacillus carrying medium containing 1%
glucose. Cells were harvested by centrifugation, washed
once in 10 ml of TES (50 mM NaCl, 100 mM Tris, 70 mM
disodium EDTA, pH 8.0), and suspended in 3 ml of TES
supplemented with 25% (wthol) sucrose. Then 2 ml of TES
containing 25% sucrose, 120 mg of lysozyme (grade VI;
Sigma Chemical Co., St. Louis, Mo.), and 140 U of mutanolysin (Sigma) was added. The cells were incubated at 42°C for
2 h, 3 mg of proteinase K (type XI; Sigma) was added, and
the preparation was incubated at 37°C for 30 min.
After cooling, 1 ml of a 5% NaCl solution, 1 ml of 20%
sodium dodecyl sulfate, and 11.4 ml of TES were added to
the cell suspension. If required, the suspension was gently
stirred with a glass rod until it became homogeneous and
then heated at 65°C for 15 min. The cell lysate was extracted
once with redistilled phenol-chloroform (1:1)and once with
chloroform-isoamylalcohol (24: 1). The DNA was then precipitated with ethanol at -20°C.
The chromosomal DNA was separated from covalently
closed circular plasmid DNA by dye buoyant density centrifugation in a CsCl gradient with ethidium bromide (2). The
purity and concentration of the chromosomal DNA were
monitored spectrophotometrically at 260 and 280 nm. The
molecular weight of each chromosomal DNA was at least 32
x lo6.
Digestion with restriction endonucleases. All restriction
enzymes were obtained from Boehringer Mannheim Scandinavia AB, Bromma, Sweden. We tested mainly hexanucleotide-recognizing enzymes, including ApaI, Asp 718,
BamHI, BclI, ClaI, EcoRI, HpaI, NaeI, PstI, PvuII, SacI,
SalI, SmaI, SphI, XbaI, XhoI, and NotI. Generally, 0.75 pg
of DNA in 20 pl of buffer was digested for 4 to 6 h with 10 U
of enzyme at the recommended temperature. Asp 718, CfaI,
and EcoRI were found to be the best enzymes; i.e., they
gave reproducible digestion patterns and complete DNA
digestion into a suitable number of fragments. Each of these
enzymes generated about 30 to 50 DNA fragments having
molecular weights of more than 4.2 x lo6.
Agarose gel electrophoresis. Gel electrophoresis was carried out by using submerged horizontal agarose slab gels.
The gels consisted of 150 ml of 0.9% agarose (ultrapure DNA
grade; low electroendo-osmosis; Bio-Rad Laboratories,
Richmond, Calif.) and were cast as slab gels (150 by 235
mm). This size of gels gave sharp, well-separated bands
down to a molecular weight of 4.2 x lo6 (Fig. 1). Longer gels
separated a wider range of fragments, but the bands were not
sharp enough for reproducible evaluation.
The gels were run at a constant voltage of 40 V for 69 h
with cooling to about 15°C. Because of the extended period
of electrophoresis, the buffer (89 mM Tris, 23 mM H3P0,,
2.5 mM disodium EDTA, pH 8.3) was recirculated during
runs (16 ml/min). Minimal band distortion 'and maximal
sharpness were achieved by applying the sample DNA in
Ficoll loading buffer (2 g of Ficoll, 8 ml of water, 0.25%
bromphenol). Approximately 0.75 pg of DNA was loaded
into each well (width, 8 mm). The gels were stained for 1 h
with ethidium bromide (1.5 pg/ml) and destained in distilled
water. DNA fragments were visualized at 302 nm with a UV
transilluminator (UVP Inc., San Gabriel, Calif.) and photographed .
Reading of band patterns. Band patterns on photo-negatives were scanned with a laser densitometer (UltroScan
XL; LKB-Produkter AB, Bromma, Sweden). Each track
FIG. 1. Agarose gel electrophoresis of L. plantarum ATCC 8014
digested with Asp 718, ClaI, and EcoRI (lanes 1 through 3, respectively) and L. plantarum DSM 20174T digested with Asp 718, ClaI,
and EcoRI (lanes 4 through 6, respectively). The bottom band of
each lane is pHC 79 digested with EcoRI (internal standard).
was scanned five times with a total scan width of 4,800 pm,
and data points were collected at 40-pm intervals.
In order to remove background noise, smoothing was
performed on a sliding window of data points. These points
were taken in groups of 21, with each collection moving 1
point. (Instruction manual for model LKB 2222-020 Ultroscan XL laser densitometer, LKB-Produkter AB, 1987). The
data were processed by using the LKB 2400 GelScan XL
software package (LKB-Produkter AB). Absorbance curves
were integrated, and relative peak areas were calculated.
The absorbance curves from Asp 718 digestions were divided into 40 intervals, and the curves from CfaI and EcoRI
digestions were divided into 20 intervals each.
Molecular weights were calculated from a standard curve
made from undigested lambda DNA (Boehringer Mannheim
Scandinavia AB), lambda DNA digested with HindIII, and
lambda DNA digested with Asp 718. To compensate for
differences in mobility between different gels, each lane
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sun, 18 Jun 2017 07:21:49
VOL.40, 1990
I
-4
DSM 20015
I
I
1048=
1063.
U
E
.
RESTRICTION ENDONUCLEASE PATTERNS
,
.13
k
191
1
0
.c,
0
M ZUl-f4
1073
_ _
I
.03 -
n
M
E
;d"
mi
DSY 20016
-.07
-
-.17
-
H
*
0
Q)
ATCC 10776
n
-8.6 I
-8.25
U
I
I
I
I
v1
I
-5.25 -2.25
.75
3.75
First principal component
FIG. 2. First and second PC scores with the projection based on
the complete data set. The first PC score vector separates the L.
reuteri strains (m) from L. plantarum (0).The second PC score
vector contains little class-separating information. The axes are in
arbitrary units.
contained pHC 79 (molecular weight, 4.2 x lo6; Boehringer
Mannheim Scandinavia AB) linearized with EcoRI as an
internal standard.
Multivariate data analysis. Restriction endonuclease patterns were compared by using the SIMCA-3B, version TMP
computer program (Sepanova AB, Enskede, Sweden). The
analysis consisted of a PCA (14) on the whole set of data
without direct reference to class separation. In addition, the
SIMCA method was used to investigate dissimilarity between classes. SIMCA has been described in detail by Wold
(27) and Blomquist et al. (4). The basic idea of SIMCA is that
multivariate data which are measured for a group of similar
objects (a class) are well approximated by a simple principalcomponent (PC) model. This gives the potential to create
models of classes (for example, species) and thus allows
unknown objects to be classified. This is accomplished by
fitting separate PC models to each class. Unknown objects
can then be assigned to a single class, several classes, or no
class.
RESULTS
Data analysis. The data set consisted of 10 objects (strains)
and 80 variables (total, relative peak areas of intervals). The
data set plot of the first and second PCs gave good separation
between the two species (Fig. 2). The loading plot in Fig. 3
shows the contribution of each variable to the separation of
the objects. The even distribution in Fig. 3 shows that the
separation achieved in Fig. 2 was due to the influence of
many variables.
A class model was created for L. reuteri 1073,1068,1063,
1048, and 1044 (Table 1).Closest to the class model was L.
reuteri DSM 20016T, with a distance of 1.3 (residual standard
deviation) to the class model. The other reference strain of
L. reuteri had a distance of 1.7, and the L. plantarum strains
had distances between 2.4 and 2.9.
In order to check the reproducibility of the test system,
five parallel digestions of L. plantarum DSM 20174T DNA
were made, and the results were compared by using PCA
with the results of digestions of DNAs from strains ATCC
FIG. 3. First and second loading vectors giving the loading plot
corresponding to Fig. 2. The axes are in arbitrary units.
8014 (95% DNA-DNA homology to strain DSM 20174T [lo])
and ATCC 10776 (57% DNA-DNA homology to strain DSM
20174T [lo]). The three strains were clearly separated in a
score plot (Fig. 4). Figure 4 shows the maximum separation
of the seven digestions of L. plantarum DNAs without
interference from the L. reuteri DNAs. The five parallel
digestions of strain DSM 20174T DNA clustered closely
together in the top right corner of the plot and were well
separated from the strain ATCC 8014 digestion. Thus, we
demonstrated that the reproducibility of the method was
good enough to distinguish between strains with a level of
DNA-DNA homology of 95%. Moreover, the results shown
in Fig. 4 suggest that the lower limit of resolution is significantly below a 5% difference in DNA-DNA homology.
DISCUSSION
According to Axelsson and Lindgren (I), the levels of
DNA-DNA homology between strain DSM 20016T and
strains 1063, 1068, 1073, 1044, and 1048 were 82, 72,72, 61,
and 54%, respectively. Strain 1063 was more distant from
the type strain in the PCA score plot (Fig. 2), but otherwise
the interrelationships among the strains were in general
accordance with the results of the DNA-DNA homology
studies. L. ptantarum DSM 20174= has been shown to have
95% DNA-DNA homology to strain ATCC 8014, but only
57% homology to strain ATCC 10776 (lo), which is in
TABLE 1. Class model of L. reuteri strains from pig intestines
obtained by SIMCA classification
Organism
L. plantarum DSM 20174T ................................
L. plantarum ATCC 8014 ..................................
L. plantarum ATCC 10776 ................................
L. reuteri DSM 2001ST .....................................
L. reuteri DSM 20015 .......................................
L. reuteri 1073 ................................................
L . reuteri 1068 ................................................
L. reuteri 1063 ................................................
L . reuteri 1048 ................................................
L. reuteri 1044 ................................................
a
Corrected for loss at degrees of freedom.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sun, 18 Jun 2017 07:21:49
Residual standard
deviation
2.4352
2.8967
2.6355
1.3186
1.7390
0.5936a
0.4198O
0.3752"
0.5521"
0.3524"
192
STAHL ET AL.
&
INT.J. SYST.BACTERIOL.
I
lo
4
cd
PI
.r(
0
.g -4.5
-
k
P
I
g0 -7.5
-
a2
ATCC 8014
u2
-10.5
-
I
I
I
fragments to define a certain taxon. Such DNA fragments
may then be used as probes for identification of that specific
taxon. For example, if such a probe is designed for the L .
reuteri strains used in this study, fragments in the intervals
oriented to the top right of Fig. 3 ought to be chosen. In
order to find fragments that are shared by all of the L. reuteri
strains (identification markers of the species), a new loading
plot from which the L. plantarum strains are excluded can be
made. All shared fragments should then be oriented around
origo.
In conclusion, in this study we demonstrated the potential
for using REA combined with PCA and SIMCA as a tool in
systematic classification. Furthermore, this method might
provide a practical means for studying the species concept,
as well as the infrastructure, in the genus Lactobacillus.
>
ACKNOWLEDGMENTS
FIG. 4. First and second PC score projection for L . plantarum
ATCC 8014, L. plantarum ATCC 10776, and five parallel digestions
of L . plantarum DSM 20174= DNA. The axes are in arbitrary units.
agreement with our results (Fig. 2); Dellaglio et al. (10) did
not regard strain ATCC 10776 as an L . plantarum strain. The
locations of different strains in a PCA score plot (Fig. 2 and
4) reveal the interrelationships among the strains when they
are compared simultaneously; i.e., each strain is compared
with all of the rest. Therefore, the PCA score plot cannot be
directly compared with DNA-DNA homology data, in which
only two strains are compared at the same time.
Dye buoyant density centrifugation separates covalently
closed circular plasmid DNA from chromosomal DNA, but
open circular and linear forms of large plasmids may contaminate the preparations and obstruct the interpretation of
the band pattern. However, with PCA it is possible to
determine whether the distribution of the objects in the score
plot is caused by a few or several variables in the loading
plot. In this study strains 1044 and 1048 were both found to
contain a plasmid with a molecular weight of 5.5 x lo6. The
plasmids which were not digested by ClaI (data not shown)
were traced because strains 1048 and 1044 deviated from all
of the other strains in a preliminary PCA (data not shown).
In addition, the loading plot showed that this separation was
caused by one single interval in the digestion pattern of CZaI.
After the relative peak area of the plasmid bands had been
subtracted from the data, the loading plot showed an even
distribution of the variables, and strains 1044 and 1048
clustered with the other L . reuteri strains. This way to detect
plasmid interference represents an advantage over the DNADNA homology method.
Other advantages of PCA are the graphic display of the
results (Fig. 2 through 4) and the fact that the number of
objects and the number of variables are not restricted to a
certain ratio. The PC plots of the data are equivalent to
eigenvector plots (14),and the projections preserve as much
as possible of the data variance in the space made by the
variables. Furthermore, the orientation of the loading plot
(Fig. 3) is directly related to the orientation of the score plot
(Fig. 2); e.g., variables far from origo in the horizontal
direction are important for the horizontal resolution in the
score plot. This allows rapid visual interpretation of the
dominant patterns by using two plots, the score plot and the
corresponding loading plot.
The two plots can be used in searching for suitable DNA
We thank Michael Sjostrom for his kind help with SIMCA.
This work was supported by the Swedish Council for Forestry and
Agricultural Research.
LITERATURE CITED
1. Axelsson, L., and S. Lindgren. 1987. Characterization and DNA
homology of Lactobacillus strains isolated from pig intestine. J.
Appl. Bacteriol. 62:433440.
2. Bauer, W., and J. Vinograd. 1968. The interaction of closed
circular DNA with intercalative dyes. I. The superhelix density
of SV 40 DNA in the presence and absence of dye. J. Mol. Biol.
33:141-171.
3. Bjorvatn, B., V. Lund, B.-E. Kristiansen, L. Korsnes, 0.
Spanne, and B. Lindqvist. 1984. Applications of restriction
endonuclease fingerprinting of chromosomal DNA of Neisseria
meningitidis. J. Clin. Microbiol. 19:763-765.
4. Blomquist, G., E. Johansson, B. Soderstrom, and S. Wold. 1979.
Classification of fungi by means of pyrolysis-gas chromatography-pattern recognition. J. Chromatogr. 173:19-32.
5. Borch, E., and G. Molin. 1988. Numerical taxonomy of psychrotrophic lactic acid bacteria from prepacked meat and meat
products. Antonie van Leeuwenhoek J. Microbiol. Serol. 54:
301-323.
6. Bradbury, W. C., R. G. E. Murray, C. Mancini, and V. L.
Morris. 1985. Bacterial chromosomal restriction endonuclease
analysis of the homology of Bacteroides species. J. Clin. Microbiol. 21:24-28.
7. Bradbury, W. C., A. D. Pearson, M. A. Marko, R. V. Congi, and
J. L. Penner. 1984. Investigation of a Campyiobacter jejuni
outbreak by serotyping and chromosomal restriction endonuclease analysis. J. Clin. Microbiol. 19:342-346.
8. Collins, D. M., and G. W. De Lisle. 1985. DNA restriction
endonuclease analysis of Mycobacterium bovis and other members of the tuberculosis complex. J. Clin. Microbiol. 21562564.
9. Crameri, R., G. Hintermann, and R. Hutter. 1983. Deoxyribonucleic acid restriction endonuclease fingerprint characterization of actinomycete strains. Int. J. Syst. Bacteriol. 33:652-655.
10. Dellaglio, F., V. Bottazi, and M. Vescovo. 1975. Deoxyribbonucleic acid homology among Lactobacillus species of the subgenus Streptobacterium Orla Jensen. Int. J. Syst. Bacteriol.
25: 160-172.
11. Efthymion, C., and P. A. Hansen. 1962. An antigenic analysis of
Lacrobacillus acidophilus. J. Infect. Dis. 110:258-267.
12. Garcia, M. J., and E. Tabards. 1986. Separation of Mycobacterium gadium from other rapidly growing mycobacteria on the
basis of DNA homology and restriction endonuclease analysis.
J. Gen. Microbiol. 132:2265-2269.
13. Hathaway, S. C., R. B. Marshall, T. W. A. Little, S. A. Headlam,
and P. J. Winter. 1985. Identification by cross-agglutination
absorption and restriction endonuclease analysis of leptospires
of the Pomona serogroup isolated in the United Kingdom. Res.
Vet. Sci. 39:151-156.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sun, 18 Jun 2017 07:21:49
VOL. 40, 1990
RESTRICTION ENDONUCLEASE PATTERNS
14. Jolliffe, I. T. 1986. Principal component analysis. SpringerVerlag, New York.
15. London, J., and N. M. Chace. 1983. Relationships among lactic
acid bacteria demonstrated with glyceraldehyde-3-phosphate
dehydrogenase as an evolutionary probe. Int. J. Syst. Bacteriol.
33:723-737.
16. Manachini, P. L., and C. Parini. 1983. DNA-restriction endonuclease cleavage patterns, DNA sequence similarity and phenotypical characteristics in some strains of Lactobacillus helveticus and Lactobacillus jugurti. Antonie van Leeuwenhoek J.
Microbiol. Serol. 49:143-152.
17. Marschall, R. B., B. E. Wilton, and A. J. Robinson. 1981.
Identification of leptospira serovars by restriction-endonuclease
analysis. J. Med. Microbiol. 14:163-166.
18. O’Hara, M. J., D. M. Collins, and G. W. DeLisle. 1985.
Restriction endonuclease analysis of Brucella ovis and other
Brucella species. Vet. Microbiol. 10:425429.
19. Priest, F. G., and E. A. Barbour. 1985. Numerical taxonomy of
lactic acid bacteria and some related taxa, p. 137-163. In M.
Goodfellow, D. Jones, and F. G. Priest (ed.), Computer-assisted
bacterial systematics. Academic Press, Inc. (London), Ltd.,
London.
20. Razin, S., R. Harasawa, and M. F. Barile. 1983. Cleavage
patterns of the mycoplasma chromosome, obtained by using
restriction endonucleases as indicators of genetic relatedness
among strains. Int. J. Syst. Bacteriol. 33:201-206.
193
21. Razin, S., J. G. Tully, D. L. Rose, and M. F. Barile. 1983. DNA
cleavage patterns as indicators of genotypic heterogeneity
among strains of Acholeplasma and Mycoplasma species. J.
Gen. Microbiol. 129:1935-1944.
22. Regnery, R. L., T. Tzianabos, J. J. Esposito, and J. E. McDade.
1983. Strain differention of epidemic typhus rickettsiae (Rickettsia prowazekii) by DNA restriction endonuclease analysis.
Curr. Microbiol. 8:355-358.
23. Schleifer, K. H., J. Kraus, C. Dvorak, R. Kilpper-Biilz, M. D.
Collins, and W. Fischer. 1985. Transfer of Streptococcus lactis
and related streptococci to the genus Lactococcus gen. nov.
Syst. Appl. Microbiol. 6:183-195.
24. Sorensen, B., E. S. Falk, E. Wissloff-Nilsen, B. Bjorvatn, and
B. E. Kristiansen. 1985. Multivariate analysis of Neisseria DNA
restriction endonuclease patterns. J. Gen. Microbiol. 131:30993104.
25. Stackebrandt, E., V. J. Fowler, and C. R. Woese. 1983. A
phylogenetic analysis of lactobacilli, Pediococcus pentosaceus
and Leuconostoc mesenteroides. Syst. Appl. Microbiol. 4:
326-337.
26. Verger, J. M., F. Grimont, P. A. D. Grimont, and M. Grayon.
1987. Taxonomy of the genus Brucella. Ann. Inst. Pasteur
(Paris) 138:235-238.
27. Wold, S. 1976. Pattern recognition by means of disjoint principal
components models. Pattern Recogn. 8:127-139.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sun, 18 Jun 2017 07:21:49