INTERNATIONAL
JOURNAL OF SYSTEMATIC
BACTERIOLOGY,
Oct. 1990, p. 405408
0020-7713/90/040405-04$02.00/0
Copyright 0 1990, International Union of Microbiological Societies
Vol. 40,No. 4
Multivariate Analyses of Carbohydrate Data from
Lipopolysaccharides of Actinobacillus (Haemophilus)
actinomycetemcornitans, Haemop hilus aphrophilus ,
and Haemophilus paraphrophilus
ILIA BRONDZlt"
AND
INGAR OLSEN2
Research Department, National Institute of Occupational Health, Umed, Sweden, and
Department of Microbiology, Dental Faculty, University of Oslo, Oslo, Norway2
The taxonomic distinction between Actinobacillus (Haemophilus) actinomycetemcomituns and Haemophilus
aphrophilus and the taxonomic distinction between H. aphrophilus and Haemophilus paraphrophilus have been
questioned. This study was done to determine whether multivariate statistical analyses of carbohydrate data
from lipopolysaccharides could be used to distinguish between these closely related species. Lipopolysaccharides were extracted with phenol-water and purified. Carbohydrates were assessed by using gas chromatography and gas chromatography-mass spectrometry after methanolysis and derivatization with trifluoroacetic
acid anhydride. The lipopolysaccharides from all of the species contained rhamnose, fucose, galactose, glucose,
L-glycero-D-mannoheptose, and glucosamine plus galactosamine, but in varying amounts. A. actinomycetemcomitans and H . paraphrophilus also contained D-glycero-D-mannoheptose, while H. aphrophilus did not.
Sample- and variable-oriented principal-component analyses of the carbohydrate data clearly distinguished
among A . actinomycetemcomitans, H . aphrophilus, and H . paraphrophilus. Soft independent modelling of class
analogy showed that no sample in the A . actinomycetemcomitans class fell within the 95% confidence limits of
the H. aphrophilus class. H. paraphrophilus fell outside both classes.
Actinobacillus actinomycetemcomitans is morphologically and biochemically very similar to Haemophilus aphrophilus, and the taxonomic positions of these organisms have
been questioned ( 5 , 11). These species do not require hemin
and NAD for growth and therefore do not fulfil the present
criteria for inclusion in the genus Haemophilus (9). Although
the level of homology between A. actinomycetemcomitans
and the type species of the genus Haemophilus (Haemophilus injluenzae) was not examined, it was recently proposed
that A. actinomycetemcomitans should be transferred to the
genus Haemophilus as Haemophilus actinomycetemcomitans (13). Biochemical characteristics and DNA-DNA homology data do not distinguish H . aphrophilus from Haemophilus paraphrophilus, except for a single strain which
may not be a H . paraphrophilus strain (15). However, H.
aphrophilus and H . paraphrophilus can be distinguished by
their V-factor requirements. We recently demonstrated that
multivariate analyses of chemical and enzymatic characterization data for whole cells are well suited to distinguish
between closely related species of the Actinobacillus-Haemophilus-Pasteurella group (family Pasteurellaceae) (6a)
and that lipopolysaccharide (LPS) is a suitable component
for chemotaxonomic differentiation of A. actinomycetemcomitans, H . aphrophilus, and H . paraphrophilus (1, 3, 4,
6). In this study, we extended these observations by using
multivariate statistical analyses of carbohydrate data from
LPSs to distinguish between A . actinomycetemcomitans and
H. aphrophilus and between H . aphrophilus and H . paraphrophilus .
MATERIALS AND METHODS
Organisms. The 14 bacterial strains which we tested,,
including type strains, other reference strains, and clinical
strains of A . actinomycetemcomitans, H . aphrophilus, and
H . paraphrophilus, are shown in Table 1. The strains were
obtained directly from the American Type Culture Collection, Rockville, Md., the Forsyth Dental Center, Boston,
Mass., and the Royal Dental College, Aarhus, Denmark.
The organisms were cultivated in brain heart infusion broth
(Difco Laboratories, Detroit, Mich.) in air containing 10%
CO, for 5 days at 37°C. To grow H . paraphrophilus, the
broth was supplemented with filter-sterilized NAD (1 mg/
liter) and hemin (5 mg/liter). The organisms were harvested
by centrifugation, washed three times in deionized, distilled
water, and lyophilized over diphosphorous pentoxide (E.
Merck AG, Darmstadt, Federal Republic of Germany).
Lyophilized cells were kept at -20°C under nitrogen gas.
LPS. LPS was isolated by using the phenol-water procedure and was purified by ultracentrifugation (1). Methanolysis was carried out in 2 M hydrochloric acid and in
anhydrous methanol for 24 h at 85°C. The methanolysate
was derivatized in acetonitrile and trifluoroacetic acid anhydride (1).
Gas chromatography and gas chromatography-mass spectrometry. The derivatized samples were analyzed with a
model Sigma 3 gas chromatograph (The Perkin-Elmer Corp.,
Norwalk, Conn.) fitted with a model Sigma 10 electronic
integrator (1). The type CP-Sil 5 (poly-dimethylsiloxane)
glass capillary column (Chrompack, Middelburg, The Netherlands) which we used was 25 m long by 0.22 mm (inner
diameter) and had a film thickness of 0.14 Fm and a height
equivalent of a theoretical 0.25-nm plate. The identities of
the derivatized carbohydrates were determined by cochromatography with authentic standards and by gas chromatography-mass spectrometry (1). The quantities of individual
* Corresponding author.
t Present address: Norwegian Plant Protection Institute, Box 70,
1432 h - N L H , Norway.
405
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Fri, 16 Jun 2017 05:10:53
406
INT. J . SYST.BACTERIOL.
BRONDZ AND OLSEN
TABLE 1. Bacteria investigated
Sample
no.
Strainn
A . actinomycetemcomitans strains
ATCC 33384T
(= NCTC 9710T)
ATCC 29524
ATCC 29522
FDC 511
HK435
FDC N27
FDC Y4
H . aphrophilus strains
ATCC 33389T
(= NCTC 5906=)
ATCC 19415
(= NCTC 5886)
FDC 655
FDC 654
FDC 626
FDC 621
H . paraphrophilus
ATCC 29241T
(= NCTC 10557T)
Original
sourceU
Site of origin
1
ATCC
Lung abscess
2
3
4
5
6
7
ATCC
ATCC
FDC
Kilian
FDC
FDC
Chest aspirate
Mandibular abscess
Periodontitis
Pus
Periodontitis
Periodontitis
8
ATCC
Endocarditis
9
ATCC
Endocarditis
10
11
12
13
14
FDC
FDC
FDC
FDC
ATCC
Periodontitis
Periodontitis
Periodontitis
Periodontitis
Paronychia
_____
ATCC, American Type Culture Collection, Rockville, Md.; NCTC,
National Collection of Type Cultures, London, England; FDC, Forsyth
Dental Center, Boston, Mass.; Kilian, M. Kilian, Royal Dental College,
Aarhus, Denmark.
components were determined from their peak areas and
were corrected by using the molar response factor (1).
Statistical analyses. In this paper we describe the distribution of seven monosaccharides in LPSs from 14 strains of A .
actinomycetemcornitans, H . aphrophilus, and H . paraphrophilus. The variables and measurements used in this study
are shown in Table 1. The statistical analysis methods which
we used included the following two projection methods:
principal-component analysis (PCA) (8, 16, 19) and partial
least-squares discriminant analysis (7, 18). With these methods the original space for the variable measurements was
projected down onto two low-dimensional subspaces. One of
these was sample related, and the other was variable related.
The sample-related projection could be inspected for similarities among the samples, and the variable-related projection could be used to detect similarities among the variables.
This projection also identified which of the variables contributed to the sample-related projection. Cross-validation
was used to determine the complexity of both models (17,
18). In soft independent modelling of class analogy, which
also was used in this study, a class of samples was described,
and the complexity of the model was determined by crossvalidation (14). From the deviation of the samples from their
PCA model, approximate class borders were constructed by
using an F test.
RESULTS
Gas chromatography and gas chromatography-mass spectrometry. The results of the individual gas chromatographic
and gas chromatographic-mass spectrometric studies have
been described elsewhere (1, 3). Therefore, only a brief
summary is given below. The LPSs from all of the species
contained rhamnose, fucose, galactose, glucose, L-glyceroD-mannoheptose, and glucosamine plus galactosamine (Table 2). In H . paraphrophilus LPS, rhamnose and fucose were
found only in trace amounts. The amounts of galactose were
similar in H . aphrophilus and H . paraphrophilus LPSs and
were approximately two times higher in the LPSs of these
two species than in A . actinomycetemcornitans LPS. The
amount of glucose was about two times higher in H . paraphrophilus LPS than in A . actinomycetemcornitans LPS.
The amount of L-glycero-D-mannoheptose was approximately two times higher in A . actinomycetemcornitans LPS
and three times higher in H . aphrophilus LPS than in H .
paraphrophilus LPS. The amounts of glucosamine plus
galactosamine were nearly three times higher in A . actinomycetemcomitans and H . aphrophilus LPSs than in H .
paraphrophilus LPS. The most outstanding feature of the
data from the gas chromatography and gas chromatographymass spectrometry analyses was the finding that D-glyceroD-mannoheptose was present in A . actinomycetemcomitans
TABLE 2. Sugars in LPSs from species belonging to the genera Actinobacillus and Haemophilus as determined by gas chromatography
and mass spectrometry after methylation and derivatization
Relative amta of
Strain
A . actinomycetemcornitans strains
ATCC 33384T
ATCC 29524
ATCC 29522
FDC 511
HK485
FDC N27
FDC Y4
H . aphrophilus strains
ATCC 33389T
ATCC 19415
FDC 655
FDC 654
FDC 626
FDC 621
H . paraphrophilus ATCC 29241T
Fucose
Galactose
Glucose
DD-Hep
LD-Hep
9.8
13.0
11.8
6.8
8.3
8.2
13.O
5.7
10.7
6.9
7.0
10.0
9.2
8.8
10.1
9.3
7.1
8.6
9.6
9.7
10.2
36.5
31.8
30.9
29.9
30.1
32.2
31.4
14.4
11.8
12.3
16.7
16.2
12.3
12.5
17.8
17.1
23.0
26.3
21.6
20.5
18.8
5.6
6.2
8.0
4.7
4.1
7.8
5.3
7.2
8.1
8.5
6.2
10.7
6.9
<o. 1
4.6
4.8
6.0
4.5
6.7
5.0
20.2
20.3
16.5
20.8
18.9
18.4
15.2
40.9
40.1
34.2
33.6
33.4
38.6
65.9
8.0
23 .O
22.9
29.5
30.0
24.4
26.4
8.7
4.1
3.8
5.4
4.7
5.8
4.7
1.4
co.1
GlcN
+ GalN
Rhamnose
The relative amounts are expressed as the areas of the peaks in a chromatogram as percentages of the total area of the peaks of all of the substances which
were examined. The values are means based on 12 runs (strain ATCC 33384T and ATCC 33389T LPSs), 18 runs (strain ATCC 29241T LPS), or 6 runs (LPSs from
all of the remaining strains). Data were compiled from references 1 and 3. Abbreviations: DD-Hep, D-glycero-D-mannoheptose; LD-Hep, L-glyceroD-mannoheptose; GlcN + GalN, glucosamine plus galactosamine.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Fri, 16 Jun 2017 05:10:53
VOL. 40, 1990
MULTIVARIATE ANALYSES OF LPS CARBOHYDRATES
t1
9
011
407
6
010
1
' 3
3
012
00
8 9
71
0
o4
1.
.5
6
O.
O3
'
0
2
.2
4
5
t2
FIG. 1. Sample-oriented (i.e., bacterial-strain-oriented) principal-component projection ( t l / t 2 ) .The two principal components, t ,
and t,, describe 82.1 and 11.8%, respectively, of the variance in the
carbohydrate data from LPSs (1, 3). The positions of strains of A.
actinomycetemcomitans (0)(samples 1 to 7), H. aphrophilus (0)
(samples 8 to 13), and H. paraphrophilus (m) (sample 14) are shown.
For sample identification, see Table 1.
and H . paraphrophilus LPSs and absent in H . aphrophilus
LPS.
Statistical analyses. In the PCA of the carbohydrate data
from LPSs of A . actinomycetemcomitans, H . aphrophilus,
and H . paraphrophilus strains, the first two principal-component score vectors (tl and t2) were plotted against each
other (Fig. 1). This sample-oriented (i.e., bacterial-strainoriented) projection described the two largest variants of the
data matrix (1, 3). In this projection the A . actinomycetemcomitans strains (samples 1to 7 ) and H . aphrophilus strains
(samples 8 to 13) formed two homogeneous clusters which
were well separated from each other and from H . paraphrophilus (sample 14).
The variables (i.e., carbohydrates of LPSs [Table 21) that
contributed most to the first score vector ( t l ) were fucose
(variable 2), rhamnose (variable 1), D-glycero-D-mannoheptose (variable 5 ) , and glucosamine plus galactosamine
(variable 7) (Fig. 2), and the variables that contributed most
to the second score vector ( I * ) were galactose (variable 3),
glucose (variable 4), and L-glycero-D-mannoheptose (variable 6).
Soft independent modelling of class analogy analysis demonstrated that none of the samples in the A . actinomycetemcomitans class (Samples 1 to 7) fell within the 95% confidence limits of the H . uphrophilus class (samples 8 to 13)
(Fig. 3 ) . H . paraphrophilus (sample 14) fell outside both
classes.
DISCUSSION
FIG. 2. Variable-oriented (i.e., bacterial-character [carbohydrates of LPSI-oriented) principal-component projection (p1/p2).
The numbers indicate the following variables: 1, rhamnose; 2,
fucose; 3, galactose; 4, glucose; 5, D-glycero-D-mannoheptose; 6,
L-glycero-D-mannoheptose;7, ghcosamine plus galactosamine.
nomic consideration. The fatty acid content of A . actinomycetemcomitans LPS was two times higher than the fatty
acid content of H . aphrophilus LPS (6). In whole cells, fatty
acids (in contrast to carbohydrates) did not contribute to
differentiation among A . actinomycetemcomituns, H . aphrophilus, and H . paraphrophilus (6a).
Our data also suggested that PCA, partial least-squares
discriminant analysis, and soft independent modeling of
class analogy are suitable statistical methods for taxonomic
distinction. Using the same kinds of analyses for data from
SD1
I
I
I
10121389
OOOO
I
In this study, which was based on multivariate analyses of
carbohydrate data from LPSs, the data clearly differentiated
among A . actinomycetemcomitans, H . aphrophilus, and H .
paraphrophilus. This supported the hypothesis that LPS,
particularly its core region, can be an excellent component
for chemotaxonomy (6). Also the 0-chain (for serotyping)
and lipid A of LPS are valuable structures for chemotaxo-
0.2
I
1
0.4
0.6
11
0
-
SD2
FIG. 3. Residual standard deviation for each sample when the
samples were fitted to the A. actinomycetPmcomitans (0)(samples
1 to 7) class model (SD1) and the H. aphrophilus (0)(samples 8 to
13) class model (SD2). The dotted lines mark the approximate class
borders based on F tests (P = 0.05). Sample 14 (4) was an H.
p a raphrophilus strain.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Fri, 16 Jun 2017 05:10:53
408
INT. J . SYST.BACTERIOL.
BRONDZ AND OLSEN
studies of cellular carbohydrate composition, lysis kinetics
during EDTA and EDTA-plus-lysozyme treatments, and
methylene blue reduction (i.e., 23 variables) we were able to
differentiate among the same species (Brondz et al., in
press). In this study, only seven monosaccharide variables
provided the same distinction.
No class model could be made for H . paraphrophilus,
which was represented by one strain. However, the H .
paraphrophilus sample, which was the type strain, clearly
fell outside both the A . actinomycetemcomitans class and
the H . aphrophilus class. Therefore, the data from this
study, which also included the type strains of A . actinomycetemcomitans and H . aphrophilus, supported the establishment of A. actinomycetemcomitans, H . aphrophilus, and H .
paraphrophilus as distinct species (10).
D-Glycero-D-mannoheptose was found in A . actinomycetemcomitans and H . paraphrophilus LPSs but not in H .
aphrophilus LPS. We have also detected D-glycero-D-mannoheptose in other clinically important species belonging to
the Actinobacillus-Haemophilus-Pasteurellagroup, such as
H . injluenzae, Pasteurella ureae, Pasteurella haemolytica,
and Pasteurella multocida (2). This raises the question of
whether D-glycero-D-mannoheptose can be used as a chemotaxonomic marker for species belonging to the family Pasteurellaceae. A true taxonomic marker is a characteristic of
organisms that conveys information which is also conveyed
by numerous other characteristics of the organisms. A true
marker contributes to the separation of bacterial taxa, and
when it is excluded from a PCA, this exclusion only marginally influences the separation of the taxa in the analysis.
Conversely, a false marker is a characteristic that separates
taxa from each other and expresses variability within the
taxa which is not reflected in the other characteristics of the
taxa. If such a marker is excluded from a PCA, the taxa are
randomly distributed. We found that D-glycero-D-mannoheptose is a true taxonomic indicator since it was not
necessary for discrimination between the A . actinomycetemcomitans and H . aphrophilus classes (6a). If H . aphrophilus
proves to be the only species in the family Pasteurellaceae
that lacks D-glycero-D-mannoheptose, its taxonomic position may require reconsideration.
H . paraphrophilus LPS contained low levels of lipid
A-derived glucosamine and galactosamine, which might suggest the presence of a glucan (e.g., glycogen).
Important biological activity of the polysaccharide moiety
of LPS appears to reside in the lipid A-proximal inner core
region, which is characterized by the presence of 2-keto3-deoxyoctonate and heptoses. For example, synthetic
(nonpyrogenic) or bacterial core-derived heptose I-heptose
II-2-keto-3-deoxyoctonate I trisaccharide induces the release of interleukin 1 from human mononuclear cells to
supernatants (12). Lipid A is very important for the release
of intracellular interleukin 1. The lack of D-glycero-D-mannoheptose and the lower fatty acid content in H . aphrophilus
LPS compared with A . actinomycetemcomitans LPS may
contribute to the low periodontopathogenic potential of H .
aphrophilus compared with the periodontopathogenic potential of the closely related organism A . actinomycetemcornitans.
ACKNOWLEDGMENTS
We thank Norsk Dental Depots Fond For Odontologisk Forskning and Nordisk Ministerrid for financial support, stip. 35/90.
LITERATURE CITED
1. Brondz, I., and I. Olsen. 1984. Differentiation between Actinobacillus actinomycetemcomitans and Haemophilus aphrophilus
based on carbohydrates in lipopolysaccharide. J. Chromatogr.
310:261-272.
2. Brondz, I., and I. Olsen. 1985. Differentiation between major
species of the Actinobacillus-Haemophilus-Pasteurellagroup
by gas chromatography of trifluoroacetic acid anhydride derivatives from whole-cell methanolysates. J. Chromatogr. 342:
13-23.
3. Brondz, I., and I. Olsen. 1985. Sugar composition of lipopolysaccharide from Haemophilus paraphrophilus. J. Chromatogr.
345: 119-124.
4. Brondz, I., and I. Olsen. 1986. Review. Microbial chemotaxonomy. Chromatography, electrophoresis and relevant profiling
techniques. J. Chromatogr. 379:367411.
5 . Brondz, I., and I. Olsen. 1986. Review. Chemotaxonomy of
selected species of the Actinobacillus-Haemophilus-Pasteurella
group by means of gas chromatography, gas chromatographymass spectrometry and bioenzymatic methods. J. Chromatogr.
380: 1-17.
6. Brondz, I., and I. Olsen. 1989. Chemical differences in lipopolysaccharides from Actinobacillus (Haemophilus) actinomycetemcomitans and Haemophilus aphrophilus: clues to differences
in periodontopathogenic potential and taxonomic distinction.
Infect. Immun. 57:310&3109.
6a.Brondz, I., I. Olsen, and M. Sjostrom. 1990. Multivariate
analysis of quantitative chemical and enzymic characterization
data in classification of Actinobacillus, Haemophilus, and Pasteurella spp. J. Gen. Microbiol. 136507-513.
7. Geladi, P., and B. R. Kowalski. 1986. Partial least squares
regression: a tutorial. Anal. Chim. Acta. 1851-17.
8. Joliffe, I. T. 1986. Principal component analysis. SpringerVerlag, New York.
9. Kilian, M. 1976. A taxonomic study of the genus Haemophilus,
with the proposal of a new species. J, Gen. Microbiol. 93:9-62.
10. Kilian, M., and E. L. Biberstein. 1984. Genus 11. Haemophilus
Winslow, Broadhurst, Buchanan, Krumwiede, Rogers, and
Smith 1917, 561AL,p. 558-569. I n N. R. Krieg and J . G. Holt
(ed.), Bergey’s manual of systematic bacteriology, vol. 1. The
Williams & Wilkins Co., Baltimore.
11. King, E. O., and H. W. Tatum. 1962. Actinobacillus actinomycetemcomitans and Haemophilus aphrophilus. J . Infect. Dis.
111:85-94.
12. Lebbard, S., J. M. Cavaillon, M. Caroff, A. Ledur, H. Brade, R.
Sarfati, and N. Haeffner-Cavaillon. 1986. Molecular requirement
for interleukin 1 induction by lipopolysaccharide-stimulated
human monocytes: involvement of the heptosyl-2-keto-deoxyoctulosonate region. Eur. J. Immunol. 16:87-91.
13. Potts, T. V., J. J. Zambon, andR. J. Genco. 1985. Reassignment
of Actinobacillus actinomycetemcomitans to the genus Huemophilus as Haemophilus actinomycetemcomitans comb. nov.
Int. J. Syst. Bacteriol. 35337-341.
14. Sjostrom, M., and S. Wold. 1977. Chemometrics: theory and
application. Am. Chem. SOC.Symp. Ser. 52:243-282.
15. Tanner, A. C. R., R. A. Visconti, S. S. Socransky, and S. C. Holt.
1982. Classification and identification of Actinobacillus actinomycetemcomitans and Haemophilus aphrophilus by cluster
analysis and deoxyribonucleic acid hybridizations. J . Periodontal Res. 17585-596.
16. Wold, S. 1976. Pattern recognition by means of disjoint principal
components models. Pattern Recogn. 8:127-139.
17. Wold, S. 1978. Cross-validatory estimation of the number of
components in factor and principal components models. Technometrics 20:397405.
18. Wold, S., C. Albano, W. J. Dunn 111, K. Esbensen, S. Hellberg,
E. Johanson, and M. Sjostrom. 1983. Pattern recognition: finding
and using regularities in multivariate data, p. 147-188. In H.
Martens and H. Russwurm (ed.), Food research and data
analysis. Applied Sciences Publishers, New York.
19. Wold, S., K. Esbensen, and P. Geladi. 1987. Principal component analysis. Chemometr. Intellig. Lab. Syst. 2:37-52.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Fri, 16 Jun 2017 05:10:53