Glycobiology vol. 14 no. 6 pp. 511±519, 2004
DOI: 10.1093/glycob/cwh073
Advance Access publication on March 24, 2004
Carbohydrates act as receptors for the periodontitis-associated bacterium
Porphyromonas gingivalis: a study of bacterial binding to glycolipids
Ulrika Hellstrom1,2, Eva C. Hallberg2,3, Jens Sandros2,
Lennart Rydberg3, and Annika E. BaÈcker4
2
Department of Oral Pathology, Faculty of Odontology, The Sahlgrenska
Academy at G
oteborg University, Box 450, SE 405 30 G
oteborg, Sweden;
3
Department of Clinical Chemistry and Transfusion Medicine,
The Sahlgrenska Academy at G
oteborg University, SE 413 45 G
oteborg,
Sweden; and 4Department of Endodontics and Oral Diagnostics,
Faculty of Odontology, The Sahlgrenska Academy at G
oteborg
University, Box 450, SE 405 30 G
oteborg, Sweden
Received on October 30, 2003; revised on February 19, 2004; accepted on
March 4, 2004
In this study we show for the first time the use of carbohydrate
chains on glycolipids as receptors for the periodontitisassociated bacterium Porphyromonas gingivalis. Previous
studies have shown that this bacterium has the ability to adhere
to and invade the epithelial lining of the dental pocket. Which
receptor(s) the adhesin of P. gingivalis exploit in the adhesion
to epithelial cells has not been shown. Therefore, the binding
preferences of this specific bacterium to structures of carbohydrate origin from more than 120 different acid and nonacid
glycolipid fractions were studied. The bacteria were labeled
externally with 35S and used in a chromatogram binding assay.
To enable detection of carbohydrate receptor structures for
P. gingivalis, the bacterium was exposed to a large number of
purified total glycolipid fractions from a variety of organs
from different species and different histo-blood groups.
P. gingivalis showed a preference for fractions of human and
pig origin for adhesion. Both nonacid and acid glycolipids were
used by the bacterium, and a preference for shorter sugar
chains was noticed. Bacterial binding to human acid glycolipid
fractions was mainly obtained in the region of the chromatograms where sulfated carbohydrate chains usually are found.
However, the binding pattern to nonacid glycolipid fractions
suggests a core chain of lactose bound to the ceramide part as a
tentative receptor structure. The carbohydrate binding of the
bacterium might act as a first step in the bacterial invasion
process of the dental pocket epithelium, subsequently leading
to damage to periodontal tissue and tooth loss.
Key words: adhesion/bacterial overlay/chromatogram
binding assay/glycolipid/Porphyromonas gingivalis
Introduction
Over the years more than 400 different bacterial species
have been isolated from the oral cavity. Only a small
1
To whom correspondence should be addressed; e-mail:
[email protected]
number of these strains are considered to be periodontally
harmful isolates (Moore and Moore, 1994). Previous studies
have shown that the oral pathogen Porphyromonas gingivalis
plays a significant role in the development of periodontitis,
a disease that affects the supporting structure of the tooth
and subsequently leads to tooth loss if left untreated
(Haffajee and Socransky, 1994). A large array of potential
virulence factors produced by the bacterium, including
(among others) production of proteases, fimbriae, polysaccharide capsule, and hemagglutinating factors, enables the
induction of periodontal tissue damage (Holt and Bramanti,
1991). The adhesive and invasive potential of P. gingivalis
that provides the bacterium with a temporary shelter
against antibodies and phagocytic cells of the host has
been carefully examined in previous studies. However, few
of the interactions between this microorganism and the host
that lead to periodontal tissue destruction are known. The
bacterium uses an adhesin located on the tip of its fimbriae
to attach to the surface of the epithelial cells. This binding is
inhibited if the fimbriae are blocked by monoclonal antibodies (Isogai et al., 1988). It is also known that P. gingivalis
strains, which express only a minor arsenal of fimbriae or
no fimbriae at all, show a decreased binding capacity to
epithelial cells (Lee et al., 1992).
It has been shown that a number of bacteria can bind to
glycoconjugates on the cell surfaces (Kelm and Schauer,
1997; Mouricout, 1997), for example, Streptococcus suis
(Haataja et al., 1994), Escherichia coli (Yang et al., 1994),
and Helicobacter pylori (Boren et al., 1993). Which receptor(s) the adhesin of P. gingivalis exploit in the adhesion to
epithelial cells has not been shown. Theories and studies
based on various proteins and glycosylated proteins have
been published, but none have yet been able to characterize
the structure of the bacterial binding epitope. In vitro studies with glycoprotein from oral epithelium suggest the use
of glycoconjugates as bacterial receptors (Agnani et al.,
2003). Adhesion to glycoproteins can occur either to the
protein or the carbohydrate part of the molecule. A recently
published study showed that the addition of soluble saccharides to the bacterial solution can abolish the binding of
P. gingivalis to oral epithelial cells (Agnani et al., 2003).
This fact suggests that the carbohydrate moiety is somehow
involved in the adhesion process. Previous studies have
also shown that P. gingivalis can hydrolyze proteins and
thereby expose previously hidden epitopes (i.e., cryptitopes)
(Kontani et al., 1997). This procedure might be adopted by
the bacterium when adhering to carbohydrate epitopes on
glycolipids.
Glycoproteins exist in different forms in the membranes
of eukaryotic cells. A characteristic feature of glycoproteins
is the dominating protein moiety of the molecule, compared
Glycobiology vol. 14 no. 6 # Oxford University Press 2004; all rights reserved.
511
U. Hellstrom et al.
to proteoglycans, which consist of a dominating carbohydrate part. Carbohydrates existing on glycoproteins can
often also be found on glycolipids on the cell surface. The
carbohydrate epitopes on glycolipids and glycoproteins are
in many cases identical, but the core chains do vary (Oriol,
1995). When compared to glycoproteins, the core chains on
glycolipids are generally shorter and less branched. They
are also quite distinct from the core sequences of glycoproteins. The expression of carbohydrate chains can also vary
between different animal species (BaÈcker et al., 1997; Bj
ork
et al., 1987; Smith et al., 1975; Umesaki et al., 1989),
between different organs (BaÈcker et al., 1999; Ulfvin et al.,
1989), and also between different cells (Strokan et al., 1998a).
Based on these facts, the aim of the present study was to
investigate whether the periodontal pathogen P. gingivalis
can use glycolipids as receptors by screening a large array of
glycolipid structures with a chromatogram binding assay.
(A)
(B)
1
1
2
2
3
3
Results
More than 120 glycolipid fractions representing a wide range
of different carbohydrate structures were separated by thinlayer chromatography (TLC). The TLC plate was used
for bacterial overlay assay with 35S-labeled P. gingivalis
FDC381, and an identical TLC plate was stained with
anisaldehyde reagent for chemical detection of the glycolipids. Binding was visualized by autoradiography. In
general, more binding sites were detected in the autoradiograms after exposure of P. gingivalis to nonacid (one to four
binding sites) compared to exposure to acid glycolipid fractions (one to two binding sites) (Figures 1±6). When the
autoradiograms where applied on the anisaldehyde-stained
TLC plates with nonacid glycolipid fractions, binding of the
bacteria was mostly to medium moving fractions in the
chromatogram. This region of the TLC plate contains
glycolipids with three to five saccharides in their carbohydrate chain. In acid fractions bacterial binding mostly
occurred to fast moving fractions which is the part of the
chromatogram that contains glycolipids with one to three
saccharides in their carbohydrate chains.
Binding of P. gingivalis to nonacid glycolipid fractions of
human origin was observed in the region of the TLC plates
with one to four sugar residues (Table I, Figure 1). Considering human acid glycolipid fractions, bacterial binding
was mainly obtained in the short sugar residue region of the
TLC plates, a region where sulfated carbohydrate chains
usually are found (Table I, Figures 2 and 3). No binding was
detected to human acid fractions originating from blood
group A plasma, colon, liver, small intestine, and blood
group A1 transplantation liver (data not shown). The total
acid fraction from ureter showed a weak staining on the
autoradiogram, but binding was not always obtained to this
fraction (Figure 2, lane 4). Neither could positive binding be
confirmed with purified fractions of ganglioside or sulfatide
from this organ (data not shown).
The binding pattern of nonacid glycolipid fractions from
pig organs was generally more distinct compared to autoradiograms of bacterial overlay assay with glycolipid fractions from human organs (Table II, Figures 4 and 5).
Binding to acid glycolipid fractions of pig origin could not
be detected (Figure 6).
512
Lane
1
2
3
4
5
1
2
3
4
5
Fig. 1. Thin-layer chromatogram of nonacid glycolipid fractions of
human origin with positive binding of 35S-labeled P. gingivalis. (A)
Detection of glycolipids with anisaldehyde reagent. (B) Glycolipids
detected by autoradiography after binding of 35S-labeled P. gingivalis.
The binding assay was performed as described in Materials and methods.
Approximately 2±10 mg of the glycolipid mixtures were added to the lanes
(1±5), depending on their complexity. The solvent system used was
chloroform/methanol/water (60:35:8, by volume). The lanes contained
kidney A (lane 1, reference), transplantation kidney A1 Le(aÿb‡) (lane 2),
vein A1 Le(aÿb‡) (lane 3), ureter A1 Le(aÿb‡) (lane 4), and artery
A1 Le(aÿb‡) (lane 5). The sugar residues in the glycolipid chains are
indicated on the left side.
Organs of sheep origin were also used in the bacterial
overlay assay. Total acid glycolipid fractions of blood
group A (Strokan et al., 1998b) were screened. All fractions
tested negative for binding with radiolabeled bacteria (data
not shown).
Binding was also obtained to gangliotriaosylceramide
and gangliotetraosylceramide, the pure reference glycolipid
fractions used for studying binding analogy with other
bacteria (Figure 7).
Discussion
Periodontitis is known to be a multifactorial disease,
involving a variety of bacterial species. The complexes of
microorganisms involved in the process appear to relate to
the severity of the periodontal destruction (Socransky et al.,
1988). In the present study we focus on P. gingivalis strain
FDC381 because previously published results have shown
that this particular strain has the ability to adhere to and
invade oral epithelial cells and human pocket epithelium
in vitro (Sandros et al., 1993, 1994). P. gingivalis seems to act
as a key bacteria found in a majority of the microbial
complexes causing severe and aggressive periodontitis with
a resulting loss of periodontal attachment. However, it is
Novel glycolipid binding of P. gingivalis
(A)
(A)
(B)
1
2
1
2
1
2
Lane
Lane 1
2
3
4
5
6
7
8
9
10
11
(B)
1
2
1
2
3
4
5
6
7
8
9
10
11
Fig. 2. Thin-layer chromatogram of acid glycolipid fractions of human
origin with positive binding of 35S-labeled P. gingivalis. (A) Detection of
glycolipids with anisaldehyde reagent. (B) Glycolipids detected by
autoradiography after binding of 35S-labeled P. gingivalis. The binding
assay was performed as described in Materials and methods.
Approximately 2±10 mg of the glycolipid mixtures were added to the lanes
(1±11), depending on their complexity. The solvent system used was
chloroform/methanol/water with 0.2% CaCl2 (60:40:9, by volume). The
lanes contained kidney A (lane 1, reference), intestine B (lane 2), kidney B
(lane 3), ureter A1 Le(aÿb‡) (lane 4), sulfolipid fraction from kidney B
(lane 5), sulfatide fraction from pancreas (lane 6), kidney (lane 7),
transplantation kidney (lane 8), transplantation kidney (lane 9),
transplantation kidney (lane 10), and sulfatide fraction from kidney
(lane 11). The sugar residues in the glycolipid chains are indicated on
the left side.
1
2
3
4
5
1
2
3
4
5
Fig. 3. Thin-layer chromatogram of acid glycolipid fractions of human
origin with positive binding of 35S-labeled P. gingivalis. (A) Detection of
glycolipids with anisaldehyde reagent. (B) Glycolipids detected by
autoradiography after binding of 35S-labeled P. gingivalis. The binding
assay was performed as described in Materials and methods.
Approximately 2±10 mg of the glycolipid mixtures were added to the lanes
(1±5), depending on their complexity. The solvent system used was
chloroform/methanol/water with 0.2% CaCl2 (60:40:9, by volume). The
lanes contained kidney A (lane 1, reference), kidney A (lane 2),
ganglioside fraction from kidney (lane 3), ganglioside fraction from
transplantation kidney (lane 4), and sulfated sulfolipid fraction from
kidney (lane 5). The sugar residues in the glycolipid chains are indicated
on the left side.
important to keep in mind the complexity of this disease
and the cooperation of the periodontally harmful bacterial
species existing in the dental crevice. The binding epitopes
used by P. gingivalis for attachment to and invasion of oral
epithelium are not fully understood.
The fact that bacteria use short carbohydrate sequences
as receptors is well known (Karlsson, 1995) and has been
studied at different locations, such as in the human intestine
(Boren et al., 1993; Holgersson et al., 1991a) and oral cavity
(Str
omberg and Boren, 1992). The blood group of the individual (e.g., ABO and Lewis blood group systems) is of
importance. In fact, in some cases the blood group has
been shown to be a crucial factor for the bacterial capacity
to adhere to the specific tissue. This is exemplified by the
bacterium H. pylori, which binds to stomach mucosa in
individuals expressing the Leb structure (a five-sugar-long
carbohydrate) in the tissue (Boren et al., 1993). Both neutral
(nonacid) and acid carbohydrate chains are known to act as
Ê ngstr
bacterial receptors (A
om et al., 1998; Boren et al.,
1993; Holgersson et al., 1991a; Karlsson, 1995; MillerPodraza et al., 1996; Teneberg et al., 2002, 2004). It is not
known whether the binding of P. gingivalis to carbohydrates is dependent on the type of blood group or not.
We found that it was adequate to cultivate the bacteria
for 3 days prior to the bacterial overlay assays. At this point
P. gingivalis seems to express properties necessary for
513
U. Hellstrom et al.
(A)
Lane
(A)
(B)
1
2
2
3
3
4
4
1
2
3
4
1
2
3
1
4
Fig. 4. Thin-layer chromatogram of nonacid glycolipid fractions of pig
origin with positive binding of 35S-labeled P. gingivalis. (A) Detection of
glycolipids with anisaldehyde reagent. (B) Glycolipids detected by
autoradiography after binding of 35S-labeled P. gingivalis. The binding
assay was performed as described in Materials and methods.
Approximately 2±10 mg of the glycolipid mixtures were added to the
lanes (1±4), depending on their complexity. The solvent system used was
chloroform/methanol/water (60:35:8, by volume). The lanes contained
human acid kidney A (lane 1, reference), pig liver O (lane 2), pig kidney O
(lane 3), and pig small intestine O (lane 4). The sugar residues in the
glycolipid chains are indicated on the left side.
adhesion to glycolipid structures. The bacteria in the study
were cultured on agar plates. Other glycolipid binding bacteria, like H. pylori, show the same binding pattern when
cultured on agar as in broth (Teneberg et al., 2002). It is also
known that the use of different coating solutions in the
chromatogram binding assay can affect the reproducibility
of binding (Teneberg et al., 2002). The use of 2% bovine
serum albumin (w/v) in phosphate buffered saline (PBS)
with 0.1% Tween 20 (v/v) and 0.01% sodium azid (w/v)
was found to be useful in our bacterial overlays.
The bacterial overlay assay showed binding of
P. gingivalis to human nonacid and acid fractions and to
pig nonacid fractions. Acid fractions of pig origin tested
negative despite the close similarity of glycolipid expression
in human and pig organs. Binding was found in the foursugar region but not to longer sugar chains in the glycolipid
fractions. The bacterial binding sites are presumably not
located on the terminal sugars of the glycolipid carbohydrate chain but on the core chain. This is an important
observation because previous studies have shown that
P. gingivalis has the ability to enzymatically cleave proteins
(Kontani et al., 1997) and thereby reveal cryptitopes of
protein origin. Enzymatic cleavage might also be used by
the bacterium on structures of carbohydrate origin, thus
enabling adhesion of the bacterium to cell surfaces. If we
expose P. gingivalis solely to carbohydrate epitopes that
normally are expressed only in oral mucosa, there would
in fact be a reduced chance of finding the actual binding
514
(B)
Lane
1
2
3
4
1
2
3
4
Fig. 5. Thin-layer chromatogram of nonacid glycolipid fractions of pig
origin with positive binding of 35S-labeled P. gingivalis. (A) Detection of
glycolipids with anisaldehyde reagent. (B) Glycolipids detected by
autoradiography after binding of 35S-labeled P. gingivalis. The binding
assay was performed as described in Materials and methods.
Approximately 2±10 mg of the glycolipid mixtures were added to the lanes
(1±4), depending on their complexity. The solvent system used was
chloroform/methanol/water (60:35:8, by volume). The lanes contained
nonacid glycolipid fraction from human acid kidney A (lane 1,
reference), pig salivary gland O (lane 2), pig heart O (lane 3), and pig
spleen O (lane 4). The sugar residues in the glycolipid chains are indicated
on the left side.
epitopes used by the bacterium in vivo. The glycolipid
expression in oral epithelium varies between different individuals because of, among other things, genetic conditions
such as the type of blood group (Ravn and Dabelsteen,
2000). Therefore, it is of great importance to use a variety of
species (in this case human, pig, and sheep) to find structures
of carbohydrate origin that are normally not exposed in
human tissue. These structures can be exposed as cryptitopes, for example, as previously mentioned. Consideration
must also be taken of the fact that it is difficult to purify
glycolipids from human pocket epithelium because it is very
hard to gain access to the large amount of tissue necessary
for glycolipid purification. Studies on glycolipid expression
in gingival tissue only consists of assays with monoclonal antibodies on surfaces of gingival epithelial cells (Mackenzie
et al., 1989). The glycolipids constitute only a small part of
the total weight of the tissue which results in approximately
only 20 mg total glycolipid fraction from 200 g tissue.
Another important reason for using glycolipids from different species and different organs is that information can be
determined about the possible influence and/or involvement
on the binding of the bacterium to the ceramide part of the
glycolipid chain. Studies with H. pylori have shown that the
ceramide part is involved in the binding of this bacterium to
Ê ngstr
glycolipids (A
om et al., 1998).
Novel glycolipid binding of P. gingivalis
(A)
Table I. Glycolipid fractions of human origin with positive binding of
35
S-labeled P. gingivalis in the bacterial overlay assay
(B)
Binding
Sugar
intensitya regionb References
Nonacid glycolipid fractions of human origin
Blood group A fractions
1
2
1
2
Transplantation
kidney A1
3
Holgersson et al., 1991b
Kidney A1 Le(a‡bÿ)
3
Holgersson et al., 1991b
Kidney A1 Le(aÿb‡)
1, 2, 3
Holgersson et al., 1990a
Ureter A1 Le
1, 2
Holgersson et al., 1990a
Artery A1 Le(aÿb‡)
2
Holgersson et al., 1990a
3
Holgersson et al., 1991b
(aÿb‡)
Blood group B fractions
Kidney
Acid glycolipid fractions of human origin
Blood group A fractions
Lane
1
2
3
4
5
1
2
3
4
5
Fig. 6. Thin-layer chromatogram of acid glycolipid fractions of pig
origin with negative binding of 35S-labeled P. gingivalis. (A) Detection of
glycolipids with anisaldehyde reagent. (B) Glycolipids detected by
autoradiography after binding of 35S-labeled P. gingivalis. The binding
assay was performed as described in Materials and methods.
Approximately 2±10 mg of the glycolipid mixtures were added to the lanes
(1±5), depending on their complexity. The solvent system used was
chloroform/methanol/water with 0.2% CaCl2 (60:40:9, by volume). The
lanes contained acid glycolipid fractions from kidney A (lane 1,
reference), ganglioside fraction from pig heart (lane 2), sulfatide fraction
from pig heart (lane 3), ganglioside fraction from pig kidney (lane 4), and
sulfatide fraction from pig kidney (lane 5). The sugar residues in the
glycolipid chains are indicated on the left side.
The blood group type of the organ that the glycolipid
fraction originated from did not seem to be significant.
The blood group epitopes are in most cases terminally
expressed on precursor chains with four to five sugar residues. One can speculate whether the presence of a blood
group epitope might inhibit bacterial binding to the carbohydrate. Instead, binding might occur to the core chain or
the intermediate part of the carbohydrate chain depending
on the 3D structure of the carbohydrate chain. A core chain
similarity is indicated in the bacterial binding region in the
non-acid glycolipid fractions from pig. In this region of the
TLC plate the globo chain and its precursors are often
found (BaÈcker et al., 1998).
Binding of P. gingivalis to acid glycolipid fractions from
organs of human origin was noticed in the upper part of the
chromatogram. In this region of the TLC plates sulfatide
fractions are usually found, which indicates that this may be
a tentative structure for adhesion of this specific bacterium.
Sulfatide fractions have also been included in our screening
of glycolipid structures and positive binding was obtained
(Figure 2, lane 5, 6 and 11; and Figure 3, lane 5), a result
that supports our hypothesis. This has recently also been
indicated in the case of glycoprotein binding of P. gingivalis
(Agnani et al., 2003). In sulfatide glycolipid fractions the
sugar chain shows a one- or two-sugar residue length with
Kidney
1, 2
Holgersson et al., 1991b
Kidney
1, 2
Holgersson et al., 1991b
Ureter A1 Le(aÿb‡)
3
Holgersson et al., 1990a
Blood group B fractions
Intestine
1, 2
Holgersson et al., 1991a
Kidney
1, 2
Holgersson et al., 1991b
Sulfolipid fraction
kidney
1, 2
Holgersson et al., 1991b
1, 2
Breimer, 1983
Nonspecified blood group
Sulfatide fraction
pancreas
Kidney
1, 2
Holgersson et al., 1991b
Transplantation kidney
1, 2
Holgersson et al., 1991b
Transplantation kidney
1, 2
Holgersson et al., 1991b
Transplantation kidney
1, 2
Holgersson et al., 1991b
Sulfatide fraction kidney
1
Holgersson et al., 1991b
a
Asterisks mark significant darkening observed by visual inspection of
the autoradiogram. One asterisk indicates weaker staining and four
asterisks indicate very strong staining of the b-sensitive film.
b
Estimated number of sugar residues in the bacterial binding region of
the glycolipid chain when comparing the anisaldehyde stained TLC
plates to previously published structural characterization studies of the
respective organ.
terminal sulfate group (SOÿ
3 ). The core structure tentatively
consists of a galactose connected to glucose before connecting to the polar ceramide chain.
Sugar chains consisting of one-, two-, three-, and foursugar residues are used by P. gingivalis as receptor structures in nonacid fractions from pig organs. A comparison of
the binding areas in this study, with earlier published structural characterization studies of the pig organs (BaÈcker
et al., 1998), indicated that the bacterial glycolipid receptor
structure might be a so-called type 4 chain, that is, the globo
chain and its core structures (Holgersson et al., 1992).
Considering the binding in the two-sugar region, it is not
possible to distinguish between binding to lactosylceramide
515
U. Hellstrom et al.
Table II. Glycolipid fractions of pig origin with positive binding of
35
S-labeled P.gingivalis in the bacterial overlay assay
Binding
intensitya
Sugar
regionb
(A)
(B)
References
Nonacid glycolipid fractions of pig origin
Blood group A fractions
Salivary gland
1
BaÈcker et al., 1998
Blood group O fractions
Spleen
1, 3
Holgersson et al., 1990b
Liver
1, 3
BaÈcker et al., 1998
Kidney
1, 2, 3
BaÈcker et al., 1998
Small intestine
2, 3, 4
BaÈcker et al., 1997
a
Asterisks mark significant darkening observed by visual inspection of
the autoradiogram. One asterisk indicates weaker staining and four
asterisks indicate very strong staining of the b-sensitive film.
b
Estimated number of sugar residues in the bacterial binding region of
the glycolipid chain when comparing the anisaldehyde-stained
TLC plates to previously published structural characterization studies
of the respective organ.
and/or to galabiosylceramide, because these glycolipids do
not separate on the TLC plates due to similar polarities.
Our results also showed that P. gingivalis has the ability to
adhere to gangliotriaosylceramide (GgO3) and gangliotetraosylceramide(GgO4).Thesepurifiedreferenceglycolipid
fractions were used to study analogy of binding patterns to
other bacteria. Previous studies has shown that, among
others, Actinobacillus pleuropneumoniae (Abul-Milh et al.,
1999) and E. coli (Teneberg et al., 2004) can bind to GgO3
and GgO4. The fact that P. gingivalis also can use these
glycolipids as receptor structures shows that this bacterium
has a similar binding pattern. What type of other pure
glycolipid structures P. gingivalis recognizes will be fully
described in our following study.
Our screening of a large library of glycolipid carbohydrate chains with TLC and bacterial overlay assay showed
that P. gingivalis can bind to sugar residues on glycolipids.
A screening of more than 120 different fractions by TLC
where each fraction was tested at least three times is very
time-consuming. Therefore, this screening will be followed
in a forthcoming study by further purification of the glycolipid fractions with positive binding of the bacterium.
In this screening study we show for the first time that the
periodontal pathogen P. gingivalis strain FDC381 can use
glycolipid sugar chains as bacterial receptors. Both nonacid
and acid glycolipids can be used by the bacterium and
shorter carbohydrate chains are mainly preferred. We have
also found that sulfated glycolipids probably are favored by
the bacterium in acid glycolipid fractions. Glycolipids based
on the globo chain with a core chain of lactosylceramide
bound to the ceramide part, seem to be preferred in the
nonacid glycolipid fractions.
The carbohydrate binding of the bacterium might act as a
first step in the bacterial invasion process of the pocket
epithelium, subsequently leading to periodontal damage in
516
Lane
1
2
1
2
Fig. 7. Thin-layer chromatogram of pure reference glycolipid fractions
with positive binding of 35S-labeled P. gingivalis. (A) Detection of
glycolipids with anisaldehyde reagent. (B) Glycolipids detected by
autoradiography after binding of 35S-labeled P. gingivalis. The binding
assay was performed as described in Materials and methods. 8 mg of the
glycolipid mixtures were added to the lanes (1±2). The solvent system
used was chloroform/methanol/water (60:35:8, by volume). The lanes
contained gangliotriaosylceramide from guinea pig erythrocytes (lane 1),
and gangliotetraosylceramide from mouse feces (lane 2). These reference
glycolipid fractions were kindly provided by M.D., Ph.D. Teneberg; for
a more detailed description, we refer to Teneberg et al. (2004).
the tissue. A mapping of the receptor structures that
P. gingivalis uses for adhesion and invasion of eukaryotic
cells would increase the knowledge of the origin of periodontitis. This knowledge might in turn be used in new treatment strategies, for example receptor analogs, to inhibit
bacterial binding and the production of vaccines and
other antimicrobial therapies specifically directed against
P. gingivalis. The future goal is to reduce the incidence,
onset, and spread of periodontal disease.
Materials and methods
Glycolipid preparations
Non-acid glycolipids from different species, organs, and
blood groups were isolated according to the previously
reported method of Karlsson (Karlsson, 1987). These nonacid glycolipid fractions are referred to as total fractions.
Fractions with acid properties were separated from the
nonacid fractions during the purification procedure and
used separately in the screening process. These fractions
are referred to as total acid fractions. In some cases the
glycolipid fractions have been further purified (e.g., acid
fractions from human kidneys, nonacid fractions originating from pig small intestine, salivary gland, kidney, liver,
and spleen) by high-performance liquid chromatography
(HPLC) (BaÈcker et al., 1998). The samples have then been
pooled and are referred to as, for example, fractions with
Novel glycolipid binding of P. gingivalis
one, two, or three sugar residues. Some of the longer
carbohydrate chain fractions have also undergone structural analysis by masspectrometry and/or nuclear magnetic
resonance for the identification of the component(s)
(BaÈcker et al., 1997, 1998; Gustavsson et al., 1996).
Sample criteria
The aim of the experiments was to perform a screening of a
wide range of glycolipid structures with TLC and bacterial
overlay assay. In this way P. gingivalis is exposed to a large
number of different carbohydrate structures, which enables
detection of receptor structures that are used by the bacterium. Different types of organs with different blood groups
were analyzed and compared for differences in binding
patterns and/or to study special bacterial preferences for
any particular blood group epitope. Several fractions from
the same kind of organ, but from different organ donors,
were used in the experiments as well as organs with different
blood group subgroups. Glycolipid fractions of both acid
and nonacid origin were used to see if, for example, sialic
acid is necessary for adhesion. In the majority of the screening experiments, total glycolipid fractions were used.
Fractions used in TLC and bacterial overlay
The following section is a listing of all fractions that were
used in the TLC and bacterial overlay assays. Any subgrouping, such as serological typing (e.g., A2), blood
group Lewis (Le(ab)), and Secretor (Se) status, is shown in
parentheses. Fractions named transplantation indicate
organs that originate from a human transplantation donor.
Fractions of human origin. Total fractions originating
from human organs of blood group A used for screening
included acid and nonacid kidney (A2), nonacid
transplantation kidney (A1), nonacid transplantation
kidney (A1 Le(a‡bÿ)), nonacid transplantation kidney (A2
Le(a‡bÿ)), nonacid liver (A1B), nonacid liver (A1 Le(aÿb‡)),
nonacid liver (A1B Le(aÿbÿ)), nonacid liver (A1 Le(a‡bÿ)),
nonacid liver (A2B Le(aÿb‡)), acid transplantation liver
(A1), nonacid jejunum, nonacid ileum (A1), nonacid pancreas, acid plasma (A1 Le(aÿbÿ)Se), acid plasma (A2
Le(aÿb‡)), nonacid thrombocytes (A1), and nonacid thrombocytes (A2).
Total fractions originating from human organs of blood
group B type used for screening included acid and nonacid
kidney, nonacid transplantation kidney (Le(aÿb‡)), acid
intestine, acid and nonacid ureter (Le(aÿb‡)), acid and nonacid artery (Le(aÿb‡)), acid and nonacid vein (Le(aÿb‡)),
nonacid liver, nonacid liver (Le(aÿb‡)), acid erythrocytes
(Le(aÿb‡)), and acid plasma (A1 Le(aÿb‡)).
Total fractions originating from human organs of
blood group AB type used for screening included nonacid
liver (A1B) and nonacid transplantation kidney (AB
Le(aÿb‡)).
Total fractions originating from human organs of blood
group O type used for screening included nonacid kidney,
nonacid transplantation kidney (Le(aÿb‡)), nonacid erythrocytes (Le(aÿb‡)), and nonacid plasma (Le(aÿb‡)).
Total fractions originating from human organs with
pooled and/or mixed nonspecified blood group included
acid kidney, acid and nonacid human liver, acid ureter,
acid transplantation kidney, acid and nonacid colon, and acid
small intestine.
HPLC-purified human fractions used in the screening
included ganglioside fraction from colon, brain ganglioside
fraction, pancreas ganglioside fraction, pancreas sulfatide
fraction, ureter ganglioside fraction, transplantation kidney
ganglioside fraction, sulfolipid fraction from ureter (residual
stroma) blood group B, sulfolipid fraction from kidney
blood group B, kidney ganglioside fraction, sulfolipid fraction from kidney, and a pure globoside fraction.
Fractions of nonhuman origin. Total fractions originating
from pig organs of blood group O used for screening
included nonacid small intestine, nonacid kidney, nonacid
liver, nonacid spleen, and acid erythrocytes.
Total fractions originating from pig organs of blood
group A included nonacid salivary gland and nonacid heart.
A total fraction originating from about 80 pig blood group
A and O aortas was also used.
HPLC-purified pig organ fractions used included ganglioside, globotriaosylceramide and globotetraosylceramide
fractions from heart; sulfolipid fractions from heart, salivary gland, kidney, small intestine, liver, and spleen; and
ganglioside fractions from intestine, kidney, salivary gland,
heart, and liver. Fractions from small intestine, kidney, liver,
salivary gland, spleen, and heart have also been purified and
pooled into fractions with nonspecified monosaccharides,
disaccharides, trisaccharides, tetrasaccharides, and fractions with tetrasaccharide glycolipids (e.g., four to eight
sugars in the sugar chain), respectively.
Total fractions originating from sheep organs of blood
group A type included acid pancreas, acid small intestine,
acid colon, acid lung, and acid kidney. Purified reference
glycolipid fractions included gangliotriaosylceramide from
guinea pig erythrocytes and gangliotetraosylceramide from
mouse feces.
Thin Layer Chromatography
TLC was performed as previously described (Hansson et al.,
1985) on aluminum-backed silica gel 60 high-performance
TLC plates (Merck, Darmstadt, Germany and HP-KF,
Whatman, Maidstone, U.K.). Glycolipid fractions (2±10 mg/
lane) were added to the TLC plates and separated in a
specific mixture of organic solvents and water. The nonacid components were chromatographed in chloroform/
methanol/water, 60:35:8 by volume, and the acid components in chloroform/methanol/water with 0.2% CaCl2,
60:40:9 by volume. On all TLC plates, a reference fraction
from human kidney blood group A (an acid glycolipid
fraction) that was positive for bacterial binding was added
in lane 1.
For each chromatogram with bacterial overlay, two sets
of identical glycolipid fractions were separated on the same
TLC plate. In this manner we received two TLC plates with
as identical a separation pattern as possible. The TLC plate
was dried at room temperature until the organic solvents
had evaporated from the silica gel, and thereafter it was cut
in two halves, thus separating the identical TLC sets. The
TLC plate that was used in the bacterial overlay experiment
517
U. Hellstrom et al.
was coated in 0.3% (w/v) polyisobutylmethacrylate, P28
(Sigma-Aldrich), in diethyl ether/n-hexane (1:1 by volume),
or in 0.5% P28 in diethyl ether/n-hexane (1:4 by volume)
and left to polymerize overnight to avoid unspecific binding
to nonpolymerized monomers. The identical TLC plate was
visualized chemically by anisaldehyde reagent and used
for inspection of the separation pattern and for identification of the bacterial binding region in the autoradiogram
(Karlsson, 1987). All glycolipid-containing lines were visualized by chemical detection with anisaldehyde reagent. This
TLC plate was then compared with the identical plate used
in the chromatogram binding assay, where only lines with
positive binding can be seen.
Bacterial strains and growth conditions
In this study P. gingivalis FDC381 (collection of Forsyth
Dental Center, Boston, MA), a strain isolated from a periodontitis patient, was used. The bacteria were grown on
Brucella agar plates (BBL Microbiology Systems, Cockeysville, MD) enriched with 5% defibrinated horse blood, 0.5%
hemolyzed blood, and 5 ml/ml menadione in jars with 95%
H2 and 5% CO2 at 37 C. A suspension of 5 ml L-[35S]methionine and L-[35S]cysteine in PBS with a concentration
of 14.3 mCi/ml and t1/2 of 87.4 days (Redivue ProMix L[35S] in vitro labeling mix, Amersham Biosciences, Uppsala,
Sweden) was added to the agar plates. After 3 days of
growth, bacterial cultures were collected and washed three
times by centrifugation at approximately 5000 rpm in PBS.
A spectrophotometer analysis at 550 nm was performed
for evaluation of the bacterial concentration and compared
to an absorbance standard curve for P. gingivalis. The
bacteria were suspended in PBS with 2% bovine serum
albumin (w/v), 0.1% Tween 20 (v/v), and 0.01% sodium
azid (w/v) and diluted to a final concentration of 1.5±
3 108 bacteria/ml used in the bacterial overlay assays.
Bacterial overlay and detection of bacterial binding
The method used in the bacterial overlay assays is a modified version of the method by Karlsson and Str
omberg
(1987). The TLC plates were coated with 2% bovine serum
albumin (w/v), 0.1% Tween 20 (v/v), and 0.01% sodium azid
(w/v) in PBS for 2 h at room temperature to block nonspecific binding sites. Thereafter, the coating solution was carefully removed from the TLC plates. Approximately 5 ml of
the bacterial suspension was sprinkled over the TLC plates
with a Pasteur pipette and then incubated at room temperature for 9 h or overnight. The TLC plates were gently
washed four times with PBS, dried in room temperature
for at least 1 h, and autoradiographed for approximately
7±14 days using a b-sensitive film (Kodak BioMax MR,
Amersham Biosciences). After development of the film (on
Agfa Curix HT-530 U equipment), bacterial binding was analyzed by visual inspection. Significant staining/darkening
of the film was registered as positive binding and marked
with asterisks (one to four, depending on intensity of the
binding; one asterisk ˆ weak staining, four asterisks ˆ strong
staining). When positive binding was detected in the autoradiogram, the film was placed on top of the identical
anisaldehyde-stained TLC plate by means of easily identified landmarks. In this way the specific sugar binding region
518
was identified. A comparison to the TLC plate in the
reference article for that specific organ enabled estimation
of the glycolipid structure that P. gingivalis can use as a
receptor.
Acknowledgments
Stina Olsson is gratefully acknowledged for valuable
help with the bacteria. We also thank Lola Svensson
for technical assistance with the TLC plates and
M.D., Ph.D. Teneberg for discussions and troubleshooting
concerning the bacterial overlay method and for lending us
the pure reference glycolipid fractions. This work was supported by the Swedish Research Council (grant no. K200273X-14018-02B), a research grant from TUA-SAM, the
Magn. Bergwall Foundation, Wilhelm and Martina Lundgren's Scientific Fund, the Royal Society of Arts and
Sciences in G
oteborg, the Swedish Dental Society, and the
Dental Society of G
oteborg.
Abbreviations
HPLC,
high-performance
PBS, phosphate buffered
chromatography.
liquid
saline;
chromatography;
TLC, thin-layer
References
 ., Dubreuil, J.D., and Jacques, M. (1999)
Abul-Milh, M., Paradis, S.E
Binding of Actinobacillus pleuropneumoniae lipopolysaccarides to
glycosphingolipids evaluated by thin-layer chromatography. Infect.
Immun., 67, 4983±4987.
Agnani, G., Tricot-Doleux, S., Houalet, S., and Bonnaure-Mallet, M.
(2003) Epithelial cell surface sites involved in the polyvalent adherence
of Porphyromonas gingivalis: a convincing role for neuraminic acid and
glucuronic acid. Infect. Immun., 71, 991±996.
Ê ngstr
A
om, J., Teneberg, S., Milh, M.A., Larsson, T., Leonardsson, I.,

Olsson, B.M., Olwega
Ê rd Halvarsson, M., Danielsson, D., NaÈslund, I.,
and others. (1998) The lactosylceramide binding specificity of
Helicobacter pylori. Glycobiology, 8, 297±309.
BaÈcker, A.E., Breimer, M.E., Samuelsson, B.E., and Holgersson, J. (1997)
Biochemical and enzymatic characterization of blood group ABH and
related histo-blood group glycosphingolipids in the epithelial cells of
porcine small intestine. Glycobiology, 7, 943±953.
BaÈcker, A.E., Holgersson, J., Samuelsson, B.E., and Karlsson, H. (1998)
Rapid and sensitive GC/MS characterization of glycolipid released
Gala1,3Gal-terminated oligosaccharides from small organ specimens
of a single pig. Glycobiology, 8, 533±545.
BaÈcker, A.E., Thorbert, S., Rakotonirainy, O., Hallberg, E.C., Olling, A.,
Gustavsson, M., Samuelsson, B.E., and Soussi, B. (1999) Liquid
chromatography ``on-flow'' 1H nuclear magnetic resonance on native
glycosphingolipid mixtures together with gas chromatography/mass
spectrometry on the released oligosaccharides for screening and
characterization of carbohydrate-based antigens from pig lungs.
Glycoconj. J., 16, 45±58.
Bj
ork, S., Breimer, M.E., Hansson, G.C., Karlsson, K.-A., and Leffler, H.
(1987) Structures of blood group glycosphingolipids of human small
intestine. A relation between the expression of fucolipids of epithelial
cells and the ABO, Le and Se phenotype of the donor. J. Biol. Chem.,
262, 6758±6765.
Boren, T., Falk, P., Roth, K.A., Larson, G., and Normark, S. (1993)
Attachment of Helicobacter pylori to human gastric epithelium
mediated by blood group antigens. Science, 262, 1892±1895.
Breimer, M.E. (1983) Tissue specificity of glycosphingolipids as expressed
in pancreas and small intestine of blood group A and B human
individuals. Arch. Biochem. Biophys., 228, 71±85.
Novel glycolipid binding of P. gingivalis
Gustavsson, M.L., Gannedahl, G., BaÈcker, A.E., Larson, G., Olling, A.,
Tufveson, G., and Samuelsson, B.E. (1996) Anti-carbohydrate
antibodies associated with hyperacute rejection in a vascularized
mouse heart-to-rat xenotransplantation model. Transplantation, 61,
957±963.
Haataja, S., Tikkanen, K., Nilsson, U., Magnusson, G., Karlsson, K.-A.,
and Finne, J. (1994) Oligosaccharide-receptor interaction of the
Gala1-4Gal binding adhesin of Streptococcus suis. Combining site
architecture and characterization of two variant adhesin specificities.
J. Biol. Chem., 269, 27466±27472.
Haffajee, A.D. and Socransky, S.S. (1994) Microbial etiological agents of
destructive periodontal diseases. Periodontol. 2000, 5, 78±111.
Hansson, G.C., Karlsson, K.-A., Larsson, G., Str
omberg, N., and
Thurin, J. (1985) Carbohydrate-specific adhesion of bacteria to thinlayer chromatograms: a rational approach to the study of host cell
glycolipid receptors. Anal Biochem., 146, 158±163.
Holgersson, J., Clausen, H., Hakomori, S.I., Samuelsson, B.E.,
and Breimer, M.E. (1990a) Blood group A glycolipid antigen expression in kidney, urethra, kidney artery, and kidney vein from a blood
group A1Le(aÿb‡) human individual. J. Biol. Chem., 265, 20790±20798.
Ê ., Samuelsson, B.E., and Breimer, M.E.
Holgersson, J., Jovall, P.-A
(1990b) Structural characterization of non-acid glycosphingolipids in
kidneys of single blood group O and A pigs. J. Biochem., 108, 766±777.
Ê ., and Breimer, M.E. (1991a) GlycosphingHolgersson, J., Jovall, P.-A
olipids of human large intestine: detailed structural characterization
with special reference to blood group compounds and bacterial
receptor structures. J. Biochem. (Tokyo), 110, 120±131.
Ê ., Samuelsson, B.E., and Breimer, M.E. (1991b)
Holgersson, J., Jovall, P.-A
Blood group type glycosphingolipids of human kidneys. Structural
characterization of extended globo-series compounds. Glycoconj. J., 8,
424±433.
Holgersson, J., Breimer, M.E., and Samuelsson, B.E. (1992) Basic
biochemistry of cell surface carbohydrates and aspects of the tissue
distribution of histo-blood group ABH and related glycosphingolipids.
APMIS Suppl., 27, 18±27.
Holt, S.C. and Bramanti, T.E. (1991) Factors in virulence expression and
their role in periodontal disease pathogenesis. Crit. Rev. Oral Biol.
Med., 2, 177±281.
Isogai, H., Isogai, E., Yoshimura, F., Suzuki, T., Kagota, W., and
Takano, K. (1988) Specific inhibition of adherence of an oral strain of
Bacteroides gingivalis 381 to epithelial cells by monoclonal antibodies
against the bacterial fimbriae. Arch. Oral Biol., 33, 479±485.
Karlsson, K.-A. (1987) Preparation of total nonacid glycolipids for
overlay analysis of receptors for bacteria and viruses and for other
studies. Methods Enzymol., 138, 212±220.
Karlsson, K.-A. (1995) Microbial recognition of target-cell glycoconjugates. Curr. Opin. Struct. Biol., 5, 622±635.
Karlsson, K.-A. and Str
omberg, N. (1987) Overlay and solid-phase
analysis of glycolipid receptors for bacteria and viruses. Methods
Enzymol., 138, 220±232.
Kelm, S. and Schauer, R. (1997) Sialic acids in molecular and cellular
interactions. Int. Rev. Cytol., 175, 137±240.
Kontani, M., Kimura, S., Nakagawa, I., and Hamada, S. (1997)
Adherence of Porphyromonas gingivalis to matrix proteins via a
fimbrial cryptic receptor exposed by its own arginine-specific protease.
Mol. Microbiol., 24, 1179±1187.
Lee, J.Y., Sojar, H.T., Bedi, G.S., and Genco, R.J. (1992) Synthetic
peptides analogous to the fimbrillin sequence inhibit adherence of
Porphyromonas gingivalis. Infect. Immun., 60, 1662±1670.
Mackenzie, I.C., Dabelsteen, E., and Mandel, U. (1989) Expression of
blood group antigen-related carbohydrates by human gingival
epithelia. J. Periodontal Res., 24, 289±97.
Miller-Podraza, H., Milh, M.A., Bergstr
om, J., and Karlsson, K.-A. (1996)
Recognition of glycoconjugates by Helicobacter pylori: an apparently
high-affinity binding of human polyglycosylceramides, a second sialic
acid-based specificity. Glycoconj. J., 13, 453±460.
Moore, W.E. and Moore, L.V. (1994) The bacteria of periodontal
diseases. Periodontol. 2000, 5, 66±77.
Mouricout, M. (1997) Interactions between the enteric pathogen and the
host. An assortment of bacterial lectins and a set of glycoconjugate
receptors. Adv. Exp. Med. Biol., 412, 109±123.
Oriol, R. (1995) ABO, Lewis and secretion. Serology genetics and tissue
distribution. Plenum Press, NY.
Ravn, V. and Dabelsteen, E. (2000) Tissue distribution of histo-blood
group antigens. APMIS, 108, 1±28.
Sandros, J., Papapanou, P., and Dahlen, G. (1993) Porphyromonas
gingivalis invades oral epithelial cells in vitro. J. Periodontal Res., 28,
219±226.
Sandros, J., Papapanou, P.N., Nannmark, U., and Dahlen, G. (1994)
Porphyromonas gingivalis invades human pocket epithelium in vitro.
J. Periodontal Res., 29, 62±69.
Smith, E.L., McKibbin, J.M., Karlsson, K.-A., Pascher, I.,
Samuelsson, B.E., and Li, S.C. (1975) Characterization of dog small
intestinal fucolipids with human blood group H activity. Biochemistry,
14, 3370±3376.
Socransky, S.S., Haffajee, A.D., and Dzink, J.L. (1988) Relationship of
subgingival microbial complexes to clinical features at the sampled
sites. J. Clin. Periodontol., 15, 440±444.
Strokan, V., M
olne, J., Svalander, C.T., and Breimer, M.E. (1998a)
Heterogeneous expression of Gala1-3Gal xenoantigen in pig kidney: a
lectin and immunogold electron microscopic study. Transplantation,
66, 1495±1503.
Strokan, V., Rydberg, L., Hallberg, E.C., M
olne, J., and Breimer, M.E.
(1998b) Characterization of human natural anti-sheep xenoantibodies.
Xenotransplantation, 5, 111±121.
Str
omberg, N. and Boren, T. (1992) Actinomyces tissue specificity may
depend on differences in receptor specificity for GalNAcb-containing
glycoconjugates. Infect. Immun., 60, 3268±3277.
Ê ., A
Ê ngstr
Teneberg, S., Leonardsson, I., Karlsson, H., Jovall, P.-A
om, J.,
Ê ., Wadstr
Danielsson, D., NaÈslund, I., Ljungh, A
om, T., and
Karlsson,
K.-A.
(2002)
Lactotetraosylceramide,
a
novel
glycosphingolipid receptor for Helicobacter pylori, present in human
gastric epithelium. J. Biol. Chem., 277, 19709±19719.
Ê ngstr
Ê . (2004) Carbohydrate recogniTeneberg, S., A
om, J., and Ljungh, A
tion by enterohemorrhagic Escherichia coli: characterization of a novel
glycosphingolipid from cat small intestine. Glycobiology, 14,
187±196.
Ulfvin, A., Breimer, M.E., Holgersson, J., Vinas, J., and Samuelsson, B.E.
(1988) ABO blood grouping in erythrocytes and antigen expression in
different organs. Librairie Arnette, Paris.
Umesaki, Y., Takamizawa, K., and Ohara, M. (1989) Structural and
compositional difference in the neutral glycolipids between epithelial
and non-epithelial tissue of the mouse small intestine. Biochim.
Biophys. Acta, 1001, 157±162.
Yang, Z., Bergstr
om, J., and Karlsson, K.-A. (1994) Glycoproteins with
Gala4Gal are absent from human erythrocyte membranes, indicating
that glycolipids are the sole carriers of blood group P activities. J. Biol.
Chem., 269, 14620±14624.
519