Bacteroides fragilis signals through Toll-like
receptor (TLR) 2 and not through TLR4
Mohammad Alhawi,13 John Stewart,1 Clett Erridge,2 Sheila Patrick3
and Ian R. Poxton1
Correspondence
Ian R. Poxton
i.r.poxton@ed.ac.uk
1Medical Microbiology, Centre for Infectious Diseases, University of Edinburgh College of Medicine
and Veterinary Medicine, Chancellor’s Building, 49 Little France Crescent, Edinburgh EH16 4SB,
UK
2Department of Cardiovascular Sciences, University of Leicester, Clinical Science Wing, Glenfield
General Hospital, Leicester LE3 9QP, UK
3Centre for Infection and Immunity, School of Medicine, Dentistry and Biomedical Sciences,
Queen’s University Belfast, 97 Lisburn Road, Belfast BT9 7BL, UK
Received 2 February 2009
Accepted 9 April 2009
Although it is desirable to identify the interactions between endotoxin/LPS and the innate immune
mechanism, it is often not possible to isolate these interactions from other cell wall-related
structures of protein or polysaccharide origin. There is no universally accepted method to extract
different LPSs from different bacteria, and their natural state will be influenced by their interactions
with the associated molecules in the bacterial outer membrane. It is now believed that Toll-like
receptor (TLR) 4 is the main signal transducer of classical LPS (i.e. Escherichia coli LPS), while
TLR2 is used by certain non-classical LPSs. There are contradictory reports as to whether
Bacteroides fragilis LPS, a non-classical LPS, signals primarily through TLR2 or TLR4. This study
was designed to address this problem. Different non-purified and purified B. fragilis LPSs
extracted by different methods together with different heat-killed, whole-cell populations of B.
fragilis were used to elucidate the TLR specificity. All of these B. fragilis preparations showed a
significant signalling specificity for TLR2 but not for TLR4. This indicates that changing the
extraction methods, with or without applying a repurification procedure, and varying the cell
populations do not alter the TLR specificity of B. fragilis LPS.
INTRODUCTION
Species belonging to the genus Bacteroides are the
predominant Gram-negative, anaerobic bacteria found in
the normal faecal/colonic microbiota. Bacteroides fragilis is
of great importance as an opportunistic pathogen and is
commonly associated with bacteraemia, soft tissue infec-
tions, intra-abdominal infections and abscesses (Mancuso
et al., 2005) often following trauma to the colon. Other
Bacteroides species such as Bacteroides vulgatus and
Bacteroides thetaiotaomicron are found more commonly
than B. fragilis in faeces, resulting in the commonly held
view that it is not found in high concentrations in the gut;
however, in some individuals it is the most common
species of Bacteroides found attached to the colonic mucosa
(Poxton et al., 1997). In terms of pathogenicity, the LPSs of
B. fragilis have long been known to be of lower
endotoxicity than those of other enteric bacteria such as
Escherichia coli (Hofstad et al., 1977). However, from
studies in our laboratory we showed that the endotoxicity
was dependent on the extraction technique used
(Delahooke et al., 1995).
B. fragilis produces antigenically distinct and within strain
variable large capsules (LCs) and small capsules (SCs)
distinguishable by light microscopy when cultures are
grown in a glucose-defined medium. In addition, bacterial
cells which appear to be non-capsulate by light microscopy
may produce a marginal micro-capsule (MC) of approxi-
mately 35 nm in size visible as an electron-dense layer
adjacent to the outer membrane, visible by electron
microscopy. The electron-dense MC is also visible beneath
the LC by electron microscopy, but is not associated with
the SC (Patrick et al., 1986). On primary isolation,
individual cultures usually contain a mixture of LC, SC
and MC cells. Populations of B. fragilis enriched for each of
the different capsules can, however, be obtained by using
Percoll density-gradient centrifugation (Reid et al., 1987).
Abbreviations: AP, aqueous phenol; BWP, boiling water/proteinase K;
LAL, Limulus amoebocyte lysate; LC, large capsule; MC, micro-capsule;
PCP, phenol/chloroform/petroleum spirit; PS, polysaccharide; SC, small
capsule; TLR, Toll-like receptor; TM/TMP, Triton/magnesium chloride
without and with proteinase K treatment.
3Present address: College of Health Sciences, King Abdulaziz
University, PO Box 128098, Jeddah 21362, Saudi Arabia.
Journal of Medical Microbiology (2009), 58, 1015–1022 DOI 10.1099/jmm.0.009936-0
009936 G 2009 SGM Printed in Great Britain 1015

Examination of the MC- and LC-enriched population
indicates that there are multiple antigenically different
polysaccharides variably expressed within single popula-
tions (Patrick et al., 1999). The determination of the
complete genome sequence of B. fragilis NCTC 9343
indicated that it lacks a single operon with similar genetic
structure to the loci involved in O-antigen or lower-
molecular-mass lipid A linked K antigen synthesis in E. coli.
Unusually, however, there are eight PS loci (A–H) that are
related to variable MC expression and each contains genes
predicted to encode O-antigen polymerase (Wzy) and
flippase (Wzx)-like proteins. The variable expression at
seven of these loci is controlled by site-specific serine
recombinase directed inversion of the promoter region
(Cerdeno-Tarraga et al., 2005; Patrick et al., 2003).
Investigation of NCTC 9343 by targeted gene deletion
suggests that the antigenically variable MC equates to
enteric LPS, but instead of the classical low-molecular-mass
O-antigen, it contains a high-molecular-mass polysacchar-
ide that forms the antigenically variable electron-dense MC
visible by electron microscopy outside and adjacent to the
outer membrane (Patrick et al., 2009). Thus the LPS of B.
fragilis differs from classical enteric LPS not only with
respect to phosphorylation of the lipid A diglucosamine
and 3-deoxy-D-manno-octulosonate (reviewed by Patrick,
2002), but also with respect to the O-antigen.
It is recognized widely today that innate immunity against
LPS is considered to be one of the principal defences
against Gram-negative bacterial infection. Of particular
interest in this regard was the discovery of Toll-like
receptors (TLRs), suggested to be a cornerstone in the LPS
immuno-sensing machinery since they guided the elucida-
tion of the cascade of molecules which collectively
recognize LPS and eventually lead to the production of
the innate inflammatory response (Alexander & Rietschel,
2001). Although TLR4, in particular, is considered to be
the main signal transducer of classic LPS (Hirschfeld et al.,
2000; Lien et al., 2000; Poltorak et al., 1998), there are
contradictory reports as to whether TLR2 (Erridge et al.,
2004, 2007) or TLR4 (Mancuso et al., 2005) plays a
principal role in signalling B. fragilis LPS as a non-classical
LPS.
The main aim of this study was to investigate whether
different B. fragilis LPS preparations from different
extraction methods and different phenotypic subpopula-
tions of B. fragilis signalled through the same Toll-like
receptor.
METHODS
Bacterial strains and culture conditions. Strains of B. fragilis
(NCTC 9343: MPRL 1669), E. coli O18K2 (MPRL 1275) and
Porphyromonas gingivalis (NCTC 11834) were used. MPRL numbers
indicate strains from our laboratory culture collection. B. fragilis was
grown in proteose-peptone yeast extract medium (Holbrook et al.,
1977) or defined medium (Van Tassell & Wilkins, 1978) under
anaerobic conditions at 37 uC. E. coli O18K2 was grown in nutrient
broth under aerobic conditions at 37 uC. P. gingivalis was grown in
brain heart infusion (BHI) broth (Oxoid) supplemented with haemin
(7.7 mM) under anaerobic conditions at 37 uC. All cultures were
checked for purity by Gram stain and subculture under aerobic and
anaerobic conditions.
Extraction of LPSs. Five different B. fragilis LPS preparations were
obtained by extraction of lyophilized bacterial cells by four different
extraction methods, namely: (i) the aqueous phenol (AP) method as
described originally by Westphal & Luderitz (1953); (ii) the phenol/
chloroform/petroleum spirit (PCP) method of Galanos et al. (1969)
with the modification of Qureshi et al. (1982); (iii) the Triton/
magnesium chloride method, applied to extract two kinds of LPS with
(TMP) and without (TM) proteinase K treatment according to the
work of Uchida & Mizushima (1987); and finally (iv) the boiling
water/proteinase K treatment (BWP) method, which was applied to
extract LPSs as proposed by Eidhin & Mouton (1993). E. coli and P.
gingivalis LPSs were extracted using the AP method (i) only. Full
details of extraction methods (i) and (ii) are described in Hancock &
Poxton (1988).
Repurification of LPS preparations. All LPS samples were
subjected to a repurification step to eliminate the protein con-
taminants which might be responsible for TLR2 signalling according
to the procedure described by Manthey & Vogel (1994) and further
detailed by Hirschfeld et al. (2000).
Preparation of heat-killed bacteria. For TLR specificity assays,
three different subpopulations of B. fragilis were prepared by
discontinuous Percoll density-gradient centrifugation of broth
cultures grown in a defined medium (Van Tassell & Wilkins, 1978)
which contained a mixed population of LC, SC and electron-dense
MC layer cells. Populations enriched for expression of each of the LC,
SC or MC were prepared by subculture from the 0–20% Percoll, 20–
40% Percoll and 60–80% Percoll interface layers, respectively.
Enrichment for capsule expression was monitored by negative capsule
staining (Patrick et al., 1986). The bacteria were heat-killed in a 95 uC
water bath for 30 min.
Analysis of LPS preparations. After LPSs were extracted, the
preparations were examined by three procedures: firstly, PAGE (12%
without SDS in the gel buffers, but included in the sample buffer) was
prepared with the buffer system of Laemmli (1970) according to the
method of Poxton & Brown (1979). To visualize LPSs, gels were
processed according to the method of Tsai & Frasch (1982) as
described by Hancock & Poxton (1988). Secondly, protein bound to
LPS samples was detected by colloidal gold total protein stain (Bio-
Rad) according to the manufacturer’s instructions. Finally, the
endotoxic activity of LPS samples was determined by Limulus
amoebocyte lysate assay (LAL assay) using the Pyrochrome LAL kit
(Associates of Cape Cod) according to the manufacturer’s instruc-
tions. The A405 was read on an Anthos 2001 automated plate reader.
Cells and transfection assays. LPS and heat-killed bacteria
signalling via TLR2 or TLR4/MD2 was assessed using a TLR-deficient
human embryonic kidney (HEK) 293 cell line transfection assay as
previously described (Erridge et al., 2007). Briefly, HEK-293 cells were
plated in 96-well plates at 86103 cells per well and transfected after
24 h using GeneJuice (Novagen) according to the manufacturer’s
instructions. Amounts of construct per well were 10 ng human
pTLR2, 30 ng human pTLR4/MD2 (Invivogen), 30 ng pCD14 (kind
gift of Professor Christopher Gregory, University of Edinburgh, UK),
20 ng renilla reporter construct and 10 ng luciferase-reporter
construct driven by the NF-kB-regulated E-selectin promoter
(pELAM), with the balance made up with empty pCMV. A plasmid
encoding human mCD14 as a co-transfectant was always included in
every transfection assay for both TLR2 and TLR4 transfections. The
M. Alhawi and others
1016 Journal of Medical Microbiology 58