P0884453-Ch37.qxd
10/19/05
8:49 PM
Page 659
C
H
A
P
T
E
R
37
Comparative three-dimensional
structure of cholesterol-dependent
cytolysins
Galina Polekhina, Susanne C. Feil, Julian Tang, Jamie Rossjohn, Kara Sue Giddings,
Rodney K. Tweten, and Michael W. Parker
INTRODUCTION
brane. Several excellent reviews have been published
on these toxins (Alouf and Geoffroy, 1991; Alouf, 2000;
Gilbert et al., 2000; Palmer, 2001; Heuck et al., 2001;
Tweten et al., 2001; Gilbert, 2002), so this review will
focus on recent advances in structural studies of CDC
toxins. Biological aspects of the toxins discussed in this
chapter are covered in other chapters of Section III, particularly in Chapter 38 “Perfringolysin O and intermedilysin: mechanisms of pore formation by the
cholesterol-dependent cytolysins.”
The cholesterol-dependent cytolysins (CDCs) are one
of the most widely distributed toxins known. The toxin
gene or its product has been identified in numerous
species from seven different genera of Gram-positive
bacteria, including the Clostridium, Bacillus, Streptococcus,
Listeria, Brevibacillus, Paenibacillus, and most recently the
Arcanobacterium (Tweten et al., 2001). The fact that this
gene is so widely distributed among these various
pathogenic bacteria suggests that it plays an important
role in the pathogenic mechanism of these organisms.
For example, listeriolysin O is an essential virulence
factor that is responsible for the release of the intracellular pathogen Listeria monocytogenes from the phagocytic vacuole (Jones and Portnoy, 1994).
The CDCs exhibit many unique features, including
an absolute dependence of their cytolytic activity on
the presence of cholesterol in the membrane and also
the formation of very large oligomeric pores, more
than 150 Å in diameter, on the membranes of cells.
These toxins have been shown to be cytolytic to many
eukaryotic cell types, although the bulk of the literature has focused on the hemolytic activity of these toxins. The crystal structures of two CDCs have been
solved, and experimental approaches combining
molecular biology techniques and various biophysical
analyses have helped uncover fundamental features by
which these toxins assemble and insert into the memThe Comprehensive Sourcebook of Bacterial Protein Toxins
Alouf and Popoff
CDC PRIMARY STRUCTURES
There are more than 20 members of the CDC family so
far identified, and there exists a high degree of sequence
similarity (40–80%) among them, suggesting they all
have similar activities and 3D structures. The primary
structures of streptolysin O (SLO), pneumolysin (PLY),
and perfringolysin O (PLO) were revealed in the late
1980s by DNA sequence analyses of their genes (Kehoe
et al., 1987; Tweten et al., 1988; Walker et al., 1987). These
studies showed that all three toxins exhibited a significant level of identity in their primary structures (Figure
37.1), and that each contained a single cysteine residue
in a highly conserved region near the C-terminal end of
each protein. It was this cysteine, upon modification
with thiol-specific reagents, that was responsible for the
loss in cytolytic activity of these toxins (Iwamoto et al.,
659
Copyright © 2006, Elsevier Ltd.
All rights reserved.
P0884453-Ch37.qxd
660
10/19/05
8:49 PM
Page 660
COMPARATIVE THREE-DIMENSIONAL STRUCTURE OF CHOLESTEROL-DEPENDENT CYTOLYSINS
FIGURE 37.1 Sequence alignment of various cholesteroldependent cytolysins. Numbering
refers to the PFO sequence;
strictly conserved regions are
shaded in dark gray, and highly
conserved regions are shaded in
light gray. The secondary structure of PFO, as seen in the crystal
structure, is shown above the
alignments (helices as cylinders
and strands as arrows). The
TMH1, TMH2, and undecapeptide regions are shown boxed, in
that order, from the N-terminus.
1987). However, this single cysteine residue was shown
not to be required for the in vitro cytolytic activity of
SLO and PLY (Saunders et al., 1989; Pinkney et al., 1989),
and more recently for PFO (Shepard et al., 1998).
Substitution of the cysteine with alanine by in vitro
mutagenesis yielded a toxin molecule that was similar
in activity to that of the cysteine-containing wild-type.
However, substitution of the cysteine with serine or
glycine caused a significant decrease in the cytolytic
activity of SLO and PLY (Saunders et al., 1989; Pinkney
et al., 1989). Therefore, even though the sulfhydryl
group is not required for the in vitro cytolytic activity of
these toxins, the cysteine residue apparently occupied a
site within the toxin structure that is critical to the function of the CDC. Why the cysteine has been retained at
this position in these toxins, when alanine would function equally well and is not susceptible to oxidation, is
not clear. Of the 11 sequenced CDCs, only pyolysin
(PLO) from Arcanobacterium pyogenes and intermedilysin (ILY) from Streptococcus intermedius (Figure
37.1) have an alanine substituted for the cysteine.
Although it is clear that the sulfhydryl is not required
for cytolytic activity, it is not known if it has some as yet
undefined role in vivo.
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
10/19/05
8:49 PM
Page 661
CRYSTAL STRUCTURE OF PFO
A cysteine-containing, highly conserved undecapeptide sequence, ECTGLAWEWWR, is present in eight of
the 11 sequenced CDCs (Figure 37.1). The remaining
three toxins exhibit various substitutions in this region,
some of which are conservative and others of which are
not. In addition to containing the cysteine residue, this
region also contains a conspicuously large number of
tryptophan residues: 10 of the 11 sequenced toxins contain three tryptophans in the undecapeptide. The
undecapeptide sequence of seeligeriolysin from
Listeria seeligleri has a single residue change in this
sequence in which a phenylalanine is substituted for
an alanine. However, the more recently discovered and
sequenced CDCs, PLO (Billington et al., 1997) and ILY
(Nagamune et al., 1996; Nagamune et al., 2000), exhibit
significant differences in this region (Figure 37.1). Both
have an alanine substituted for the cysteine residue,
thus making them resistant to inactivation due to oxidation of the sulfhydryl. These two toxins also have
significant differences in the last 4–5 residues of the
undecapeptide. PLO has a conservative change of Glu
to Asp (Asp498 of PLO), but also contains an insertion
of a proline between Asp498 and Trp500. ILY also
exhibits the same aspartate to glutamate and cysteine
to alanine changes as PLO, but instead of inserting a
proline, a proline has been substituted at position 494,
a position where tryptophan would normally be found
(Figure 37.1). Therefore, ILY only contains two tryptophans in this region, whereas all of the other toxins
have three tryptophans. The role of the highly conserved undecapeptide in the cytolytic mechanism has
not yet been completely clarified.
There are many additional differences in the primary
structures of these toxins, none of which has been
linked to a unique function of a particular CDC.
However, one difference is worth noting and is unique
to the structure of SLO, which is produced by
Streptococcus pyogenes. When the SLO amino acid
sequence is aligned with the primary structures of the
other CDCs, an additional 70–75 amino acids are present at its amino terminus. The additional residues in
SLO are located between the predicted signal peptide
and the region of SLO that aligns with the approximate
amino termini of the secreted forms of the other toxins.
This additional sequence does not align with any of the
other CDCs and does not exhibit any significant
sequence similarity with any other protein. It has been
shown that SLO can be nicked with the cysteine proteinase of S. pyogenes between residues Lys 77 and Leu
78 (Gerlach et al., 1993). This cleavage removes 46
amino acids from the secreted SLO and yields a 55.5
kDa protein. Both the uncleaved and cleaved forms are
hemolytically active. Although this cleavage removes a
significant portion of the extra sequence, there still
661
remains an extra 26 residues on the small form of SLO,
which does not exhibit any sequence similarity with
the other CDCs. If the amino terminal region has a
function specific for SLO, it has yet to be demonstrated.
PFO—AN ARCHETYPICAL
CDC
Clostridium perfringens is a normal inhibitant of the gastrointestinal tract of humans and animals, as well as
being commonly found in soil. It causes human gas
gangrene and food poisoning, as well as several enterotoxemic diseases in animals. The pathogenesis of gas
gangrene (or clostridial myonecrosis) primarily
involves the invasion of traumatic wounds or deoxygenated tissues by the Gram-positive bacterium. C. perfringens releases numerous virulence factors, including
two lethal toxins, alpha-toxin and PFO, which are
thought to be the major virulence factors of the bacterium. PFO, or theta-toxin, induces tissue destruction
and anti-inflammatory responses and acts synergistically with alpha-toxin in gangrenous lesions (Rood,
1998). PFO is produced by all strains of C. perfringens.
Its gene encodes a polypeptide that consists of a 27
amino acid signal peptide followed by 500 amino acids
of the mature protein.
CRYSTAL STRUCTURE OF
PFO
The crystal structure of PFO was determined in 1997
(Rossjohn et al., 1997). The structure was originally
determined to a resolution of 2.7 Å, but has since been
extended to a resolution of 2.2 Å (Rossjohn, J.,
Polekhina, G., Feil, S.C. McKinstry, W.J., Tweten, R.K.
and Parker, M.W., unpublished results). The PFO molecule is a very elongated molecule with its long axis
measuring approximately 115 Å (Figure 37.2). A
notable feature of the secondary structure is that it is
very rich in beta-sheet. Although the molecule did not
closely resemble any other molecule for which a crystal structure was known, its shape and secondary
structure content were reminiscent of a number of
other toxins including aerolysin (Parker et al., 1994),
Staphyloccus α-hemolysin (Song et al., 1996), the protective antigen of anthrax toxin (Petosa et al., 1997), and
LukF (Olson et al., 1999; Pédelacq et al., 1999).
The crystal structure demonstrated that PFO is built
from four domains: the N-terminal domain or domain
1 (residues 37–53, 90–178, 229–274, 350–373), domain 2
(residues 54–89, 374–390), domain 3 (residues 179–228,
275–349), and the C-terminal domain or domain 4
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
662
10/19/05
8:49 PM
Page 662
COMPARATIVE THREE-DIMENSIONAL STRUCTURE OF CHOLESTEROL-DEPENDENT CYTOLYSINS
tures of domain 3 suggested the possibility that it could
readily flex away from domain 2 so as to relieve the
energetically unfavorable stress at the domain 1–3
interface.
Domain 4 was of particular interest because it houses
the undecapeptide sequence, which had previously
been implicated in cholesterol and membrane binding.
This sequence was found to form an extended loop
with a single turn of helix at the tip of the molecule. The
loop curls back into one of the beta-sheets so that
the tip of the loop defined by Trp 464 is nestled into
the hydrophobic core of a number of long surface
side chains. This immediately suggested that this
site might be the cholesterol-binding site with the
tryptophan side chain mimicking how a cholesterol
molecule would bind if the loop was displaced.
The undecapeptide motif itself consists of mostly
hydrophobic residues, so its displacement would generate a hydrophobic “dagger” that would be capable of
inserting into a membrane. Domain 4 is connected to
the rest of the protein through a single linking peptide,
suggesting that it too could be quite mobile in solution.
STRUCTURE/FUNCTION
STUDIES OF PFO
FIGURE 37.2 Crystal structure of PFO. A ribbon representation
indicating the location of domains and the undecapeptide motif (in
dark shade). The figure was drawn using MOLSCRIPT (Kraulis,
1991).
(residues 391–500) (Figure 37.2). Domain 1, located at
one end of the molecule, adopts an alpha-beta topology with a long helix packing against a core of antiparallel beta-sheet. Domain 2 consists of a single layer
of anti-parallel beta-sheet connecting one end of the
molecule to the other. Domain 3 also adopts alpha-beta
topology with a core anti-parallel beta-sheet surrounded by helical layers on both sides. Domain 4
adopts a beta-sandwich topology, a common fold
found in a variety of proteins.
Two of the domains, domains 3 and 4, exhibited
unusual features. The core beta-sheet that runs through
domains 1 and 3 has a highly pronounced curvature at
the domain interface. The packing of domain 3 onto the
rest of the protein is far from complementary and
involves predominantly polar contacts. (Normally,
domain interfaces have very complementary surfaces
and regions of hydrophobic contacts). These two fea-
A major advance in understanding the mechanism
of CDC membrane insertion came with the discovery
that domain 3 harbored regions that formed the transmembrane (TM) β-barrel. The studies of Shepard et al.
and Shatursky et al. revealed that a series of small
helices on either side of a central β-sheet could unfurl
into β-hairpins, TMH1 and TMH2, that were capable of
spanning membrane bilayers (Figure 37.3). Such conformational changes were unprecedented among toxins and hence represented a new paradigm for how
pore-forming proteins generated a membrane-spanning domain.
PFO must interact with membranes as a prerequisite
for the insertion of the domain 3 TM hairpins that form
the wall of the transmembrane β-barrel (Heuck et al.,
2000). The conserved undecapeptide and three other
short hydrophobic loops, all located at the tip of
domain 4 of PFO, penetrate the surface of the membrane and anchor the molecule for the subsequent conformational changes during pore formation (Heuck
et al., 2000; Nakamura et al., 1995; Ramachandran et al.,
2002). Domains 3 and 4 are conformationally coupled,
although they are not in direct physical contact.
Mutations in domain 3 can affect the rate at which
domain 4 interacts with the membrane, even though
the domain 4 undecapeptide interacts with the membrane prior to the insertion of the domain 3 TMHs
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
10/19/05
8:49 PM
Page 663
STRUCTURE/FUNCTION STUDIES OF PFO
663
FIGURE 37.3 A schematic model of stages of CDC insertion into membranes. From left to right: Ribbon representation of the PFO monomer
as seen in the crystal structure. This is the structure thought to exist in solution. The TMH1 and TMH2 are highlighted in dark shade, and the
undecapeptide loop is shown sprung out with a molecule of cholesterol binding to it. Next, domain 3 starts to rotate away from the body of the
molecule, and the TMH regions are released simultaneously to form extended beta hairpins. It is likely that domain 4 curls into the body of the
molecule, allowing the TMH regions to fully puncture the membrane so as to form a beta-barrel with the other monomers of the oligomer. Only
one monomer is shown for clarity. This figure was produced using MOLSCRIPT (Kraulis, 1991).
(Palmer et al., 1998; Abdel Ghani et al., 1999; Heuck
et al., 2000; Hotze et al., 2001; Ramachandran et al.,
2002). The cytolytic activity of PFO is highly sensitive
to changes in the undecapeptide motif, especially
mutations of the conserved tryptophans or chemical
modification of the thiol of the conserved cysteine,
which inhibit membrane binding (Nakamura et al.,
1995; Sekino-Suzuki et al., 1996, Alouf, 2000; Iwamoto
et al., 1987). However, some undecapeptide mutations
or modifications appear to affect the global structure of
PFO, thus suggesting that its conformation is important to the proper function of the toxin.
A detailed model of the initial stages of membrane
insertion by CDCs is now starting to emerge (Figure
37.3). In the first step, the toxin binds to cell surfaces via
domain 4. The undecapeptide loop flips out from this
domain to form a membrane-penetrating hydrophobic
dagger that anchors the protein to the membrane.
Many studies suggest that cholesterol acts as a receptor
in this step, triggering the conformational change,
although recent data suggests the role of cholesterol in
this step is toxin dependent (Giddings et al., 2003). The
partial insertion of domain 4 would bring domain 3
closer to the membrane surface and also would make
the toxin oligomerization-competent (Abdel Ghani
et al., 1999). A recent study suggests membrane binding
is associated with conformational changes in domain 3
(Abdel Ghani et al., 1999). A prepore complex assem-
bles on the membrane surface by lateral diffusion of
membrane-bound monomers (Shepard et al., 2000;
Hotze et al., 2001; Hotze et al., 2002). Electron
microscopy studies of the CDC, pneumolysin, show
that the prepore complex is bound to the membrane
via the base of domain 4 where the undecapeptide loop
resides (Gilbert et al., 1999). The TMH regions in
domain 3 are kept unfurled at this stage to prevent premature membrane insertion of isolated monomers
(Shepard et al., 2000; Heuck et al., 2000; Hotze et al.,
2001). The formation of the oligomer triggers a further
conformational change in domain 4, causing a rotation
of domain 4 away from the long axis of the molecule
and a concerted rotation of domain 3 as described previously. The rotation of domain 3 away from the molecule in the oligomer has been directly observed by
electron microscopy (Gilbert et al., 1999). These rotations are of the order of 35˚ to 45˚ for domain 4 and
would be sufficient to break all contacts between
domains 2 and 3 and allow the concerted unfurling of
the TMH regions to form β-hairpins (Hotze et al., 2001).
Electron microscopy studies of pneumolysin oligomers
support our suggestion that domain 3 swings out from
the body of the molecule to adopt a disordered conformation and that the rotation of domain 3 is coupled to
domain 4 rotations (Gilbert et al., 1999). It has been
argued that a concerted insertion of the β-hairpins
from all monomers of the oligomer then occurs based
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
664
10/19/05
8:49 PM
Page 664
COMPARATIVE THREE-DIMENSIONAL STRUCTURE OF CHOLESTEROL-DEPENDENT CYTOLYSINS
on energetic considerations (Heuck et al., 2001), and
this hypothesis is supported by recent experimental
data (Hotze et al., 2002). Our modeling of the oligomer
suggests that domain 4 must curl into the body of the
molecule so that the extended β-hairpins can penetrate
the bilayer. Very recently, a vertical collapse of about 40
Å was observed by atomic force microscopy between
the prepore and pore states of PFO. The data is consistent with domain 2 collapsing into a more globular
domain, rather than the elongated domain seen in the
crystal structure of the water-soluble oligomer
(Czajkowsky et al., 2004).
NEW CRYSTAL STRUCTURE
OF PFO
The structure of PFO has recently been solved in
another crystal form at 3 Å resolution (Rossjohn, J.,
Polekhina, G., Feil, S.C., McKinstry, W.J., Tweten, R.K.
and Parker, M.W., unpublished results). These crystals
were grown using butanol as the precipitant, and they
contain two molecules per asymmetric unit. The two
molecules pack in a head-to-tail arrangement and are
essentially identical to each other, with a root-meansquare (r.m.s) deviation of 0.6 Å over all Calpha atoms.
They also superpose well with the previously published structure, with an overall r.m.s deviation of 1.0 Å
over all Calpha atoms. The conformation of the undecapeptide motif is also very similar, as is the conformation of the TMH1 and TMH2 helical springs,
demonstrating that the observed conformations of
these critical regions are not artefacts of the cystallization process.
The overall temperature factor, a measure of the
order of the molecule, was mapped onto the structures
for both the crystal forms. The β-sandwich of domain 4
and the 5-stranded sheet of domain 3 represent the
most ordered part of the molecule in the published
crystal form. The top and neck of the molecule appears
largely mobile, with the outer region of domain 2 being
very mobile. The new crystal form shows an overall
increased mobility over the entire backbone, despite
showing no large-scale structural differences with
respect to the wild-type structure. Domain 4 and the
β-sheet of domain 3 represent the most ordered regions
of the molecule as well.
MOLECULAR BASIS OF
THIOL ACTIVATION
CDCs were initially termed the thiol-activated
cytolysins because thiol-reagents such as N-ethyl
maleimide and mercury salts (Alouf and Geoffroy,
1991) inhibited cytolytic activity, whereas thiol-reducing reagents reversed these effects. This led to the
concept of the “essential thiol.” However, it was subsequently shown that the cysteine is not essential in vivo
(see above). Each CDC contains just one cysteine
throughout the sequence, with the exception of
ivanolysin, which contains two cysteines. This solitary
cysteine is located in the undecapeptide motif, and
thus the effects of the thiol-reagents must be directed to
the cysteine within the motif. The structural basis of
inactivation by these thiol reagents is unknown.
During the course of the structure determination of
PFO (Rossjohn et al., 1997), two mercury derivatives,
mercury acetate (HgOAc) and p-chloromercuri-benzene sulfonate (PCMBS), were used. The PCMBS and
HgOAc data sets extend to 2.7 Å and 2.9 Å, respectively. To address the structural basis of toxin inactivation, the two mercury-modified forms of PFO have
been refined (Rossjohn, J., Polekhina, G., Feil, S.C.
McKinstry, W.J., Tweten, R.K. and Parker, M.W.,
unpublished results). The electron density for both
models, in particular the region around the undecapeptide motif, is well defined. The structures of
PCMBS-PFO and HgOAc-PFO superpose well with
respect to the 2.2 Å resolution wild-type structure.
Somewhat surprisingly, the region around the undecapeptide in both metal complexes is virtually identical to the wild-type (Figure 37.4). The chemical
modification of Cys 459 can be seen not to disrupt
appreciably the conformation of the undecapeptide
motif, implying that these thiol reagents must inactivate PFO via another mechanism. One possibility is
that cysteine modification affects the flexibility of the
undecapeptide loop, which has been hypothesized to
flip out from the body of the molecule on binding cholesterol to form a membrane-penetrating hydrophobic
dagger (Rossjohn et al., 1997; Ramachandran et al.,
2002; and see below).
INTERMEDILYSIN—AN
ATYPICAL CDC
S. intermedius secretes the CDC ILY (Macey et al.,
2001), and ILY appears to be a major virulence factor
of S. intermedius, as it is expressed up to 10-fold more
in abscesses (Nagamune et al., 2000). ILY exhibits
three features that distinguish it from most other
CDCs. First, it has a variant undecapeptide sequence
(GATGLAWEPWR) (Figure 37.1). The normally conserved cysteine of the CDC undecapeptide is replaced
by an alanine residue, resulting in an oxidation-resistant toxin (Nagamune et al., 1996), and a proline
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
10/19/05
8:49 PM
Page 665
INTERMEDILYSIN—AN ATYPICAL CDC
665
FIGURE 37.4 Stereoview of the mercury complex structures in the vicinity of the undecapeptide loop in domain 4. The sulfur atom of the
unique thiol of residue Cys 459 packs against the aromatic ring of Trp 467. (A) Wild-type. (B) Mercury acetate complex. (C) p-Chloromercuribenzene sulfonate complex.
residue is substituted for the second conserved tryptophan. Second, ILY is specific for human cells
(Nagamune et al., 1996; Macey et al., 2001). Nagamune
and coworkers (1996) showed that animal erythrocytes from nine species are not susceptible to ILY and
non-human primate erythrocytes are a hundred-fold
less susceptible to ILY than human erythrocytes.
Third, free cholesterol is only weakly inhibitory.
Although the presence of membrane cholesterol is
still essential to the activity of ILY, it does not appear
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
666
10/19/05
8:49 PM
Page 666
COMPARATIVE THREE-DIMENSIONAL STRUCTURE OF CHOLESTEROL-DEPENDENT CYTOLYSINS
to function as a receptor as it does for PFO (Giddings
et al., 2003). Giddings and coworkers (2003) showed
that depletion of approximately 90% of the membrane
cholesterol did not affect ILY binding or that of the
related CDC, SLO, to human erythrocytes, but it did
completely abrogate the hemolytic activity of both
toxins. It was determined by fluorescence spectroscopy that cholesterol depletion prevented the prepore to pore conversion of ILY and SLO. Therefore,
cholesterol was necessary for ILY membrane insertion, but not its membrane binding.
The unique aspects of the ILY mechanism, its
variant undecapeptide and human cell specificity,
prompted us to investigate whether the two aspects
were linked. We have pursued a detailed structural
investigation of ILY with the aim of understanding
how these distinguishing features impact on its
interaction with cells.
CRYSTAL STRUCTURE OF
ILY
The crystal structure of ILY has been determined to a
resolution of 2.6 Å (Polekhina et al., 2004; Polekhina et
al., 2005). Overall, ILY adopts a similar structure to PFO
(see next section), so only a brief overview of the structure will be presented here. It is an elongated
boomerang-shaped molecule of dimensions 50 Å × 58
Å × 114 Å (Figure 37.5). The molecule is rich in β-sheet
with 40% sheet made up of 23 β-strands contributing to
four sheets. One strand (residues 399 to 425) spans
about two-thirds of the molecule and measures 100 Å
in length. There are 15 helices, varying in length
between 4 and 21 residues.
The molecule is comprised of four discontinuous
domains. Domain 1 (residues 56 to 79, 116 to 205, 256 to
301, 377 to 400) has an α/β structure with a core of sixstranded β-sheet. Domain 2 (residues 80 to 115, 401 to
417) consists of a three-stranded β-sheet. Domain 3
(residues 206 to 255, 302 to 376) is comprised of an
α/β/α three-layer structure. The interface of domains
2 and 3, covering a surface area of 592 Å2, is constructed from the packing of two helices against the βsheet of domain 2 and consists of four potential
hydrogen-bonding interactions and 17 van der Waals
interactions. Domain 2 is covalently connected to
domain 4 through a single connection. Domain 4
(residues 418 to 528) is folded into a compact β-sandwich consisting of four- and five-stranded sheets. The
interface between domains 2 and 4, covering 242 Å2,
consists of one salt link (between Arg 92 and Glu 448),
three hydrogen-bonding interactions, and 10 van der
Waals interactions. In addition, there are numerous
FIGURE 37.5 Crystal structure of ILY. A ribbon representation
indicating the location of domains and the undecapeptide loop. The
figure was drawn using MOLSCRIPT.
sulfate ion-mediated contacts between the two
domains (see below).
ILY crystallized with two molecules in the asymmetric unit. The two molecules superimpose with a r.m.s.
deviation of 2.2 Å on Calpha positions. This large deviation is due to a relative 10˚ rotation of domain 4 with
respect to the rest of the molecule, so that the tips of
domain 4 are nearly 16 Å apart when domain 1 of each
molecule is superimposed. Superposition of domains 1
to 3 yields a r.m.s. deviation of 0.7 Å and superposition
of domain 4 yields a deviation of 0.6 Å. Hence, aside
from the rotation of domain 4, the two molecules are
very similar.
Sulfate ions were a vital ingredient for crystallization
of ILY (Polekhina et al., 2004). Four sulfate ions were
identified in the asymmetric unit and were found to
bind to the same two sites on each of the two protein
molecules. This suggests the sulfate binding sites may
be more than just an artifact of the crystallization
process. One sulfate ion is bound to Arg 92 and Arg 477
and hence bridges domains 2 and 4 (Figure 37.5).
Another sulfate is bound in an unusual manner as it is
coordinated by two consecutive residues Lys 87 and
Asn 88 (through both side chain and main chain
atoms), both from domain 2, as well as Arg 477 from
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
10/19/05
8:49 PM
Page 667
STRUCTURAL COMPARISON OF ILY AND PFO
667
the opposing molecule (Figure 37.5). It is conceivable
that the sulfate-binding sites might represent binding
pockets for the putative human protein receptor.
STRUCTURAL COMPARISON
OF ILY AND PFO
The most similar structure to ILY in the Protein Data
Bank (http://www.rscb.org/pdb) is PFO (Rossjohn
et al., 1997), the only other CDC for which a threedimensional atomic structure has been determined.
The two proteins have an overall pairwise sequence
identity of 40% (197 out of a total of 532 residues) that is
distributed relatively evenly over the entire sequence
of each molecule. Both proteins share the same number
of domains, and each domain shares the same overall
topology. The most striking difference between the two
molecules is the highly bent shape of ILY compared to
PFO, with domain 4 rotated between 20˚ and 30˚ away
from the main axis of the ILY molecule (Figure 37.6).
(The rotation is either 20˚ or 30˚, depending on which
ILY molecule in the asymmetric unit is used in the
superposition.)
Domain 4 is the most similar domain between the
two molecules with an r.m.s. deviation on 91 equivalenced Calpha atoms of 0.8 Å. The most significant difference in this domain is the conformation of the
undecapeptide loop that adopts an extended conformation in ILY, but is curled up against the β-sheet in
PFO (Figure 37.6). The extended conformation does
not appear to be an artifact of the crystallization, as the
loop of one of the two ILY molecules in the asymmetric
unit is not involved in crystal contacts, and the loops of
both molecules superimpose almost exactly. There is
also a movement of a hairpin loop by almost 10 Å at the
other end of this domain due to differences in domain 2
(Figure 37.6). Another notable difference is the distribution of electrostatic charge on the surface of domain
4 of each toxin. ILY is predominantly positively
charged, whereas PFO is negatively charged with differences particularly pronounced in the region of the
undecapeptide motif and at the C-terminus.
A comparison of the domain 4 primary structures of
ILY and PFO does not provide any obvious candidate
regions of domain 4 that may be involved in cellular
specificity of ILY (Figure 37.1). About 60% of the
residues are not conserved between the two domains,
and these differences are distributed comparatively
evenly throughout domain 4. A surface analysis over
the domain 4 region of the crystal structures is also not
illuminating. There are no compelling patches of
unique or unusual sequence on the surface of this
domain that would suggest a receptor-binding region.
FIGURE 37.6 Superposition of ILY and PFO. Calpha traces of ILY
(light gray) and PFO (black), based on a domain 1 superposition.
Domain 1 is the next closest with 1.2 Å r.m.s. deviation on 145 equivalenced Calpha atoms (Figure 37.6).
The most significant differences in this domain are the
positions of the two helices seen in PFO with the longer
helix (residues 178 to 195) shifted by about 3.5 Å and
the shorter one (residues 278 to 284) by 2.8 Å. There is
also an additional short helix (residues 122 to 127) in
ILY not seen in PFO. The N-terminal helix (residues 56
to 70) in ILY is longer and more ordered than in PFO
and is also shifted by about 3 Å compared to PFO.
Domain 3 with an r.m.s. deviation of 3.1 Å (over 130
equivalenced Calpha atoms) does not superimpose
well because, although the core β-sheet is similar
(r.m.s. deviation of 1.5 Å over 38 equivalenced atoms),
the helices packing on either side of the sheet adopt
different positions relative to the sheet (Figure 37.6).
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
668
10/19/05
8:49 PM
Page 668
COMPARATIVE THREE-DIMENSIONAL STRUCTURE OF CHOLESTEROL-DEPENDENT CYTOLYSINS
In particular, the helices on the surface of the ILY molecule (TMH2) are not as closely nestled to the sheet as
they are in PFO (with shifts of up to 7 Å), and they
exhibit high temperature factors. The helices at the
interface between domains 3 and 2 (TMH1) also shift
by up to 3 Å compared to PFO. If the two molecules are
superimposed using domain 1 only, domain 3 is shifted
by about 2 Å in ILY compared to PFO (Figure 37.6).
Domain 2 does not superimpose well due to a twist in
the β-sheet in the middle of the domain, causing a
r.m.s. deviation of 2.9 Å over 51 equivalenced Calpha
atoms (Figure 37.6). This deviation is responsible for
the pronounced kink in the shape of the molecule
compared to PFO.
The solution of the ILY structure demonstrates that
the basic features of the CDC 3D structure appear to be
conserved among the CDCs. A phylogenetic tree generated by the alignment of the known CDC primary
structures (15 in all), using the nearest-neighbor
method of Saitou and Nei (1987), shows that PFO and
ILY are two of the most distantly related CDCs (data
not shown), yet they exhibit similar 3D structures.
Therefore, it is likely that all CDCs will exhibit 3D
structures similar to that of PFO and ILY.
CHOLESTEROL-BINDING
SITES
Previous work has implicated cholesterol as the receptor for the CDC family, where it is thought to act by
concentrating the toxin in cholesterol-rich domains of
the target cell membrane to promote oligomerization
(Alouf and Geoffroy, 1991). We have previously
hypothesized that cholesterol could bind to PFO by
displacing the undecapeptide loop and binding to one
face of the β-sheet in domain 4 (Rossjohn et al., 1997).
A striking feature of the ILY structure is the extended
conformation of the undecapeptide loop (Figure 37.6),
a conformation predicted for PFO when it binds cholesterol. We have computationally docked a cholesterol
molecule into the ILY structure and found no preferred
clustering of cholesterol molecules into the putative
binding site (Polekhina et al., 2005). The conformation
of the undecapeptide loop in ILY provided a basis for
modeling the equivalent loop, in its sprung-out form,
in PFO. We then docked cholesterol into the modified
PFO structure and found that the majority of docks
yielded a close clustering of cholesterol molecules with
the same orientation and position in the putative binding pocket. The cholesterol molecule was found to
stack against the aromatic side chains of Trp 466 and
467 and the 3β-hydroxy moiety, forming a hydrogen
bond to Glu 407 of PFO. A comparison of the key
residues involved in interacting with cholesterol in
PFO with the equivalent residues in ILY showed that
some are not conserved. Glu 407 (PFO) is replaced by
Tyr 434 in ILY, and this residue does not appear capable
of providing a hydrogen bond to the 3β-hydroxyl of
cholesterol. Trp466 (PFO) is substituted by Pro493 in
ILY; the latter residue is likely responsible for the
extended conformation of the undecapeptide observed
in ILY and also disrupts the aromatic stacking interaction between cholesterol and the tryptophans
suggested in the PFO model.
In summary, PFO possesses a plausible binding site
in domain 4 for a cholesterol molecule if it is assumed
that the undecapeptide adopts an extended conformation, as observed in the crystal structure of ILY.
Surprisingly, ILY is not predicted to possess such a site.
We hypothesize that ILY does not need a cholesterolbinding site because its undecapeptide is already
“sprung out.” These suggestions are in keeping with
earlier work that has shown that free cholesterol is only
weakly inhibitory for ILY cytolytic activity in contrast
to PFO (Nagamune et al., 1996). Nevertheless, cholesterol does play a role in the insertion process of ILY
(Giddings et al., 2003), indicating that cholesterol may
play multiple roles in CDC activity.
CONCLUSION
The crystal structure of PFO has proved a valuable
basis for numerous structure-function studies that
have revealed many details of the membrane insertion process (Shepard et al., 1998; Shatursky et al.,
1999; Heuck et al., 2000; Shepard et al., 2000; Hotze
et al., 2001; Hotze et al., 2002; Ramachandran et al.,
2002; Heuck et al., 2003; Czajkowsky et al., 2004;
Ramachandran et al., 2004). The structure of ILY, an
atypical CDC, demonstrates that the overall protein
fold first seen in PFO is almost certainly maintained
in all members of the family. Key differences include
different orientations of domain 4 with respect to the
rest of the protein and different conformations of the
undecapeptide loop, presumably in response to
sequence differences in the region. Further structural
studies are required to illuminate the role of cholesterol in CDC activity and the molecular detail of the
CDC oligomer.
ACKNOWLEDGMENTS
This work was supported by a grant from the National
Health and Medical Research Council of Australia
(NHMRC) to Michael Parker and the National
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
10/19/05
8:49 PM
Page 669
REFERENCES
Institutes of Health (NIAID, AI37657) to Rodney
Tweten. Galina Polekhina is supported by a NHMRC
RD Wright Research Fellowship, Jamie Rossjohn is a
Wellcome Trust Senior Research Fellow, and Michael
W. Parker is a NHMRC Senior Principal Research
Fellow.
REFERENCES
Abdel Ghani, E.M., Weis, S., Walev, I., Kehoe, M., Bhakdi, S. and
Palmer, M. (1999). Streptolysin O: inhibition of the conformational change during membrane binding of the monomer prevents oligomerization and pore formation. Biochemistry 38,
15204–15211.
Alouf, J.E. and Geoffroy, C. (1991). The family of the antigenicallyrelated, cholesterol-binding (“sulphydryl-activated”) cytolytic
toxins. In: Sourcebook of Bacterial Toxins (eds. J.E. Alouf, and J.J.
Freer), pp. 147–186. Academic Press, London.
Alouf, J. E. (2000). Cholesterol-binding cytolytic protein toxins. Int.
J. Med. Microbiol. 290, 351–356.
Billington, S.J., Jost, B.H., Cuevas, W.A., Bright, K.R. and Songer, J.G.
(1997). The Arcanobacterium (Actinomyces) hemolysin, pyolysin, is
a novel member of the thiol-activated cytolysin family. J. Bacteriol.
179, 6100–6106.
Czajkowsky, D.M., Hotze, E.M., Shao, Z. and Tweten, R.K. (2004).
Vertical collapse of a cytolysin prepore moves its transmembrane
beta-hairpins to the membrane. EMBO J. 23, 3206–3215.
Gerlach, D., Kohler, W., Gunther, E. and Mann, K. (1993). Purification
and characterization of streptolysin O secreted by Streptococcus
equisimilis (group C). Infect. Immun. 61, 2727–2731.
Giddings, K.S., A.E. Johnson, A.E. and Tweten, R.K. (2003).
Redefining cholesterol’s role in the mechanism of the cholesterol-dependent cytolysins. Proc. Natl. Acad. Sci. USA 100,
11315–11320.
Gilbert, R.J.C., Jiménez, J.L., Chen, S., Tickle, I.J., Rossjohn, J., Parker,
M.W., Andrew, P.W. and Saibil, H.R. (1999). Two structural transitions in membrane pore formation by pneumolysin, the poreforming toxin of Streptococcus pneumoniae. Cell 97, 647–655.
Gilbert, R.J.C., Jiménez, J.L., Chen, S., Andrew, P.W. and Saibil, H.R.
(2000). Structural basis of pore formation by cholesterol-binding
toxins. Int. J. Med. Microbiol. 290, 389–394.
Gilbert, R.J.C. (2002). Pore-forming toxins. Cell. Mol. Life Sci. 59,
832–844.
Heuck, A.P., Hotze, E.M., Tweten, R.K. and Johnson, A.E. (2000).
Mechanism of membrane insertion of a multimeric beta-barrel
protein: perfringolysin O creates a pore using ordered and coupled conformational changes. Mol. Cell 6, 1233–1242.
Heuck, A.P., Tweten, R.K. and Johnson, A.E. (2001) β-barrel poreforming toxins: intriguing dimorphic proteins. Biochemistry 40,
9065–9073.
Heuck, A.P., Tweten, R.K. and Johnson, A.E. (2003). Assembly and
topography of the prepore complex in cholesterol-dependent
cytolysins. J. Biol. Chem. 278, 31218–31225.
Hotze, E.M., Wilson-Kubalek, E.M., Rossjohn, J., Parker, M.W.,
Johnson, A.E. and Tweten, R.K. (2001). Arresting pore formation
of a cholesterol-dependent cytolysin by disulfide trapping synchronizes the insertion of the transmembrane beta-sheet from a
prepore intermediate. J. Biol. Chem. 276, 8261–8268.
Hotze, E.M., Heuck, A.P., Czajkowsky, D.M., Shao, Z., Johnson, A.E.
and Tweten, R.K. (2002). Monomer-monomer interactions
669
drive the prepore to pore conversion of a beta barrel-forming,
cholesterol-dependent cytolysin. J. Biol. Chem. 277, 11597–11605.
Iwamoto, M., Ohno-Iwashita, Y. and Ando, S. (1987). Role of the
essential thiol group in the thiol-activated cytolysin from
Clostridium perfringens. Eur. J. Biochem. 167, 425–430.
Jones, S. and Portnoy, D.A. (1994). Characterization of Listeria monocytogenes pathogenesis in a strain expressing perfringolysin O
instead of listeriolysin O. Infect. Immun. 62, 5608–5613.
Kehoe, M.A., Miller, L., Walker, J.A. and Boulnois, G.J. (1987).
Nucleotide sequence of the streptolysin O (SLO) gene: structural
homologies between SLO and other membrane-damaging, thiolactivated toxins. Infect. Immun. 55, 3228–3232.
Kraulis, P. (1991). MOLSCRIPT: a program to produce both detailed
and schematic plots of proteins. J. Appl. Crystallogr. 24, 946–950.
Macey, M.G., Whiley, R.A., Miller, L. and Nagamune, H. (2001).
Effect on polymorphonuclear cell function of a human-specific
cytotoxin, intermedilysin, expressed by Streptococcus intermedius.
Infect. Immun. 69, 6102–6109.
Nagamune, H., Ohnishi, C., Katsuura, A., Fushitani, K., Whiley, R.A.,
Tsuji, A. and Matsuda, Y. (1996). Intermedilysin, a novel cytotoxin
specific for human cells secreted by Streptococcus intermedius
UNS46 isolated from a human liver abscess. Infect. Immun. 64,
3093–3100.
Nagamune, H., Whiley, R.A., Goto, T., Inai, Y., Maeda, T., Hardie,
J.M. and Kourai, H. (2000). Distribution of the intermedilysin
gene among the anginosus group streptococci and correlation
between intermedilysin production and deep-seated infection
with Streptococcus intermedius. J. Clin. Microbiol. 38, 220–226.
Nakamura, M., Sekino, N., Iwamoto, M. and Ohno-Iwashita, Y.
(1995). Interaction of θ-toxin (perfringolysin O), a cholesterolbinding cytolysin, with liposomal membranes: change in the aromatic side chains upon binding and insertion. Biochemistry 34,
6513–6520.
Olson, R., Nariya, H., Yokota, K., Kamio, Y. and Gouaux, E. (1999).
Crystal structure of Staphylococcal LukF delineates conformational changes accompanying formation of a transmembrane
channel. Nature Struct. Biol. 6, 134–140.
Palmer, M. Vulicevic, I., Saweljew, P., Valeva, A., Kehoe, M. and
Bhakdi. S. (1998). Streptolysin O: a proposed model of allosteric
interaction between a pore-forming protein and its target lipid
bilayer. Biochemistry 37, 2378–2383.
Palmer, M. (2001). The family of thiol-activated, cholesterol-binding
cytolysins. Toxicon 39, 1681–1689.
Parker, M.W., Buckley, J.T., Postma, J.P.M., Tucker, A.D., Leonard, K.,
Pattus, F. and Tsernoglou, D. (1994). Structure of the Aeromonas
toxin proaerolysin in its water-soluble and membrane-channel
states. Nature 367, 292–295.
Pédelacq, J-D., Maveyraud, L., Prévost, G., Baba-Moussa, L.,
González, A., Courcelle, E., Shepard, W., Monteil, H., Samama, JP. and Mourey, L. (1999). The structure of a Staphylococcus aureus
leucocidin component (LukF-PV) reveals the fold of the watersoluble species of a family of transmembrane pore-forming toxins. Structure 7, 277–287.
Petosa, C., Collier, R.J., Klimpel, K.R., Leppla, S.H. and Liddington,
R.C. (1997). Crystal structure of the anthrax toxin protective antigen. Nature 385, 833–838.
Pinkney, M., Beachey, E. and Kehoe, M. (1989). The thiol-activated
toxin streptolysin O does not require a thiol group for cytolytic
activity. Infect. Immun. 57, 2553–2558.
Polekhina, G., Giddings, K.S., Tweten, R.K. and Parker, M.W.
(2004). Crystallization and preliminary x-ray analysis of the
human specific toxin intermedilysin. Acta Crystallogr. D 60,
347–349.
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )
P0884453-Ch37.qxd
670
10/19/05
8:49 PM
Page 670
COMPARATIVE THREE-DIMENSIONAL STRUCTURE OF CHOLESTEROL-DEPENDENT CYTOLYSINS
Polekhina, G., Giddings, K.S., Tweten, R.K. and Parker, M.W.
(2005). Insights into the action of the superfamily of cholesteroldependent cytolysins from studies of intermedilysin. Proc. Natl.
Acad. Sci. USA 102, 600–605.
Ramachandran, R., Heuck, A.P., Tweten, R.K. and Johnson, A.E.
(2002). Structural insights into the membrane-anchoring mechanism of a cholesterol-dependent cytolysin. Nature Struct. Biol. 11,
823–827.
Ramachandran, R., Tweten, R.K. and Johnson, A. E. (2004).
Membrane-dependent conformational changes initiate cholesterol-dependent cytolysin oligomerization and intersubunit betastrand alignment. Nature Struct. Mol. Biol. 11, 697–705
Rood, J.I. (1998) Virulence genes of Clostridium perfringens. Annu. Rev.
Microbiol. 52, 333–360.
Rossjohn, J., Feil, S.C., McKinstry, W.J., Tweten, R.K. and Parker,
M.W. (1997). Structure of a cholesterol-binding, thiol-activated
cytolysin and a model of its membrane form. Cell 89, 685–692.
Saitou, N. and Nei, M. (1987). The neighbor-joining method: a new
method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4,
406–425.
Saunders, F.K., Mitchell, T.J., Walker, J.A., Andrew, P.W. and
Boulnois, G.J. (1989). Pneumolysin, the thiol-activated toxin of
Streptococcus pneumoniae, does not require a thiol group for in vitro
activity. Infect. Immun. 57, 2547–2552.
Shatursky, O., Heuck, A.P., Shepard, L.A., Rossjohn, J., Parker, M.W.,
Johnson, A.E. and Tweten, R.K. (1999). The mechanism of membrane insertion for a cholesterol-dependent cytolysin: a novel
paradigm for pore-forming toxins. Cell 99, 293–299.
Sekino-Suzuki, N., Nakamura, M., Mitsui, K.I. and Ohno-Iwashita,
Y. (1996). Contribution of individual tryptophan residues to the
structure and activity of θ-toxin (perfringolysin-O), a cholesterolbinding cytolysin. Eur. J. Biochem. 241, 941–947.
Shepard, L.A., Heuck, A.P., Hamman, B.D., Rossjohn, J., Parker,
M.W., Ryan, K.R., Johnson, A.E. and Tweten, R.K. (1998).
Identification of a membrane-spanning domain of the thiolactivated, pore-forming toxin Clostridium perfringens perfringolysin O: an alpha-helical to beta-sheet transition identified
by fluorescence spectroscopy. Biochemistry 37, 14563–14574.
Shepard, L.A., Shatursky, O., Johnson, A.E. and Tweten, R.K. (2000).
The mechanism of assembly and insertion of the membrane complex of the cholesterol-dependent cytolysin perfringolysin O:
Formation of a large prepore complex. Biochemistry 39, 10284–10293.
Song, L. Hobaugh, M.R., Shustak, C., Cheley, S., Bayley, H. and
Gouaux, J.E. (1996). Structure of Staphylococcal α-hemolysin, a
heptameric transmembrane pore. Science 274, 1859–1866.
Tweten, R.K. (1988) Nucleotide sequence of the gene for perfringolysin O (theta toxin) from Clostridium perfringens has significant homology with the genes for streptolysin O and
pneumolysin. Infect. Immun. 56, 3235–3240.
Tweten, R.K., Parker, M.W. and Johnson, A.E. (2001). The cholesteroldependent cytolysins. Curr. Top. Microbiol. Immunol. 257, 15–33.
Walker, J.A., Allen, R.L., Falmagne, P., Johnson, MK. and Boulnois,
G.J. (1987). Molecular cloning, characterization, and complete
nucleotide sequence of the gene for pneumolysin, the sulfhydrylactivated toxin of Streptococcus pneumoniae. Infect. Immun. 55,
1184–1189.
I I I . TO X I N S A C T I N G O N T H E S U R FA C E O F TA R G E T C E L L S ( E X C E P T S U P E R A N T I G E N S )