A Structural Model for Binding of the Serine-Rich Repeat Adhesin GspB to Host Carbohydrate Receptors
et al. (2011) A Structural Model for Binding of the Serine-Rich Repeat Adhesin GspB to
Host Carbohydrate Receptors. PLoS Pathog 7(7): e1002112. doi:10.1371/journal.ppat.1002112
A Structural Model for Binding of the Serine-Rich Repeat Adhesin GspB to Host Carbohydrate Receptors
Tasia M. Pyburn 0 1
Barbara A. Bensing 0 1
Yan Q. Xiong 0 1
Bruce J. Melancon 0 1
Thomas M. Tomasiak 0 1
Nicholas J. Ward 0 1
Victoria Yankovskaya 0 1
Kevin M. Oliver 0 1
Gary Cecchini 0 1
Gary A. Sulikowski 0 1
Matthew J. Tyska 0 1
Paul M. Sullam 0 1
T. M. Iverson 0 1
Partho Ghosh, University of California San Diego, United States of America
0 a Current address: Vanderbilt Program in Drug Discovery, Department of Pharmacology, Nashville, Tennessee, United States of America b Current address: Molecular Structure Group, University of California , San Francisco, California , United States of America
1 1 Department of Pharmacology, Vanderbilt University Medical Center, Nashville, Tennessee, United States of America, 2 Vanderbilt Institute of Chemical Biology, Nashville, Tennessee, United States of America, 3 Department of Medicine, Veterans Affairs Medical Center and the University of California, San Francisco, California, United States of America, 4 Department of Medicine, Harbor-UCLA Medical Center, Torrance, California, United States of America, 5 Department of Chemistry, Vanderbilt University , Nashville , Tennessee, United States of America, 6 Molecular Biology Division, Veterans Affairs Medical Center , San Francisco , California, United States of America, 7 Department of Biochemistry & Biophysics University of California, San Francisco, California, United States of America, 8 Department of Cell and Developmental Biology, Vanderbilt University Medical Center, Nashville, Tennessee, United States of America, 9 Department of Biochemistry, Vanderbilt University Medical Center , Nashville, Tennessee , United States of America
GspB is a serine-rich repeat (SRR) adhesin of Streptococcus gordonii that mediates binding of this organism to human platelets via its interaction with sialyl-T antigen on the receptor GPIba. This interaction appears to be a major virulence determinant in the pathogenesis of infective endocarditis. To address the mechanism by which GspB recognizes its carbohydrate ligand, we determined the high-resolution x-ray crystal structure of the GspB binding region (GspBBR), both alone and in complex with a disaccharide precursor to sialyl-T antigen. Analysis of the GspBBR structure revealed that it is comprised of three independently folded subdomains or modules: 1) an Ig-fold resembling a CnaA domain from prokaryotic pathogens; 2) a second Ig-fold resembling the binding region of mammalian Siglecs; 3) a subdomain of unique fold. The disaccharide was found to bind in a pocket within the Siglec subdomain, but at a site distinct from that observed in mammalian Siglecs. Confirming the biological relevance of this binding pocket, we produced three isogenic variants of S. gordonii, each containing a single point mutation of a residue lining this binding pocket. These variants have reduced binding to carbohydrates of GPIba. Further examination of purified GspBBR-R484E showed reduced binding to sialyl-T antigen while S. gordonii harboring this mutation did not efficiently bind platelets and showed a significant reduction in virulence, as measured by an animal model of endocarditis. Analysis of other SRR proteins revealed that the predicted binding regions of these adhesins also had a modular organization, with those known to bind carbohydrate receptors having modules homologous to the Siglec and Unique subdomains of GspBBR. This suggests that the binding specificity of the SRR family of adhesins is determined by the type and organization of discrete modules within the binding domains, which may affect the tropism of organisms for different tissues.
Funding: This work was supported by the American Heart Association grants 09GRNT2180065 (YQX), 09GRNT2310188 (MJT), and 09GRNT2220122 (TMI), Pilot
Project funds from the Vanderbilt Institute of Chemical Biology (TMI and GAS), Pilot funds from the VICTR CTSA UL1 RR024975 from NCRR/NIH (TMI), Vanderbilt
University IDEAs award (MJT), the Department of Veterans Affairs (PMS and GC), and grants GM079419 (TMI), AI079558 (TMI), AI041513 (PMS), AI057433 (PMS),
DK075555 (MJT) and GM61606 (GC) from the National Institutes of Health. TMT and KMO were supported by T32 GM65086 (Chemical-Biology Interface Training
Grant). BJM was supported by T90 DA022873 (Integrative Training in Therapeutic Discovery Training Grant). Portions of this research were carried out at
beamlines 9-2, 11-1, and 12-2 at Stanford Synchrotron Radiation Lightsource (SSRL), a national user facility operated by Stanford University on behalf of the U.S.
Department of Energy, Office of Basic Energy Sciences. The SSRL Structural Molecular Biology Program is supported by the Department of Energy, Office of
Biological and Environmental Research, and by the National Institutes of Health, National Center for Research Resources, Biomedical Technology Program, and the
National Institute of General Medical Sciences. The Life Sciences Collaborative Access Team (LS-CAT) ID21-D, ID21-F, ID21-G at Advanced Photon Source (APS). Use
of the Advanced Photon Source was supported by the U. S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No.
DE-AC0206CH11357. Use of the LS-CAT Sector 21 was supported by the Michigan Economic Development Corporation and the Michigan Technology Tri-Corridor for the
support of this research program (Grant 085P1000817).
Competing Interests: The authors have declared that no competing interests exist.
The serine-rich repeat (SRR) glycoproteins of Gram-positive
bacteria are an expanding family of microbial adhesins and
virulence factors . These proteins consist of a distinctive
signal sequence and export-targeting region at the N-terminus, a
short SRR region (,50170 amino acids), a ligand binding region,
a second lengthy SRR region (,4004000 amino acids), and a cell
wall anchoring motif at the C-terminus (Fig. 1) [1,2,4,69]. The
binding regions of the SRR glycoproteins contain significant
sequence variation, which appears to account for their broad
range of binding targets, including platelet membrane and salivary
The binding of bacteria to human platelets is thought to
be important for development of infective endocarditis, a
life-threatening infection of the cardiovascular system.
Streptococcus gordonii is a leading cause of endocarditis.
This pathogen uses a protein called GspB to attach to
carbohydrates on human platelets. While this binding
interaction appears to be mediated by a specific,
contiguous domain within GspB, little is known about
the molecular details of the interaction between GspB and
the carbohydrate receptors on its human host. We
therefore determined the crystal structure of the region
of GspB that binds to platelet carbohydrates, both alone
and in complex with a synthetic carbohydrate receptor.
Using this structure as a guide, we were able to produce
three strains of S. gordonii that lacked the ability to bind to
platelet carbohydrates. One of these isogenic variants was
studied more in-depth and lacked the ability to bind to
human platelets in vitro and was reduced in virulence
when tested in vivo. These studies provide the first
structural information detailing the molecular interactions
between any serine-rich repeat adhesin and its host
receptor, and identify how different, related adhesins
may have evolved different specificities for host receptors.
glycoproteins [1,4,911], endothelial cells , epithelial cells
, erythrocytes [14,15], and keratins [16,17]. Expression of
SRR proteins has been associated with increased virulence in
several animal models of infection, including endocarditis [6,18],
meningitis , pneumonia , and bacteremia [7,19].
A number of bacterial surface components have been shown to
mediate platelet binding in vitro, either by interacting directly with
receptors on the platelet membrane, or via bridging molecules,
such as fibrinogen . The contribution of these interactions
to virulence, however, has been assessed for relatively few of these
adhesins. Previous studies have focused on the molecular basis for
the SRR adhesin mediated binding of gram-positive bacteria to
human platelets, and the role of this process in the pathogenesis of
infective endocarditis. This interaction appears to be important for
the attachment of blood-borne bacteria to platelets on the surface
of damaged cardiac valves, thereby initiating infection. In
addition, the subsequent deposition of platelets onto the infected
endocardium may be due in part to bacterium-platelet binding,
resulting in the formation macroscopic vegetations, which are the
hallmark lesions of this disease . Three of the SRR proteins
(GspB of Streptococcus gordonii strain M99, Hsa of S. gordonii strain
Challis, and SrpA of Streptococcus sanguinis strain SK36) bind human
platelets through their interaction with glycocalicin, which is the
carbohydrate-rich extracellular portion of the platelet membrane
glycoprotein GPIba [9,11]. While the specific carbohydrate
receptor for SrpA has not yet been identified, GspB and Hsa
recognize sialylated trisaccharides [1,11,31,32]. Dot blot assays
using immobilized carbohydrates have demonstrated that GspB
has high fidelity for sialyl-T antigen (i.e. NeuAca(23)Galb(1
3)GalNAc), one of the major carbohydrates on GPIba , while
Hsa binds to glycocalicin via either sialyl-T antigen or sialyllactose
(Neu5Aca(26)Galb(14)Glc) [31,32]. Binding of these SRR
adhesins to platelets is a high affinity process, with the interaction
between GspB and glycocalicin having a KD of 2.3861028 M 
and appears to be a major factor in the pathogenesis of infective
endocarditis, since the loss of GspB or Hsa expression results in a
marked reduction in virulence .
Structural information can enhance the understanding of the
determinants of binding specificity. Here, we report the
highresolution crystal structure of GspBBR, both alone and in complex
with the disaccharide a-2,3-sialyl (1-thioethyl)galactose, a
precursor to synthetically-produced sialyl-T antigen. From these
structures, we identified that a subdomain of GspBBR resembling
mammalian sialic acid binding proteins binds to a-2,3-sialyl
(1thioethyl)galactose. Site directed mutagenesis and in vivo studies in
a rat model of infective endocarditis verified that this carbohydrate
binding site within GspBBR mediates binding of S. gordonii strain
M99 to sialyl-T antigen, the host carbohydrates of the platelet
Figure 1. Schematic of the SRR adhesins. Selected SRR adhesins are aligned based on the N-terminus of their binding regions, with the largest
binding region at the top and the smallest binding region at the bottom. The length of each domain and the entire protein is drawn to-scale.
SP = signal peptide (dark gray), SRR1 = first serine rich repeat (light gray), BR = binding region (red), SRR2 = second serine rich repeat (dark gray),
CWA = cell wall anchoring motif (light gray). The SRR adhesins used in this figure are S. aureus SraP (SraP); S. epidermidis SRR1 (seSRR), S. agalactiae
SRR1 (SRR1GBS), S. gordonii strain M99 GspB (GspB), S. parasanguinis Fap1 (Fap1), S. pneumoniae PsrP (PsrP), S. gordonii strain Challis Hsa (Hsa), and S.
sanguinis SrpA (SrpA).
membrane receptor GPIba, and intact platelets, and that this
interaction is important for virulence. Our analysis of the structure
further identified that GspBBR contains an unusually modular fold,
prompting us to re-analyze the sequences of the SRR family of
adhesins. We determined that other structurally uncharacterized
SRR adhesins also contain modules within their binding regions,
suggesting that particular subdomains may be included, removed,
or interchanged, manifesting in the broad range of binding
partners observed in the family.
Overall structure of GspBBR
We determined the structure of GspBBR to 1.4 A resolution
using the method of Multiwavelength Anomalous Dispersion from
a single Dy3+ derivative (Fig. 2, Table 1, Table 2). GspBBR folds
into an elongated rod, with dimensions of ,130 A630 A630 A.
This rod is comprised of three apparently independently-folded
subdomains arranged in a linear fashion, like beads on a string.
The secondary structure of each of the three subdomains is
predominated by b-strands. Interestingly, the first two subdomains
are organized around core folds that resemble those found within
the eukaryotic immunoglobulin (Ig) superfamily (Fig. 3). Ig-folds
have previously been identified in prokaryotic proteins ,
and it has been noted that some of these bacterial proteins with
Igfolding topologies contain sequence similarity to their eukaryotic
counterparts, while the others lack the residues conserved in the
core of eukaryotic proteins with Ig-folds. GspBBR has homology
with bacterial proteins that do not contain detectable sequence
similarity to eukaryotic proteins with Ig-folds, lacking even the
cysteines that normally form a disulfide bond.
Several topology variants of Ig-folds have been characterized. A
structural homology search using the EMBL DaliLite server 
identified that the N-terminal subdomain of GspBBR contains a
folding topology, strand inserts, and inter-sheet angle reminiscent of
the DE-variant of the Ig-fold  (Fig. 3AC). This Ig-fold topology
is found in prokaryotic proteins and was first identified within the
Aregion of the Staphylococcus aureus CNA protein . CNA belongs to
the family of microbial surface components recognizing adhesive
matrix molecules (MSCRAMMs) of Gram-positive pathogens [34
36]. Given its structural similarity, this N-terminal subdomain of
GspBBR will be termed the CnaA subdomain. The highest structural
similarity between the CnaA subdomain of GspBBR and any other
structurally characterized protein is to the C-terminal subdomain of
the binding region of SRR adhesin Fap1 from Streptococcus
parasanguinis (Fap1NR-b) . The RMS deviation of the structural
alignment between the CnaA subdomain of GspBBR and the
Fap1NR-b subdomain is 2.5 A for 200 Ca atoms.
Figure 2. Structure of GspBBR. A. Two views of the structure of GspBBR separated by a 90u rotation. The N-terminus of the binding region is
located at the top of the figure and the C-terminus is located at the bottom. b-Strands are colored magenta, a-helices are colored blue, and loops are
colored black. A cation bound within the Siglec domain is shown as a purple sphere. B. Representative experimental electron density (blue mesh) at
2.0 A resolution, contoured at 1.0 s, and depicted superpositioned onto residues 515520 of the final model. C. A m|Fo|2d|Fc| omit electron density
map at 1.4 A resolution (green mesh) calculated in REFMAC5  after the removal of the model contoured at 3.0 s and depicted superpositioned
onto residues 515520 of the final model. D. Folding diagram of the Unique subdomain. b-Strands are colored magenta and the a-helix is colored
35.5, 98.5, 99.1 34.5, 98.5, 99.1 34.5, 98.5, 99.0
No. of protein atoms 352
33.7, 96.8, 100.2 33.3, 86.7, 117.9 69.9, 34.0, 83.4
Surprisingly, a structural homology search using EMBL
DaliLite  identified that the second subdomain contained a
topology and strand inserts reminiscent of the V-set Ig fold
adopted by eukaryotic sialic acid binding immunoglobulin-like
lectins  (Siglecs; Fig. 3DF). However, the C9 and C0
strands normally inserted into the V-set Ig-fold are instead
replaced by a long loop inserted at the same location (Fig. 3D
F). The RMS deviation of the structural alignment between the
second subdomain of GspBBR and Siglec-5 is 2.9 A for 210 Ca
atoms despite only 6% sequence identity. Like GspB, Siglecs bind
carbohydrate receptors. Accordingly, the second domain is termed
the Siglec subdomain. This subdomain of GspBBR contained
electron density consistent with a cation-binding site (see
Supporting Text S1, Fig. S1, Supporting Protocol S1).
This was rather unexpected since Siglecs themselves do not bind
cations in the carbohydrate-binding domain. The
seven-coordinate number suggests that under physiological conditions, this
should be a Ca2+ binding site. We explored the role of Ca2+ and
other cations for the binding of S. gordonii strain M99 to
glycocalicin. However, metal depletion, metal substitution, and
site directed mutagenesis of the residues coordinating the ion were
all consistent with the cation not being essential for carbohydrate
receptor recognition (data not shown).
While still predominated by b-strands, the C-terminal subdomain
of GspBBR does not contain an Ig-fold. In fact, a structural homology
No. of unique reflections 23918
search using DaliLite  did not identify any proteins of known
structure with significant similarity to this C-terminal subdomain
(Fig. 2D). As a result, it will be termed the Unique subdomain.
Identification of the receptor binding site
We sought to identify the details of the interaction between
GspBBR and the host receptor sialyl-T antigen. Free sialyl-T
antigen is a rare reagent that is not commercially available and is
challenging to synthesize. Therefore, we developed a 4-step
synthesis for a-2,3-sialyl (1-thioethyl)galactose
(NeuAca(23)(1CH3CH2S)Galb) (see Supporting Protocol S1; Fig. S2) which
is the disaccharide truncation of and a synthetic precursor to
sialylT antigen (NeuAca(23)Galb(13)GalNAc). To confirm that
a2,3-sialyl (1-thioethyl)galactose binds to GspBBR, we assessed the
ability of this disaccharide to inhibit the binding of S. gordonii to
glycocalicin. In the presence of 44 mM a-2,3-sialyl
(1-thioethyl)galactose, binding of S. gordonii to glycocalicin was reduced by 90%
(Fig. 4A). Given the structural similarity to the native host
receptor, this strongly suggests that a-2,3-sialyl
(1-thioethyl)galactose competes directly for the sialyl-T antigen binding site.
We determined the co-crystal structure of a-2,3-sialyl
(1thioethyl)galactose in complex with GspBBR. New electron density
consistent with bound disaccharide (Fig. 4B) was apparent within
a defined pocket (Fig. 4C) located at the N-terminus of the C
strand and C-terminus of the F strand in the Siglec domain
(Fig. 3F) of the Siglec subdomain (Fig. 4D). While the similarity
of the fold of the Siglec subdomain to mammalian Siglecs might
predict a similar binding pocket, both the overall location of the
a2,3-sialyl (1-thioethyl)galactose binding site on the domain and the
specific contacts between GspBBR to the carbohydrate differ from
that of structurally characterized mammalian Siglecs 
(Fig. 5). In fact, in GspBBR a helix is positioned in the location
where carbohydrates bind to Siglecs, precluding the use of a
structurally similar binding site. Instead, the local secondary
structure surrounding the a-2,3-sialyl (1-thioethyl)galactose
binding site resembles a b-grasp domain, which is a motif that
commonly binds the a-2,3-linkage of sialic acid based multivalent
Figure 3. Ig-fold topologies. A. Topology diagram of a C-set Ig-fold. B. Topology diagram of the DE-variant of a C-set Ig-fold identified in
MSCRAMMs . C. Topology diagram of the GspBBR CnaA subdomain. D. Topology diagram of a V-set Ig-fold. E. Topology diagram of a Siglec 
showing the binding location of the sialylated carbohydrate in pink. F. Topology diagram of the GspBBR Siglec subdomain showing the binding
location of the sialyl-T antigen in pink.
carbohydrates . In GspBBR, the backbone carbonyls of
Ala506 and Arg449, the Ng1 and Ng2 of Arg484, the backbone
amide nitrogen of Thr483 and the side chain hydroxyl of Tyr443
make direct contacts to the a-2,3-sialyl (1-thioethyl)galactose
disaccharide, while the side chain hydroxyl of Tyr485 makes
water-mediated contacts (Fig. 4B, Table 3).
A comparison of the structure of GspBBR determined with and
without the disaccharide did not show significant structural changes
localized within the binding pocket, which suggests that it is
preformed. Thus, while the physiologically-relevant receptor
trisaccharide sialyl-T antigen is a rare reagent, we can qualitatively
suggest its binding location by modeling the third carbohydrate onto
a-2,3-sialyl (1-thioethyl)galactose. In our model, all three
carbohydrates of the host receptor fit into a pre-formed binding pocket on
the Siglec subdomain, with the third carbohydrate of sialyl-T
antigen extending toward the Unique subdomain (Fig. S3).
While disaccharide binding to GspBBR did not appear to induce
obvious changes in conformations of side chains within the binding
pocket, the interdomain angle between the CnaA and Siglec
subdomains unexpectedly straightened by 40u (Fig. 6; Video S1).
The change in orientation between the CnaA and Siglec
subdomains involves rotation around a single hinge consisting of
residues Lys398-Asp399-Thr400, where the side chain Asp399 is
one of the ion coordinating residues.
Functional relevance of the carbohydrate binding site
To confirm that this crystallographically-identified a-2,3-sialyl
(1-thioethyl)galactose binding site is indeed important for
GspBmediated binding to GPIba, we generated four isogenic variants of
strain M99 with point mutations within the Siglec subdomain of
GspB (Table 4). Three of these variants contained mutations
within the crystallographically-identified binding site (Y443F,
R484E and Y485F), while a control mutation (E401A) was located
within the Siglec domain, but away from the receptor binding site.
Importantly, none of the point mutations affected surface
expression of GspB (Fig. 7A). Each of these isogenic variants
had an 86% or greater decrease in binding to GPIba in vitro, as
compared with the parent strain (p,0.001) (Fig. 7B). This
decrease is comparable to that of the gspB null mutant. In contrast,
binding by the E401A variant was not significantly different from
that of M99 (p = 0.44).
We then selected the R484E mutation for more detailed study.
The purified GST-GspBBR fusion protein harboring the R484E
substitution exhibited a marked decrease in binding to biotinylated
sialyl-T antigen (Fig. 7C). The R484E substitution also resulted in
reduced binding to human platelets in vitro as assessed by
quantifying the amount of input inoculum bound to fixed platelets
(Fig. 7D). We next visualized platelets by Differential Interference
Contrast (DIC) and fluorescence microscopy and quantified the
number of platelets with surface-bound bacteria. In the presence
of the DAPI-labeled PS846 (M99DgspB) or the PS2116 (M99
gspbR484E) strain, substantially fewer platelets had bacteria bound
as compared to platelets in the presence of wild-type S. gordonii
strain M99 (Fig. 8, Table 5). It should be noted that while the
platelet numbers were normalized when placed onto the cover slip,
during the experiment the wild type bacteria seemed to form
microscopic aggregates with the platelets (data not shown) but this
was not observed when platelets were mixed with strains PS2116
or PS846. This confirms the importance of this residue for
Impact of Siglec-mediated binding on virulence
It has previously been demonstrated that loss of GspB
expression results in a marked decrease in virulence, as measured
by a rat model of infective endocarditis . To assess whether
binding via the Siglec subdomain to host carbohydrate receptors
contributes to virulence, we examined the impact of the R484E
substitution on the ability of S. gordonii strain M99 to produce
endocardial infection. We first compared the relative virulence of
M99 with strain PS2116, (M99 gspbR484E). Catheterized rats were
simultaneously infected intravenously with 261025 CFU of M99
and PS2116. After 72 h, the animals were sacrificed and the
relative levels of bacteria within tissues determined. Animals
coinfected with M99 and PS2161 (which carries the spec resistance
marker just upstream of gspB) served as controls. When assessed at
the above time-point, animals co-infected with M99 and PS2116
had significantly reduced densities of the mutant strain within
vegetations (mean 6 S.D. = 6.8161.70 log10 CU/g. veg.) as
compared with parental strain M99 (7.4761.69 log10 CFU/g. veg;
P,0.02). Loss of sialyl-T antigen binding was also associated with
significantly reduced bacterial densities within kidneys and spleens
(P,0.02 and P,0.001, respectively; Table 6). In parallel studies,
no differences were observed between the M99 and the control
strain (PS2161), as measured by CFU per gram of tissue within
vegetations, kidneys, or spleens (data not shown). We also analyzed
these findings by calculating a competition index, in which the
ratio of M99 and PS2116 within tissues was normalized for the
CFU of each strain within the inoculum. When analyzed by this
approach, the densities of the GspB mutant strain PS2116
remained significantly reduced in all tissues, as compared with
M99 (P,0.002), while the densities of the control strain were
We then compared the relative virulence of PS2116 with PS846
(M99DgspB). Infective endocarditis was produced as above, using
an inoculum containing the two strains in a 1:1 ratio. When
assessed at 72 h post-infection, PS2116 and PS846 had similar
densities of organisms within all tissues (Table 6). Of note, the
levels of both strains with vegetations were markedly lower than
those achieved by wild-type M99 in the above studies. Thus, these
two strains appeared comparably attenuated in the setting of
endocarditis, indicating that platelet binding by the Siglec domain
may be the predominant GspB interaction contributing to
virulence in endocarditis.
The binding regions of SRR adhesins have a modular
While overall sequence analysis of members of the SRR family
has identified unifying sequence trends (Fig. 1), sequence
comparisons of the binding regions have shown little homology.
A structural comparison between the three distinct subdomains of
GspBBR (Fig. 2A) and the two distinct subdomains of the SRR
adhesin Fap1  immediately indicates that these family
members adopt unrelated folds in their binding domains. Like
GspBBR, Fap1NR appears to be composed of independently-folded
subdomains; however its binding region only contains two
modules whereas the binding region of GspBBR contains three.
Intriguingly, Fap1NR contains a helical subdomain at its
Nterminus (Fap1NR-a) that does not resemble any of the subdomains
of GspBBR, and a CnaA subdomain at its C-terminus (Fap1NR-b)
that resembles the N-terminal CnaA subdomain of GspBBR
(GspBBR-C). This suggested to us that members of the SRR family
could undergo reorganization of the modules within their host
binding regions, with particular modules or combinations of
modules conferring specific properties.
Accordingly, we re-analyzed the binding regions of selected
members of the family using a new strategy, where we used
BLAST  and ClustalW  to query the sequences of the
binding regions of SRR adhesins with input sequences
corresponding to subdomains of either GspBBR or Fap1NR, or of short
regions (,200 amino acids) of sequences of structurally
uncharacterized binding regions of SRR adhesins. These modified
sequence comparisons strongly suggest that the binding regions
of members of the SRR family have indeed evolved to contain
modules (Fig. 9), and show several distinct groupings. For
example, all three assessed SRR adhesins with carbohydrate
binding partners contain the Siglec and Unique subdomains in
tandem, strongly suggesting that the inclusion of these modules
within the binding region confers lectin-like binding
These focused sequence alignments additionally identify that a
CnaA subdomain may be common in SRR adhesins, appearing in
three of the eight binding regions that we assessed. As opposed to
the combination of the Siglec and Unique subdomains, which are
associated with carbohydrate binding, each adhesin containing a
CnaA subdomain may have a different binding specificity (Fig. 9).
Interestingly, in the SRR adhesins that were assessed, the CnaA
subdomains were always found paired with sequence for another
It is important to note that several of the binding regions of the
SRR adhesins do not have significant sequence identity to any
protein of known structure. For example, the binding region of S.
agalactiae SRR1 (SRR1GBS) likely contains two subdomains,
SRR1GBS-a at the N-terminus, which does not contain detectable
sequence similarity to any protein of known fold, and SRR1GBS-b
at the C-terminus, which has sequence similarity to MSCRAMMs
and likely adopts a similar folding topology to CnaA. Likewise, no
part of the binding regions of S. pneumoniae PsrP, S. aureus SraP, or
S. epidermidis seSRRBR exhibits detectable sequence similarity to
any known protein, but the latter two have high sequence
similarity to each other.
SRR adhesins have been identified in a variety of Gram-positive
pathogens, and have been implicated as virulence factors in a wide
spectrum of infections [5,6,12,18]. The diversity of these infections
(e.g., endocarditis, meningitis, pneumonia) and the broad scope of
their anatomic locations are consistent with the binding regions of
the SRR proteins differing considerably in their selectivity. Indeed,
although only a few ligands for this family of proteins have been
identified, they range from carbohydrates (such as sialyl-T antigen)
 to proteins (such as keratins) [16,17]. To date, however, the
structural basis for this selectivity has been unknown.
In several members of the family (e.g. the SRR adhesins of S.
gordonii and S. pneumoniae) the domain between the two SRR
regions mediates binding. However, for many of these SRR
adhesins (i.e. SraP of S. aureus), the host receptor has yet to be
identified [6,43], and the delineation of the binding region is
assumed based upon sequence comparisons within the family.
Among the best characterized of the SRR adhesins is GspB from
S. gordonii, which has demonstrated binding affinity for the sialyl-T
antigen carbohydrate decorating platelet glycoprotein GPIba. Our
crystal structure of the binding region, GspBBR, identified a
modular organization with three subdomains (Fig. 2A), two of
which are organized around Ig-folds. Proteins containing Ig-folds
are commonly found within the mammalian immune system,
where they exhibit a variety of functions; however, Ig-folds are not
uncommon within pathogens, where they act exclusively as
virulence factors. The first characterized bacterial protein to
contain an Ig-fold was PapD, a chaperone for the assembly of pili
in E. coli . The Ig-fold has been observed in several other types
of virulence factors, including components of pili
(carbohydratebinding fimbriae) [41,52], and in adhesins of varying specificity for
host receptors, including MSCRAMMs  and invasins [38
Co-crystallization of GspBBR with a-2,3-sialyl
(1-thioethyl)galactose identified a specific binding site for this disaccharide within
the Siglec subdomain (Fig. 4BD), and the introduction of single
point mutations within this region significantly reduced levels of
binding of S. gordonii to glycocalicin. Moreover, the R484E
substitution in GspB reduced platelet binding by M99 (Fig. 7D,
Fig. 8, Table 5), and had a marked reduction in binding of the
GST-GspBBR fusion protein to sialyl-T antigen (Fig. 7C),
confirming the importance of the Siglec subdomain for
carbohydrate binding. These results indicate that this binding pocket
within the Siglec subdomain is required for binding to sialyl-T
antigen, and that the interaction of this GspB subdomain with
sialyl-T antigen is important for binding to platelets.
When examined in a co-infection model of infective
endocarditis, the isogenic variant of M99 expressing GspB-R484E (strain
PS2116) was also significantly reduced in virulence (most notably a
78% reduction in bacterial levels within vegetations as compared
with the parent strain). When PS2116 was tested with PS846
(M99DgspB) in this model, the strains had comparable but reduced
densities within target tissues (as compared with WT M99),
indicating that they were similarly reduced in virulence. These
findings indicate that the predominant property of GspB
contributing to the pathogenesis of endocarditis is its interaction
with sialyl-T antigen. In previous in vivo studies, co-infection with
M99 and PS846 yielded more pronounced differences between the
strains, as measured by densities of organisms within target tissues
. This may reflect subtle experimental differences, such as
inoculum size, but could also be due to residual binding activity of
the mutated GspB. Of note, PS2116 had low but detectable levels
of platelet binding in vitro. Similarly, the GST-GspBBR-R484E had
measurable levels of binding to sialyl-T antigen, as compared to
GST alone. This residual binding seen in vitro, presumably due to
the other key residues identified by crystallography, may account
for the residual virulence of the PS2116 strain. Alternatively,
although ligands for GspB other than sialyl-T antigen have not
been identified, it is conceivable that GspB has other interactions
S. gordonii strain M99 wild-type
S. gordonii strain M99 gspBR484E
S. gordonii strain M99DgspB
59gspB::spec control strain
control residue not located near the binding pocket. C. Binding of
sialylT antigen to purified GST-GspBBR. The indicated amounts of GST or
GSTGspBBR wild-type (wt) and variant (R484E) proteins were immobilized in
microtiter wells, and the binding of biotinylated sialyl-T antigen to each
protein was detected by using peroxidase-conjugated streptavidin
along with a chromogenic peroxidase substrate. Binding is expressed as
the mean 6 standard deviation (n = 3). The o represents purified
GSTGspBBR-R484E, the x represents purified GST, and the represents
purified GST-GspBBR. D. Binding of M99 and derivative strains to human
platelets. Binding is expressed as the percent of input bacteria that
remained bound to platelets after repeated washing of the wells (mean
6 standard deviation).
in vivo that contribute to virulence in this setting, beyond those
mediating platelet binding.
Upon binding of the a-2,3-sialyl (1-thioethyl)galactose to
GspBBR, a 40u interdomain angle straightening is observed
between the CnaA and Siglec subdomains (Fig. 6, Video S1).
The interdomain hinge is centered at the ion binding site, making
the interdomain angle change reminiscent of calcium dependent
interdomain angle changes observed in C-cadherin, which also
contains modules of Ig-fold topology . There are several
possibilities for the origin and role of this interdomain
reorganization in GspBBR. It is first possible that GspBBR exhibits natural
flexibility between these two domains, and that the difference in
crystal contacts artifactually resulted in the straighter form of the
protein being trapped in the co-structure of GspBBR with
a-2,3sialyl (1-thioethyl)galactose. Intriguingly, however, an interdomain
straightening is also observed in small angle x-ray scattering
(SAXS) of the binding region of SRR protein Fap1NR upon
lowering the pH to 5.0, where Fap1NR has highest affinity for its
host receptor . Those experiments, performed in solution, are
not limited by crystal contacts and suggest that interdomain
straightening upon ligand binding could be a conserved feature of
members of the SRR family. Physiological roles of interdomain
straightening upon ligand binding include improved
hydrodynamics in the cardiovascular system or oral cavity. The elongation
upon ligand binding is perhaps more compellingly reminiscent of
the interdomain straightening that occurs upon the binding of
Pselectin and b3-integrin to their respective ligands . These
proteins have been suggested to form catch bonds in order to
have increased affinity to their ligand upon the application of
tensile force. Each of these possibilities for the observed
straightening of GspBBR upon ligand binding is currently under
While both Siglecs and GspBBR bind to carbohydrate receptors,
the structural details of disaccharide binding to this novel bacterial
Siglec bear little resemblance to the binding of sialic acid moieties
to characterized mammalian Siglecs  (Fig. 5). Indeed,
comparison of the binding sites reveals that the secondary
structure of GspBBR has a helix that is found at the same location
where 39 sialyllactose binds to Siglec-5 (Fig. 5A), thus eliminating
the possibility of analogous carbohydrate binding. Instead,
GspBBR appears to use a b-grasp domain to form the specific
host receptor binding site, a domain found in another group of
sialic acid binding proteins, staphylococcal superantigen-like
proteins (SSLs) that forms a V-like cleft for carbohydrate binding
. Nevertheless, both mammalian Siglecs and GspBBR have
binding pockets predominated by tyrosines and arginines. With a
closer look at each binding site, in Siglecs, the sialic acid moiety of
the binding partner makes a salt bridge with an arginine residue.
This salt bridge is not present in the binding site of GspBBR despite
the presence of R484 and its key role in binding; this arginine
instead binds to the C6 hydroxyl and the pyranose oxygen of the
1-thioethylgalactose (Table 3).
Importantly, the structure of GspBBR provides insight into the
binding repertoire of other SRR proteins, and specifically
highlights which modules mediate binding to host carbohydrate
receptors. Homologues of GspB containing regions with high
sequence similarity to the Siglec and Unique subdomains (such as
S. gordonii strain Challis Hsa and S. sanguinis SrpA) are also
demonstrated lectins. Supporting this sequence analysis, our
cocrystal structure of GspBBR with a-2,3-sialyl (1-thioethyl)galactose
demonstrates that the first two carbohydrates of host receptor bind
within a pre-formed pocket on the Siglec subdomain. Using this
experimental co-structure as a starting point, we could model
binding of the sialyl-T antigen. This model is consistent with the
third carbohydrate of this trisaccharide binding to an extended
lobe on this pocket. This region of the binding pocket is still within
the Siglec subdomain but extends toward the Unique subdomain
The structures of both GspBBR and Fap1NR identified a
CnaA-like subdomain within the binding region, and our
sequence analysis (Fig. 9) additionally predicts a CnaA
subdomain at the C-terminus of the binding region of S.
agalactiae SRR1 (SRR1GBS-b), an adhesin for keratin 4 . As
opposed to the Siglec subdomain, which contains an identifiable
receptor binding site, the precise function of the CnaA
subdomain is less clear. MSCRAMMs, such as CnaA from S.
aureus, recognize peptides through the formation of a binding site
between two Ig-like domains (Fig. S4). The bound peptide forms
an additional strand that becomes a part of the Ig-fold. In
contrast, GspBBR, Fap1NR, and SRR1GBS each apparently
contain a single CnaA subdomain amongst the modules in the
binding region (Fig. 9). Our modeling suggests that binding of a
peptide between two domains in a manner analogous to binding
of a peptide to other MSCRAMMs is not feasible. Indeed, an
additional strand found in the subdomain of GspBBR occupies
the peptide binding site of MSCRAMMs. (Fig. S4, Fig. 3C
strand A9). Nevertheless, in SRR adhesins, the CnaA
subdomain is always found together with at least one other
subdomain, suggesting that the function may either require or be
tuned by the presence of a second subdomain. Indeed, recent
studies on the binding region of Fap1 support this hypothesis.
The structures of Fap1NR-a (helical subdomain) and Fap1NR-b
(CnaA subdomain) identified a region of surface-exposed
hydrophobic residues on each domain that is predicted to be
contiguous based upon SAXS. Fap1 normally mediates binding
of this bacterium to the oral cavity, and mutagenesis of these
hydrophobic residues abrogated binding to an in vitro tooth
model consisting of saliva-coated hydroxylapatite .
Our sequence analysis additionally predicts that the binding
regions of other SRR adhesins contain modules that are
structurally distinct from those identified in either GspBBR or
Fap1NR. Two of these, S. aureus SraPBR and S. epidermidis seSRRBR,
have high sequence identity to each other, but no detectable
sequence similarity to any other SRR adhesin (Fig. 9). This
strongly suggests that they bind a common, as yet unidentified,
host receptor that is distinct from the carbohydrates and keratins
recognized by other SRR adhesin family members. By
comparison, neither S. pneumoniae PsrPBR, which has been demonstrated to
bind keratin 10, nor the N-terminal subdomain of S. agalactiae
SRR1BR, which binds keratin 4, have detectable sequence
similarity to any currently available sequence (Fig. 9). Given this
analysis, it is clear that a more detailed understanding of the
binding characteristics of the SRR adhesin family will require that
for each module type, the binding partner should be identified,
and the structure of the binding region should be determined both
alone and in complex with the appropriate host receptor.
% of Platelets with Bacteria Bound (Mean S.D.)
Materials and Methods
All procedures involving rats were approved by the Los Angeles
Biomedical Research Institute animal use and care committee,
following the National Institutes of Health guidelines for animal
housing and care. Platelets were collected using a protocol
approved by the UCSF Committee on Human Research
(H1193-25513-07) and by the VUMC Human Research
Protection Program IRB Committee (110364).
Protein expression and purification
The DNA encoding residues 233617 of GspB (GspBBR) was
cloned into the pGEX vector containing an N-terminal
glutathione S-transferase (GST) fusion tag as described . This clone
contains a single base pair change as compared to the deposited
NCBI sequence that results in a serine at position 444 instead of an
asparagine. The protein was expressed with E. coli BL21 Gold
(Stratagene) in Luria Broth Medium as described  and purified
using a GST affinity column (GE Healthcare). The GST tag was
removed from GspBBR using Factor Xa, and GspBBR was further
purified using size exclusion chromatography on a Superdex
200 10/300 GL column (GE Healthcare) with buffer containing
20 mM Tris pH 7.4 .
All crystals of GspBBR were grown using the hanging drop
vapor diffusion technique at 23uC  using 1 ml protein solution
and 1 ml of reservoir solution equilibrated against 1 ml of the
reservoir solution. Crystals of native GspBBR grew from two
chemically distinct sets of conditions. The first set of conditions
included GspBBR at a concentration of 6 mg/ml buffered in
20 mM Tris pH 7.4 and equilibrated against a reservoir solution
containing 33% Jeffamine ED-2001, 0.1 M HEPES pH 7.5 and
0.15 M KCl at 23uC. These crystals belonged to the primitive
orthorhombic space group P212121 with unit cell dimensions
a = 33.7 A, b = 96.8 A, c = 100.2 A with a = b = c = 90u. Prior to
data collection, crystals were cryo-protected in a solution
containing all of the chemical components of each reservoir
solution and 15% glycerol then flash-cooled in liquid nitrogen.
The dataset used for refinement merged to 1.4 A resolution.
The second type of native GspBBR crystals grew when the
protein was equilibrated against a reservoir solution containing
25% polyethylene glycol 3350, 0.1 M HEPES pH 7.5, 0.15 M
NH4CH3COO, and 10 mM spermidine and induced crystal
growth from 10 mg/ml GspBBR buffered in 20 mM HEPES
pH 7.5 at 18uC. These crystals also belonged to the orthorhombic
space group P212121, but had altered unit cell dimensions of
a = 33.4 A, b = 86.7 A, c = 117.9 A, a = b = c = 90u. Prior to data
collection, these crystals were flash cooled in liquid nitrogen
without additional cryo-protectant. The best dataset from this
crystal form merged to 2.0 A resolution.
Crystals of GspBBR in complex with the a-2,3-sialyl
(1thioethyl)galactose disaccharide were grown using the hanging
drop vapor diffusion method with 6 mg/mL GspBBR in buffer
containing 1 mM a-2,3-sialyl (1-thioethyl)galactose and 20 mM
Tris pH 7.4. The reservoir solution contained 8% PEG 3350,
7.5 mM CoCl2, 7.5 mM NiCl2, 7.5 mM CdCl2, 7.5 mM MgCl2,
and 0.1 M HEPES pH 7.5. Crystals grew within two days and
were cryo-protected in reservoir solution supplemented with 20%
glycerol and flash-cooled in liquid nitrogen. The disaccharide
cocrystals formed in the primitive monoclinic space group P21 with
unit cell dimensions a = 69.9 A, b = 34.0 A, c = 83.4 A with
a = c = 90u and b = 99.2u. The best data from this crystal form
merged to 1.9 A resolution.
Data collection and processing
Crystals were assessed for diffraction quality at the Stanford
Synchrotron Radiation Lightsource (SSRL) beamlines 9-2, 11-1,
and 12-2 and the Life Sciences Collaborative Access Team
(LSCAT) beamlines ID-21-D/F/G. Datasets were collected using the
beamlines, temperatures, wavelengths, and detectors listed in
Table 1 and Table 2. All data were processed using the
HKL2000  and CCP4  suites of programs.
Preparation of heavy atom derivatives and structure
A Dy3+ derivative was prepared by soaking pre-formed crystals
of GspBBR in 1 mM DyCl3 for three days. Data were collected at
three wavelengths near the Dy3+ L3 edge (Table 2). Dy3+ bound to
a single site in the protein as determined using the SHELXD 
subroutine in the program SHARP . While HoCl3 also
successfully derivatized GspBBR, non-isomorphism between
S. gordonii densities (Mean S.D. log10 CFU/g tissue)
Figure 9. Modular organization of the binding regions within adhesins of the SRR superfamily. A. A summary of BLAST  or ClustalW
 sequence alignments of the binding regions of SRR adhesins. The first number indicates sequence identity and the second indicates sequence
similarity. GspBBR has been divided into its subdomains, with GspBBR-C indicating the CnaA subdomain and GspBBR-SU indicating both the Siglec and
Unique subdomains. Only the C-terminal subdomain of SRR1GBS (SRR1GBS-b) and Fap1NR (Fap1NR-b) are used in this analysis since the N-terminal
domains do not share detectable sequence identity with other sequences used in this analysis. Boxes in grey indicate SRR pairs that do not have
detectable sequence identity within the binding region. B. A schematic summarizing (A) The colored rectangles represent similar regions of
sequence, and are drawn to scale. If known, representative structures of each domain are illustrated along the top, and binding partners are indicated
along the side.
tals prevented the use of this second derivative in a traditional
MIR calculation. The non-isomorphism was so severe that phases
calculated in SHARP  only used data from a single DyCl3
soaked crystal and did not include a native dataset for reference.
This process resulted in reasonable phasing statistics and an overall
figure of merit of 0.83 at 2.0 A resolution (Table 2). Phases were
improved by solvent flattening in DM  which produced
electron density maps of high quality (Fig. 2B). Automated chain
tracing was performed using PHENIX , which was able to
trace residues 251316 and 327601, representing 94.7% of the
model. This resulted in an initial Rcryst of 23.8% and Rfree of
The lack of isomorphism between crystals prevented transfer of
these initial coordinates to other data sets using a simple rigid body
refinement. As a result, the model from the Dy3+ data set was
transferred to the remaining data sets using the program PHASER
 followed by rigid body refinement in CNS . Each
structure was subjected to alternate rounds of model building using
the program COOT  and refinement using CNS  and
The coordinates for the a-2,3-sialyl (1-thioethyl)galactose were
built using CCP4i Sketcher  and PRODRG . Refinement
statistics for all final models are listed in Table 1. Figures were
created using PyMOL , and inter-domain rotations were
determined using DynDom .
Synthesis of the a-2,3-sialyl (1-thioethyl)galactose
Based upon work by Danifshesky and coworkers , we
developed a four-step synthesis for the sialyl-T antigen precursor,
a-2,3-sialyl (1-thioethyl)galactose (see Supporting Protocol S1).
The correct synthesis of the disaccharide was verified by NMR
(Fig. S2). The a-2,3-sialyl (1-thioethyl)galactose was resuspended
in water for all applications.
Point mutations of gspB were introduced into the S. gordonii
chromosome via a strategy that involved recombination by double
cross-over between gspB codons 487602 and a gene
approximately 300 bp upstream. This approach ensured incorporation of
only the intended mutation of gspB codons ranging from 399 to
485, and avoided possible imprecise recombination within the
SRR regions. As a first step, the 59 end of gspB (codons 1 to 486
along with 200 nts from a non-coding region upstream) was
replaced with a chloramphenicol resistance cassette as follows. A
0.5 kb segment of a gene of unknown function upstream of gspB
was amplified using PCR. The product was digested with XhoI
and ClaI and then cloned upstream of the cat gene in pC326 .
A segment spanning gspB codons 487 to 602 was then amplified
using primers B487F and B602R, digested with SpeI and NotI,
and cloned downstream of the cat gene. The resulting plasmid,
pC326D59B, was used to transform S. gordonii strain M99 as
described . One of the chloramphenicol-resistant transformants,
(M99 D59gspB::cat), was selected for subsequent gene replacement.
A series of plasmids was then constructed to facilitate the
replacement of the 59 end of gspB in M99 D59gspB::cat. The
XhoIClaI fragment from pC326D59B was cloned upstream of the spec
gene in pS326 , and a 1 kb NsiI-SpeI fragment of gspB
(spanning codons 1 to 296), was cloned downstream. A SpeI-NotI
fragment spanning codons 296 to 602 was then cloned
downstream of the NsiI-SpeI fragment. Point mutations in the
resulting plasmid, pS326B602, were generated by a two-stage
PCR procedure. In the first stage, primer 25F along with a reverse
gspB primer, or the corresponding gspB forward primer along with
primer B602R, were used to amplify the upstream or downstream
segments, respectively. The two PCR products were combined for
the second stage and then amplified using primers 25F and
B602R. The PCR product was digested with SpeI and NotI and
then used to replace the corresponding segment of pS326B602.
The incorporation of only the intended change in any segment
generated by PCR was confirmed by DNA sequence analysis.
Plasmids were then used to transform M99 D59gspB::cat, resulting
in a replacement of D59gspB::cat with a 59gspB::spec variant. As a
control, the wild-type gspB sequence along with the spec cassette
(pS326B602) was also crossed into the M99 D59gspB::cat
chromosome (generating strain PS2161). Transformants were selected on
spectinomycin and scored for the loss of chloramphenicol
resistance. Expression of the variant GspB proteins on the
bacterial cell surface was verified by western blotting as described
GspBBR binding to biotinylated sialyl-T antigen
GST, GST-GspBBR and GST-GspBBR-R484E were purified
from E. coli as described . The purified proteins were diluted to
320 mM into DPBS, serial two-fold dilutions were made, and 50 ml
of each dilution was added to wells of a 96-well microtiter plate.
After incubating the plate overnight at 4uC, unbound proteins
were removed by aspiration and wells were rinsed with 100 ml
DPBS. Biotinylated sialyl-T antigen (sialyl-T antigen conjugated to
biotin via a polyacrylamide linker; GlycoTech Corporation) was
diluted to 50 mg/ml in DPBS containing 16 Blocking Reagent
(Roche), 50 ml was added to each well, and the plate was incubated
for 2 h at RT with vigorous rocking. After removing unbound
biotin-sialyl-T antigen, wells were rinsed three times with 100 ml
DPBS, 50 ml of streptavidin-conjugated horseradish peroxidase
(0.1 mg/ml in DPBS) was added to each well and the plate was
incubated for 1 h at 23uC. The wells were washed twice with
100 ml DPBS, and then 200 ml of a solution of OPD (0.4 mg/ml
citrate-phosphate buffer) was added to each well. The contents of
the wells were mixed by gently vortexing the plate, and the
absorbance at 450 nm was measured 20 min after the addition of
the OPD substrate. Data were plotted as the means 6 standard
deviations, with n = 3.
Binding of S. gordonii to glycocalicin and platelet
The binding of S. gordonii to immobilized glycocalicin was
performed as described previously . In brief, strains were
grown for 18 hr, washed twice with DPBS, sonicated briefly to
disrupt aggregated cells, and then diluted to approximately 26107
per ml. To determine whether a-2,3 sialyl (1-thioethyl)galactose
inhibited binding to glycocalicin, the washed and sonicated
bacteria were diluted into DPBS or DPBS containing 44 mM
a2,3-sialyl (1-thioethyl)galactose, pH 7.5. The bacterial suspensions
were then applied to wells of a microtiter plate that had been
coated with glycocalicin (1.25 mg/well). After 2 h at room
temperature, the unbound bacteria were removed by aspiration.
Wells were washed three times with DPBS, and the bound
bacteria were released by trypsinization. The number of input and
bound bacteria were determined by plating serial dilutions of the
bacterial suspensions on sheep blood agar plates, and binding was
expressed as the percent of the input bound to glycocalicin. The
binding of S. gordonii to immobilized human platelets was assessed
as described previously . Results of both assays are reported as
the means 6 standard deviations, with n = 6. Differences in
binding were compared by the unpaired t-test.
DIC and fluorescence microscopy
S. gordonii strains M99 wild type, PS2116, and PS846 were
grown in 5 mL of Todd Hewitt Broth at 37uC without shaking for
18 hours. Cells were then vortexed to resuspend bacteria and spun
down at 40006 g for ten minutes. The supernatant was removed
and the bacteria washed twice with 5 mL of DPBS containing
MgCl2 and CaCl2. The bacteria were then resuspended in 5 mL
of DPBS and sonicated briefly to disrupt aggregated clumps. To
1 mL of each cell suspension, 500 nM
49,6-diamidino-2-phenylindole (DAPI) was added.
Platelets were fixed using 1.6% paraformaldehyde. 500 mL of
platelets were mounted on poly-L-lysine coated cover slips (in a
6well tray) which was spun at 4006 g for ten minutes in order to
promote platelet adherence to the cover slips. Excess platelets were
removed by washing with Tris Buffered Saline (TBS) and 1 mL of
TBS was added to each well. 500 mL of the bacterial suspension
was added to the platelets and the samples were rocked vigorously
for 30 minutes at 23uC. Excess bacteria were then removed by
washing three times with TBS. Each sample was mounted onto a
slide for microscopy. Slides were imaged using Nikon TiE Inverted
light microscope equipped with a Photometrics CoolSnap HQ
CCD camera. Platelets were imaged with a 10061.49NA objective
using DIC optics and a DAPI filter cube. Image J software was
used to create the contrast and composite images.
Rat model of infective endocarditis
A competition model of infective endocarditis was produced in
SpragueDawley female rats (250300 g) as described previously
. In brief, the animals were anesthetized with ketamine (35 mg/
kg) and xylazine (10 mg/kg). A sterile polyethylene catheter was
surgically placed across the aortic valve of each animal, such that the
tip was positioned in the left ventricle. Catheters were left in place
throughout the study. Catheterized animals were then infected
intravenously (IV) with an inoculum containing 26105 CFU of
both strains (i.e., a 1:1 mixture of i) S. gordonii strain M99 and strain
PS2116 (59gspBR484E::spec), ii) M99 and strain PS2161 (59gspB::spec
control strain), or iii) PS2116 and strain PS846 (M99 DgspB::pEVP3;
CmR) [18,73]. At 72 h post-infection, animals were sacrificed with
thiopental (100 mg, intraperitoneally). Animals were included in the
final analysis only if the catheters were correctly positioned across
the aortic valve at the time of sacrifice, and if macroscopic
vegetations were seen. All cardiac vegetations, as well as samples of
the kidneys and spleens were harvested, weighed, homogenized in
saline, serially diluted, and plated onto Todd Hewitt agar (THA) (for
the parental S. gordonii strain M99) and THA containing 100 mg/ml
of spectinomycin (for strains PS2116 and PS2161) or
chloramphenicol 5 mg/ml (for strain PS846) for quantitative culture, to
determine the number of CFU/g of S. gordonii strains within tissues.
After 48 h of incubation at 37uC, bacterial colonies were counted.
The number of bacteria within tissues was expressed as the log10
CFU per gram of tissue. Differences between means were compared
for statistical significance by the paired t-test, using p#0.05 as the
threshold for significance. The data were also analyzed by
calculating a competition index, which was defined as the ratio
of S. gordonii strain M99 and strain PS2116 or PS2161, as well as
PS2116 and PS846, within tissues for each animal, normalized by
the ratio of organisms in the inoculum. The mean of the log10
normalized ratios was tested against the hypothesized no effect
mean value of 0, as described previously, using a paired t-test.
The coordinates and structure factors for S. gordonii strain M99
GspBBR have been deposited in the Research Collaboratory for
Structural Bioinformatics Protein Data Bank with accession codes
3QC5 (native 1), 3QC6 (native 2, crystal form 2), and 3QD1
(a2,3-sialyl (1-thioethyl)galactose bound).
Figure S1 The cation binding site. Stereo view of the cation
binding site superpositioned with 2m|Fo|2d|Fc| electron density
calculated in REFMAC5  and contoured to 1.5 s. Carbons
are shown in grey, oxygens are shown in red, nitrogens are shown in
blue and the cation is shown as a purple sphere. Bond distances are
shown in A.
Figure S3 Predicted binding of sialyl-T antigen to the receptor
binding pocket. Using the co-structure of GspBBR with a-2,3-sialyl
(1-thioethyl)galactose as a starting point, we developed a model for
binding of sialyl-T antigen to the Siglec subdomain. Coordinates
for sialyl-T antigen were prepared in COOT  by manually
linking the 3-position of N-acetylgalactosamine to the 1-position of
the galactose in a-2,3-sialyl (1-thioethyl) galactose and removing
the thioethyl group. The position of the a-2,3-sialyl (1-thioethyl)
galactose group was fixed to its position in the
experimentallydetermined structure and the galactosamine manually optimized
to avoid steric clashes. In this model, all three sugars bind within a
contiguous pre-formed pocket on the surface of the protein
between the Siglec and Unique subdomains. The Siglec
subdomain is colored yellow, the Unique subdomain is colored
blue, and the modeled sialyl-T antigen is colored with gray carbons.
Figure S4 Binding of MSCRAMMs to fibrinogen receptors.
The interdomain peptide binding observed in SdrG (PDB entry
1R17) uses a dock, lock, and latch mechanism of fibrinogen
recognition  is unlikely to be used by SRR adhesins. This
interdomain mechanism requires two adjacent Ig-fold domains
that create a binding site; the fibrinogen peptide binds between
these two domains. Modeling additional interdomain angles in
GspBBR suggests that the short linker region between the CnaA
and Siglec subdomains will prevent the interdomain angle from
becoming sufficiently acute to allow the Siglec subdomain to act as
a surrogate second subdomain. In addition, GspBBR contains an
additional strand at the N-terminus of the CnaA subdomain
(Fig. 3C strand A) that occupies the peptide binding site of
MSCRAMMs. Unbound SdrG is colored gray and bound SdrG is
colored red with the fibrinogen peptide colored cyan.
Video S1 Range of interdomain angles observed in crystals of
GspBBR. The movie morphs between states observed in the crystal
structures and was produced using LSQMAN  and PyMOL
. The most extended conformation was observed in the
costructure of GspBBR with the a-2,3-sialyl (1-thioethyl)galactose
Results and discussion regarding the cation binding site.
We thank Dr. Mikio Tanabe, Dr. Timothy Panosian, Dr. Jessica Vey,
Rebecca McRae, Dr. Matthew Duvernay, Jessica Mazerick, Dr.
Rajalakshmi Nambiar, Laura Anzaldi, Summer Young, Tarjani Thaker,
and Dr. Kathryn McCulloch for experimental assistance.
Conceived and designed the experiments: GC GAS MJT PMS TMI.
Performed the experiments: TMP BAB YQX BJM VY KMO MJT.
Analyzed the data: TMP BAB YQX BJM TMT NJW MJT TMI.
Contributed reagents/materials/analysis tools: VY. Wrote the paper: TMP
BAB YQX BJM GAS MJT PMS TMI.
1. Takahashi Y , Konishi K , Cisar JO , Yoshikawa M ( 2002 ) Identification and characterization of hsa, the gene encoding the sialic acid-binding adhesin of Streptococcus gordonii DL1 . Infect Immun 70 : 1209 - 1218 .
2. Handley PS , Correia FF , Russell K , Rosan B , DiRienzo JM ( 2005 ) Association of a novel high molecular weight, serine-rich protein (SrpA) with fibril-mediated adhesion of the oral biofilm bacterium Streptococcus cristatus . Oral Microbiol Immunol 20 : 131 - 140 .
3. Wu H , Mintz KP , Ladha M , Fives-Taylor PM ( 1998 ) Isolation and characterization of Fap1, a fimbriae-associated adhesin of Streptococcus parasanguis FW213 . Mol Microbiol 28 : 487 - 500 .
4. Bensing BA , Sullam PM ( 2002 ) An accessory sec locus of Streptococcus gordonii is required for export of the surface protein GspB and for normal levels of binding to human platelets . Mol Microbiol 44 : 1081 - 1094 .
5. Sanchez CJ , Shivshankar P , Stol K , Trakhtenbroit S , Sullam PM , et al. ( 2010 ) The pneumococcal serine-rich repeat protein is an intra-species bacterial adhesin that promotes bacterial aggregation in vivo and in biofilms . PLoS Pathog 6 ( 8 ) : e1001044 . doi:10.1371/journal.ppat.1001044.
6. Siboo IR , Chambers HF , Sullam PM ( 2005 ) Role of SraP, a Serine-Rich Surface Protein of Staphylococcus aureus, in binding to human platelets . Infect Immun 73 : 2273 - 2280 .
7. Seifert KN , Adderson EE , Whiting AA , Bohnsack JF , Crowley PJ , et al. ( 2006 ) A unique serine-rich repeat protein (Srr-2) and novel surface antigen (epsilon) associated with a virulent lineage of serotype III Streptococcus agalactiae . Microbiology 152 : 1029 - 1040 .
8. Wu H , Fives-Taylor PM ( 1999 ) Identification of dipeptide repeats and a cell wall sorting signal in the fimbriae-associated adhesin, Fap1, of Streptococcus parasanguis . Mol Microbiol 34 : 1070 - 1081 .
9. Plummer C , Wu H , Kerrigan SW , Meade G , Cox D , et al. ( 2005 ) A serine-rich glycoprotein of Streptococcus sanguis mediates adhesion to platelets via GPIb . Br J Haematol 129 : 101 - 109 .
10. Takamatsu D , Bensing BA , Prakobphol A , Fisher SJ , Sullam PM ( 2006 ) Binding of the streptococcal surface glycoproteins GspB and Hsa to human salivary proteins . Infect Immun 74 : 1933 - 1940 .
11. Takamatsu D , Bensing BA , Cheng H , Jarvis GA , Siboo IR , et al. ( 2005 ) Binding of the Streptococcus gordonii surface glycoproteins GspB and Hsa to specific carbohydrate structures on platelet membrane glycoprotein Ib alpha . Mol Microbiol 58 : 380 - 392 .
12. van Sorge NM , Quach D , Gurney MA , Sullam PM , Nizet V , et al. ( 2009 ) The group B streptococcal serine-rich repeat 1 glycoprotein mediates penetration of the blood-brain barrier . J Infect Dis 199 : 1479 - 1487 .
13. Mistou MY , Dramsi S , Brega S , Poyart C , Trieu-Cuot P ( 2009 ) Molecular dissection of the secA2 locus of group B Streptococcus reveals that glycosylation of the Srr1 LPXTG protein is required for full virulence . J Bacteriol 191 : 4195 - 4206 .
14. Yajima A , Urano-Tashiro Y , Shimazu K , Takashima E , Takahashi Y , et al. ( 2008 ) Hsa, an adhesin of Streptococcus gordonii DL1, binds to alpha2-3-linked sialic acid on glycophorin A of the erythrocyte membrane . Microbiol Immunol 52 : 69 - 77 .
15. Takahashi Y , Sandberg AL , Ruhl S , Muller J , Cisar JO ( 1997 ) A specific cell surface antigen of Streptococcus gordonii is associated with bacterial hemagglutination and adhesion to alpha2-3-linked sialic acid-containing receptors . Infect Immun 65 : 5042 - 5051 .
16. Samen U , Eikmanns BJ , Reinscheid DJ , Borges F ( 2007 ) The surface protein Srr-1 of Streptococcus agalactiae binds human keratin 4 and promotes adherence to epithelial HEp-2 cells . Infect Immun 75 : 5405 - 5414 .
17. Shivshankar P , Sanchez C , Rose LF , Orihuela CJ ( 2009 ) The Streptococcus pneumoniae adhesin PsrP binds to Keratin 10 on lung cells . Mol Microbiol 73 : 663 - 679 .
18. Xiong YQ , Bensing BA , Bayer AS , Chambers HF , Sullam PM ( 2008 ) Role of the serine-rich surface glycoprotein GspB of Streptococcus gordonii in the pathogenesis of infective endocarditis . Microb Pathog 45 : 297 - 301 .
19. Obert C , Sublett J , Kaushal D , Hinojosa E , Barton T , et al. ( 2006 ) Identification of a Candidate Streptococcus pneumoniae core genome and regions of diversity correlated with invasive pneumococcal disease . Infect Immun 74 : 4766 - 4777 .
20. Petersen HJ , Keane C , Jenkinson HF , Vickerman MM , Jesionowski A , et al. ( 2010 ) Human platelets recognize a novel surface protein, PadA, on Streptococcus gordonii through a unique interaction involving fibrinogen receptor GPIIbIIIa . Infect Immun 78 : 413 - 422 .
21. Kerrigan SW , Jakubovics NS , Keane C , Maguire P , Wynne K , et al. ( 2007 ) Role of Streptococcus gordonii surface proteins SspA/SspB and Hsa in platelet function . Infect Immun 75 : 5740 - 5747 .
22. Jakubovics NS , Kerrigan SW , Nobbs AH , Stromberg N , van Dolleweerd CJ , et al. ( 2005 ) Functions of cell surface-anchored antigen I/II family and Hsa polypeptides in interactions of Streptococcus gordonii with host receptors . Infect Immun 73 : 6629 - 6638 .
23. Ganesh VK , Rivera JJ , Smeds E , Ko YP , Bowden MG , et al. ( 2008 ) A structural model of the Staphylococcus aureus ClfA-fibrinogen interaction opens new avenues for the design of anti-staphylococcal therapeutics . PLoS Pathog 4 : e1000226 .
24. McDevitt D , Nanavaty T , House-Pompeo K , Bell E , Turner N , et al. ( 1997 ) Characterization of the interaction between the Staphylococcus aureus clumping factor (ClfA) and fibrinogen . Eur J Biochem 247 : 416 - 424 .
25. Miajlovic H , Loughman A , Brennan M , Cox D , Foster TJ ( 2007 ) Both complement- and fibrinogen-dependent mechanisms contribute to platelet aggregation mediated by Staphylococcus aureus clumping factor B . Infect Immun 75 : 3335 - 3343 .
26. Loughman A , Fitzgerald JR , Brennan MP , Higgins J , Downer R , et al. ( 2005 ) Roles for fibrinogen, immunoglobulin and complement in platelet activation promoted by Staphylococcus aureus clumping factor A. Mol Microbiol 57 : 804 - 818 .
27. O'Brien L , Kerrigan SW , Kaw G , Hogan M , Penades J , et al. ( 2002 ) Multiple mechanisms for the activation of human platelet aggregation by Staphylococcus aureus: roles for the clumping factors ClfA and ClfB, the serine-aspartate repeat protein SdrE and protein A. Mol Microbiol 44 : 1033 - 1044 .
28. Fitzgerald JR , Foster TJ , Cox D ( 2006 ) The interaction of bacterial pathogens with platelets . Nat Rev Microbiol 4 : 445 - 457 .
29. Seo HS , Xiong YQ , Mitchell J , Seepersaud R , Bayer AS , et al. ( 2010 ) Bacteriophage lysin mediates the binding of streptococcus mitis to human platelets through interaction with fibrinogen . PLoS Pathog 6 ( 8 ) : e1001047 . doi:1001010.1001371/journal.ppat.1001047.
30. Durack DT ( 1975 ) Experimental bacterial endocarditis . IV. Structure and evolution of very early lesions . J Pathol 115 : 81 - 89 .
31. Yajima A , Takahashi Y , Konishi K ( 2005 ) Identification of platelet receptors for the Streptococcus gordonii DL1 sialic acid-binding adhesin . Microboil Immunol 49 : 795 - 800 .
32. Bensing BA , Lopez JA , Sullam PA ( 2004 ) The Streptococcus gordonii surface proteins GspB and Hsa mediate binding to sialylated carbohydrate epitopes on the platelet membrane glycoprotein Ib alpha . Infect Immun 72 : 6528 - 6537 .
33. Takahashi Y , Takashima E , Shimazu K , Yagishita H , Aoba T , et al. ( 2006 ) Contribution of sialic acid-binding adhesin to pathogenesis of experimental endocarditis caused by Streptococcus gordonii DL1 . Infect Immun 74 : 740 - 743 .
34. Ponnuraj K , Bowden MG , Davis S , Gurusiddappa S , Moore D , et al. ( 2003 ) A ''dock, lock, and latch'' structural model for a staphylococcal adhesin binding to fibrinogen . Cell 115 : 217 - 228 .
35. Symersky J , Patti JM , Carson M , House-Pompeo K , Teale M , et al. ( 1997 ) Structure of the collagen-binding domain from a Staphylococcus aureus adhesin . Nat Struct Biol 4 : 833 - 838 .
36. Deivanayagam CC , Wann ER , Chen W , Carson M , Rajashankar KR , et al. ( 2002 ) A novel variant of the immunoglobulin fold in surface adhesins of Staphylococcus aureus: crystal structure of the fibrinogen-binding MSCRAMM, clumping factor A . EMBO J 21 : 6660 - 6672 .
37. Holmgren A , Branden CI ( 1989 ) Crystal structure of chaperone protein PapD reveals an immunoglobulin fold . Nature 342 : 248 - 251 .
38. Cota E , Jones C , Simpson P , Altroff H , Anderson KL , et al. ( 2006 ) The solution structure of the invasive tip complex from Afa/Dr fibrils . Mol Microbiol 62 : 356 - 366 .
39. Hamburger ZA , Brown MS , Isberg RR , Bjorkman PJ ( 1999 ) Crystal structure of invasin: a bacterial integrin-binding protein . Science 286 : 291 - 295 .
40. Jedrzejczak R , Dauter Z , Dauter M , Piatek R , Zalewska B , et al. ( 2006 ) Structure of DraD invasin from uropathogenic Escherichia coli: a dimer with swapped beta-tails . Acta Crystallogr D Biol Crystallogr 62 : 157 - 164 .
41. Vengadesan K , Ma X , Dwivedi P , Ton-That H , Narayana SV ( 2011 ) A Model for Group B Streptococcus Pilus Type 1: The Structure of a 35-kDa C-Terminal Fragment of the Major Pilin GBS80 . J Mol Biol 407 : 731 - 743 .
42. Holm L , Rosenstrom P ( 2010 ) Dali server: conservation mapping in 3D . Nucleic Acids Res 38 : W545 - 549 .
43. Ramboarina S , Garnett JA , Zhou M , Li Y , Peng Z , et al. ( 2010 ) Structural insights into serine-rich fimbriae from gram-positive bacteria . J Biol Chem 285 : 32446 - 32457 .
44. May AP , Robinson RC , Vinson M , Crocker PR , Jones EY ( 1998 ) Crystal structure of the N-terminal domain of sialoadhesin in complex with 39 sialyllactose at 1.85 A resolution . Mol Cell 1 : 719 - 728 .
45. Zhuravleva MA , Trandem K , Sun PD ( 2008 ) Structural implications of Siglec-5- mediated sialoglycan recognition . J Mol Biol 375 : 437 - 447 .
46. Attrill H , Imamura A , Sharma RS , Kiso M , Crocker PR , et al. ( 2006 ) Siglec-7 undergoes a major conformational change when complexed with the alpha(2,8)- disialylganglioside GT1b . J Biol Chem 281 : 32774 - 32783 .
47. Chung MC , Wines BD , Baker H , Langley RJ , Baker EN , et al. ( 2007 ) The crystal structure of staphylococcal superantigen-like protein 11 in complex with sialyl Lewis X reveals the mechanism for cell binding and immune inhibition . Mol Microbiol 66 : 1342 - 1355 .
48. Baker HM , Basu I , Chung MC , Caradoc-Davies T , Fraser JD , et al. ( 2007 ) Crystal structures of the staphylococcal toxin SSL5 in complex with sialyl Lewis X reveal a conserved binding site that shares common features with viral and bacterial sialic acid binding proteins . J Mol Biol 374 : 1298 - 1308 .
49. Imberty A , Varrot A ( 2008 ) Microbial recognition of human cell surface glycoconjugates . Curr Opin Struct Biol 18 : 567 - 576 .
50. Altschul SF , Gish W , Miller W , Myers EW , Lipman DJ ( 1990 ) Basic local alignment search tool . J Mol Biol 215 : 403 - 410 .
51. Chenna R , Sugawara H , Koike T , Lopez R , Gibson TJ , et al. ( 2003 ) Multiple sequence alignment with the Clustal series of programs . Nucleic Acids Res 31 : 3497 - 3500 .
52. De Greve H , Wyns L , Bouckaert J ( 2007 ) Combining sites of bacterial fimbriae . Curr Opin Struct Biol 17 : 506 - 512 .
53. Sotomayor M , Schulten K ( 2008 ) The allosteric role of the Ca2+ switch in adhesion and elasticity of C-cadherin . Biophys J 94 : 4621 - 4633 .
54. Sokurenko EV , Vogel V , Thomas WE ( 2008 ) Catch-bond mechanism of forceenhanced adhesion: counterintuitive , elusive, but widespread? Cell Host Microbe 4 : 314 - 323 .
55. Pyburn TM , Yankovskaya V , Bensing BA , Cecchini G , Sullam PM , et al. ( 2010 ) Purification, crystallization and preliminary X-ray diffraction analysis of the carbohydrate-binding region of the Streptococcus gordonii adhesin GspB . Acta Crystallogr F Struct Biol Cryst Commun 66 : 1503 - 1507 .
56. Otwinowski Z , Minor W ( 1997 ) Processing of X-ray diffraction data collected in oscillation mode . Macromolecular Crystallography, Pt A 276 : 307 - 326 .
57. Bailey S ( 1994 ) The Ccp4 Suite - Programs for Protein Crystallography . Acta Crystallogr D Biol Crystallogr 50 : 760 - 763 .
58. Sheldrick GM ( 2008 ) A short history of SHELX . Acta Crystallogr A 64 : 112 - 122 .
59. de La Fortelle E , Bricogne G ( 1997 ) Maximum-likelihood heavy-atom parameter refinement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods . Macromolecular Crystallography, Pt A 276 : 472 - 494 .
60. Bailey S ( 1994 ) The Ccp4 Suite - Programs for Protein Crystallography . Acta Crystallogr D Biol Crystallogr 50 : 760 - 763 .
61. Adams PD , Afonine PV , Bunkoczi G , Chen VB , Davis IW , et al. ( 2010 ) PHENIX: a comprehensive Python-based system for macromolecular structure solution . Acta Crystallogr D Biol Crystallogr 66 : 213 - 221 .
62. Mccoy AJ , Grosse-Kunstleve RW , Adams PD , Winn MD , Storoni LC , et al. ( 2007 ) Phaser crystallographic software . J Appl Crystallogr 40 : 658 - 674 .
63. Brunger AT ( 2007 ) Version 1.2 of the Crystallography and NMR system . Nat Protoc 2 : 2728 - 2733 .
64. Emsley P , Cowtan K ( 2004 ) Coot: model-building tools for molecular graphics . Acta Crystallogr D Biol Crystallogr 60 : 2126 - 2132 .
65. Murshudov GN , Vagin AA , Dodson EJ ( 1997 ) Refinement of macromolecular structures by the maximum-likelihood method . A Acta Crystallogr D Biol Crystallogr 53 : 240 - 255 .
66. Potterton E , Briggs P , Turkenburg M , Dodson E ( 2003 ) A graphical user interface to the CCP4 program suite . Acta Crystallogr D Biol Crystallogr 59 : 1131 - 1137 .
67. Schuttelkopf AW , van Aalten DM ( 2004 ) PRODRG: a tool for high-throughput crystallography of protein-ligand complexes . Acta Crystallogr D Biol Crystallogr 60 : 1355 - 1363 .
68. DeLano WL ( 2002 ) The PyMOL Molecular Graphics System .
69. Hayward S , Berendsen HJ ( 1998 ) Systematic analysis of domain motions in proteins from conformational change: new results on citrate synthase and T4 lysozyme . Proteins 30 : 144 - 154 .
70. Schwarz JB , Kuduk SD , Chen XT , Sames D , Glunz PW , et al. ( 1999 ) A broadly applicable method for the efficient synthesis of alpha-O-linked glycopeptides and clustered sialic acid residues . J Am Chem Soc 121 : 2662 - 2673 .
71. Mitchell J , Siboo IR , Takamatsu D , Chambers HF , Sullam PM ( 2007 ) Mechanism of cell surface expression of the Streptococcus mitis platelet binding proteins PblA and PblB . Mol Microbiol 64 : 844 - 857 .
72. Takamatsu D , Bensing BA , Sullam PM ( 2004 ) Genes in the accessory sec locus of Streptococcus gordonii have three functionally distinct effects on the expression of the platelet-binding protein GspB . Mol Microbiol 52 : 189 - 203 .
73. Bensing BA , Takamatsu D , Sullam PM ( 2005 ) Determinants of the streptococcal surface glycoprotein GspB that facilitate export by the accessory Sec system . Mol Microbiol 58 : 1468 - 1481 .
74. Alphey MS , Attrill H , Crocker PR , van Aalten DM ( 2003 ) High resolution crystal structures of Siglec-7. Insights into ligand specificity in the Siglec family . J Biol Chem 278 : 3372 - 3377 .
75. Kleywegt GJ ( 1996 ) Use of non-crystallographic symmetry in protein structure refinement . Acta Crystallogr D Biol Crystallogr 52 : 842 - 857 .