Crystal structure of a Staphylococcus aureus protein

A domain complexed with the Fab fragment of a

human IgM antibody: Structural basis for
recognition of B-cell receptors and
superantigen activity
Marc Graille*, Enrico A. Stura*, Adam L. Corper†, Brian J. Sutton†, Michael J. Taussig‡, Jean-Baptiste Charbonnier*§,
and Gregg J. Silverman§¶
*Département d’Ingénierie et d’Etudes des Protéines (DIEP), Commissariat à l’Energie Atomique (CEA), C.E. Saclay, 91191 Gif-sur-Yvette Cedex, France; †The
Randall Centre, King’s College London, Guy’s Campus, London, SE1 1UL, United Kingdom; ‡Laboratory of Molecular Recognition, The Babraham Institute,
Babraham, Cambridge CB2 4AT, United Kingdom; and ¶Department of Medicine, University of California at San Diego, La Jolla, CA 92093-0663

Edited by Johann Deisenhofer, University of Texas Southwestern Medical Center, Dallas, TX, and approved February 24, 2000 (received for review
December 13, 1999)

Staphylococcus aureus produces a virulence factor, protein A (SpA), through interactions with surface membrane-associated VH3-
that contains five homologous Ig-binding domains. The interac- encoded B-cell antigen receptors (11), in vitro stimulation with
tions of SpA with the Fab region of membrane-anchored Igs can SpA can contribute to selection of these B cells and promote
stimulate a large fraction of B cells, contributing to lymphocyte their production of antibodies that may include rheumatoid
clonal selection. To understand the molecular basis for this activity, factor autoantibodies (12, 13). In vivo exposure to recombinant
we have solved the crystal structure of the complex between SpA can result in supraclonal suppression and deletion of
domain D of SpA and the Fab fragment of a human IgM antibody B lymphocytes that are susceptible based on their VH usage
to 2.7-Å resolution. In the complex, helices II and III of domain D (14, 15).
interact with the variable region of the Fab heavy chain (VH) Although the mechanism(s) are not defined, experimental
through framework residues, without the involvement of the models indicate that SpA enhances staphylococcal virulence (16,
hypervariable regions implicated in antigen recognition. The con- 17). Many features of the interactions of SpA with host B
Downloaded from https://www.pnas.org by 41.254.43.121 on November 27, 2023 from IP address 41.254.43.121.

tact residues are highly conserved in human VH3 antibodies but not lymphocytes are akin to those of superantigens for T lympho-
in other families. The contact residues from domain D also are cytes that cause a variety of inflammatory diseases including
conserved among all SpA Ig-binding domains, suggesting that each toxic shock syndrome, food poisoning, and exfoliative syndromes
could bind in a similar manner. Features of this interaction parallel (18–20), and T-cell superantigens also have been postulated to
those reported for staphylococcal enterotoxins that are superan- contribute to the pathogenesis of autoimmune disease (18, 21).
tigens for many T cells. The structural homology between Ig VH These superantigens target T-cell receptors (TcRs) from partic-
regions and the T-cell receptor V␤ regions facilitates their compar- ular variable ␤ chain (V␤) families and induce global changes in
ison, and both types of interactions involve lymphocyte receptor T lymphocyte repertoires (18).
surface remote from the antigen binding site. However, T-cell Here, we report the crystal structure of domain D of SpA
superantigens reportedly interact through hydrogen bonds with complexed with the Fab fragment of a human IgM antibody
T-cell receptor V␤ backbone atoms in a primary sequence-indepen- and describe the key contact residues from both partners in the
dent manner, whereas SpA relies on a sequence-restricted confor- interaction. The residues in domain D involved in the inter-
mational binding with residue side chains, suggesting that this action with Fab are highly conserved in other Ig-binding
common bacterial pathogen has adopted distinct molecular rec- domains of SpA, and these are distinct from the residues that
ognition strategies for affecting large sets of B and T lymphocytes.
mediate the binding of an SpA domain and Fc␥ (2). In the Fab,
the residues in contact with SpA are located in the VH region
framework ␤-strands and the interstrand loops most remote
T he common bacterial pathogen, Staphylococcus aureus,
produces a 42-kDa factor, protein A (SpA), that contains
five highly homologous extracellular Ig-binding domains in
from the antigen combining site. The structure of the complex
provides a rationale for the restricted specificity of SpA toward
VH3-encoded antibodies, the largest human VH gene family.
tandem, designated domains E, D, A, B, and C. Protein A, Hence, elucidation of the structural features of the binding
which exists in both secreted and membrane-associated forms, interactions of SpA aids our understanding of the biological
possesses two distinct Ig-binding activities: each domain can properties of a B-cell superantigen. In addition, structural
bind Fc␥ (the constant region of IgG involved in effector comparisons with the characterized interactions of bacterial
functions) and Fab (the Ig fragment responsible for antigen
recognition) (1). The Fc␥ binding site has been localized to the
elbow region at the CH2 and CH3 interface of most IgG This paper was submitted directly (Track II) to the PNAS office.
subclasses, and this binding property has been extensively used Abbreviations: SpA, Staphylococcus aureus protein A; TcR, T-cell receptor; Fc␥, constant
for the labeling and purification of antibodies (2, 3). The Fab region of IgG; VH and V␤, variable region of the Fab heavy chain and TcR ␤ chain,
respectively; CDR, complementarity determining region; rmsd, rms deviation.
specificity is less well characterized but it has been shown to
involve a site on the variable region of the Ig heavy chain (4). Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,
IMMUNOLOGY

www.rcsb.org (PDB ID code 1DEE).

Correlation with antibody sequence usage indicates that the §To whom reprint requests should be addressed. E-mail: jbcharbonnier@cea.fr or
Fab binding specificity is restricted to products of the human gsilverman@ucsd.edu.
variable region of the Fab heavy chain VH3 family that The publication costs of this article were defrayed in part by page charge payment. This
represent nearly half of inherited VH genes (5– 8) and their article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
homologues in other mammalian species (9, 10). Presumably §1734 solely to indicate this fact.

PNAS 兩 May 9, 2000 兩 vol. 97 兩 no. 10 兩 5399 –5404

 T-cell superantigens reveal how proteins produced by the same the variable domain heterodimers in the Fab are unchanged
bacterial pathogen target the two limbs of the adaptive im- between the two complexed Fab with rmsd of 0.25 Å over 903
mune system. backbone atoms. Similarly, both complexed Fab superimpose
well on the unliganded Fab with rmsd of 0.14 Å and 0.33 Å over
Materials and Methods 904 backbone atoms. The domain D assumes the triple ␣-helical
Crystallization. The Fab of the VH3–30兾1.9III-encoded 2A2 IgM bundle reported for domains B and E (28, 31). All three helices
rheumatoid factor was produced by trypsin cleavage of the IgM of domain D in the Fab-domain D complexes are clearly defined
secreted by a hybridoma created from synovial B cells of a in electron density maps and superimpose well with the reported
rheumatoid arthritis patient, as described (22). The production domain E structure (28) with rmsd of 0.52 Å over 196 main-chain
of recombinant domain D of SpA used in crystallization was atoms. For both complexes, the Fab-domain D interaction buries
described in Roben et al. (23). In the crystallization screening, a total solvent accessible surface area of 1,220 Å2 with approx-
various ratios of Fab and domain D were tested, with subsequent imately equal contributions from both molecules, as determined
optimization of ratios to improve crystal size. Crystals were with a 1.4-Å probe. This value is similar to many antigen-
grown by vapor diffusion at room temperature in sitting drops by antibody complexes and other types of complexes with Ig-
mixing a reservoir solution of 21–24% (wt兾wt) monomethyl binding proteins but lower than the average calculated for
polyethylene glycol 5,000, 100 mM sodium cacodylate, pH 6.5 reported protein–protein complexes (32).
with an equal volume of the protein solution at 5 mg兾ml. Crystals In the complex, the Fab interacts with helix II and helix III of
for data collection were enlarged by using streak seeding fol- domain D via a surface composed of four VH region ␤-strands:
lowed by macroseeding (24). Data were recorded at room B, C⬙, D, and E (Fig. 1 A–C). The major axis of helix II of domain
temperature from two crystals on a Rigaku rotating anode D is approximately 50o to the orientation of the strands, and the
generator with Supper long mirrors by using a MarResearch interhelical portion of domain D is most proximal to the C⬙
image plate detector and processed by using the HKL package strand. The site of interaction on Fab is remote from the Ig light
(25). These crystals belong to the monoclinic space group P21 chain and the heavy chain constant region. The interaction
with a ⫽ 68.5 Å, b ⫽ 78.9 Å, c ⫽ 163.2 Å, ␤ ⫽ 100.7o. The data involves the following domain D residues: Gln-26, Gly-29, Phe-
set used for structural analysis has 46,831 unique reflections in 30, Gln-32, Ser-33, and Asp-36 of helix II; Asp-37 and Gln-40 in
the 20- to 2.7-Å resolution with an overall Rmerge of 6.5% with the loop between helix II and helix III; and Asn-43, Glu-47, and
87% completeness and a 3.9-fold redundancy. In the 2.8- to Leu-51 of helix III. In the Fab, the interaction is mediated by the
2.7-Å shell, completion is 77% with an ⬍I兾␴⬎ of 1.9 and an heavy chain residues: Gly-H15 and Ser-H17 in the ␤ turn before
Rmerge of 37%. strand B; Arg-H19 of strand B; Lys-H57 and Tyr-H59 of strand
C⬙; Lys-H64, Gly-H65, and Arg-H66 before strand D; Thr-H68
Structure Determination and Refinement. The structure was solved and Ser-H70 of strand D; Gln-H81 of strand E; Asn-H82a and
by molecular replacement by using the program AMORE (26) and Ser-H82b after strand E. Six of these residues are in framework
Downloaded from https://www.pnas.org by 41.254.43.121 on November 27, 2023 from IP address 41.254.43.121.

the coordinates of the uncomplexed 2A2 Fab-rheumatoid factor region (FR) ␤ strands, whereas the other seven residues are in
from a previously solved crystal structure (B.J.S., A.L.C., and VH region FR interstrand loops on the side farthest from the
M.J.T., unpublished work). In the resulting model, the three antigen binding pocket. Both interacting surfaces are composed
Fabs in the asymmetric unit had an Rfactor of 38.2% for data in predominantly of polar side chains, with three negatively
the 15 Å to 4 Å range. After initial refinement of this model, charged residues on domain D and two positively charged
␴A-weighted 2Fo⫺Fc and Fo⫺Fc electron density maps were residues on the 2A2 Fab buried by the interaction, providing an
examined with the XTALVIEW program (27), which clearly overall electrostatic attraction between the two molecules.
showed electron densities for domain D associated with two of Of the five polar interactions identified between Fab and
the Fabs. Coordinates from all previously reported SpA domain domain D, three are between side chains. A salt bridge is formed
structures were fitted globally into these densities, with the best between Arg-H19 and Asp-36 and two hydrogen bonds are made
fit demonstrated for domain E (28), which then was modified between Tyr-H59 and Asp-37 and between Asn-H82a and
based on the sequence of domain D. For the third Fab, there is Ser-33. There are also two hydrogen bonds that are formed
no electron density at the corresponding site, and crystal packing between main-chain atoms and side-chain atoms, namely Gly-
precludes the placement of a domain D molecule there. H15 carbonyl with Gln-26, and Lys-H57 with Asp-36 carbonyl.
Noncr ystallographic symmetr y restraints of 100 kcal Except for minor variations, all interactions are common to both
mol⫺1䡠Å-2 were applied on main-chain atoms of the three Fab complexes in the asymmetric unit. From the analysis of the
molecules, using the software package XPLOR (29). The progress complex structure it appears that sequence-restricted interac-
of the refinement was judged by the decrease of Rfree after Powell tions dominate the contact surfaces between SpA and the VH
minimization and temperature factor refinement. Initially, two B region.
factors per residue were used in the refinement, whereas in the Domain D superposes well on domain E (28) with an rmsd of
last cycles a single B factor per nonhydrogen atom was refined. 0.52 Å over 196 main-chain atoms. The overall positions of the
The final Rwork and Rfree factors are 21.7% and 28.1%, respec- three helices are very similar with tilt angles of the helix I relative
tively, and the structure displays standard stereochemistry as to the two parallel helices II and III of 15o. This interhelical angle
analyzed by PROCHECK (30) with rms deviations (rmsds) from is identical in domain Z, which was engineered from domain B,
ideality of 0.008 Å on bond lengths and 1.59° on bond angles. but different from domain B itself where it is 30o (31). However,
Detectable domain D residues range from Phe-5 to Ala-56 (see this variation does not affect the relationship between helices II
Fig. 2 A for numbering). A large portion of constant region of and III that are responsible for VH binding. Importantly, each of
one Fab molecule in the asymmetric unit is poorly defined. the other SpA domains shares 75–89% sequence identity with
However, the VH and VL regions, domain D molecules, and in domain D, and there is a high conservation of all Fab interacting
particular the VH-domain D interface, are well defined in the residues in domains E, A, and B (Fig. 2A). For domain C, Gln-40
electron density maps. is replaced by Val but this substitution appears conservative
because at this site the alkyl chain mediates Fab contact, and the
Results and Discussion same can be said for the Asn-43 to Glu substitution. For the
Overall Structure of the Fab-Domain D Complex. The crystal form synthetic domain Z, the naturally occurring Gly-29 in domain B
selected for analysis has two domain D-Fab complexes and one was mutated to Ala to improve the stability to hydroxylamine
unliganded Fab in the asymmetric unit. The conformations of during purification procedures (36). In the solved Fab complex

5400 兩 www.pnas.org Graille et al.

 with domain D, the C␣ of this glycine is less than 3.5 Å away from
A
Gln-H81 and Asn-H82a, predicting that the C␤ of an alanine at
this position would perturb the interaction between the mole-
cules. These observations correlate well with the Fab-binding
activities of the various domains, as the synthetic domain Z is
devoid of this activity whereas all of the natural domains have
similar Fab dissociation constants (Kd) of 0.4–5.0 ⫻ 10⫺6 M (23,
34, 37). Moreover, presumably because of avidity effects, the
strength of the interaction between native pentameric SpA and
native decavalent VH3 IgM is increased by 2–3 orders of mag-
nitude (23).

Dual Reactivity of an Individual SpA Domain with Fc␥ and Fab. The
site responsible for Fab binding is structurally separate from the
domain surface that mediates Fc␥ binding. As first demonstrated
in a crystallographic complex (2) and recently reinvestigated in
NMR studies (38), the interaction of Fc␥ with domain B
primarily involves residues in helix I with lesser involvement of
helix II (Fig. 2 A). With the exception of the Gln-32, a minor
contact in both complexes, none of the residues that mediate the
Fc␥ interaction are involved in Fab binding. The area buried in
the Fc␥-domain B interface is 1,320 Å2, which is comparable to
the 1,220 Å2 buried in the current complex with Fab. However,
B the nature of these buried SpA residues differs significantly, as
the Fab binding is dominated by polar contacts whereas the Fc␥
interaction is predominantly hydrophobic.
To examine the spatial relationship between these different
Ig-binding sites, we superposed the SpA domains in these
complexes to construct a model of a complex between an Fab,
an SpA domain, and an Fc␥ molecule. In this ternary model, we
found an rmsd of 0.73 Å for the backbone atoms in helix I and
helix II of the SpA domains (Fig. 2B). Here, the Fab and Fc␥
Downloaded from https://www.pnas.org by 41.254.43.121 on November 27, 2023 from IP address 41.254.43.121.

form a sandwich about opposite faces of the helix II without

evidence of steric hindrance of either interaction. These findings
illustrate how, despite its small size (i.e., 56–61 aa), an SpA
domain can simultaneously display both activities, explaining
experimental evidence that the interactions of Fc␥ and Fab with
an individual domain are noncompetitive (23, 39).

Interaction Between Domain D Monomers. In the crystallographic

structure, the two molecules of domain D, designated domD-1
and domD-2, interact with one another to form an asymmetric
dimer (Fig. 2C). They are related by a rotation of 148o with an
C interface of 1,300 Å2 composed of equal contributions from each
SpA domain. In both domD-1 and domD-2, this interaction
involves the same 9 aa (Gln-9, Gln-10, Phe-13, Tyr-14, Leu-17,
Asn-18, Asn-28, Ile-31, and Lys-35) that contribute 90% of the
buried surface. Completing this interaction are Glu-24, Arg-27,
and Gln-32 from domD-1 and Phe-5 from domD-2. These
interactions are formed asymmetrically, and further asymmetry
is seen in the hydrogen bond between the side chains of Gln-32
of domD-1 and Asn-18 of domD-2, and even in the hydrogen
bond between the Asn-28 from both domains.
All hydrophobic and aromatic residues implicated in Fc␥
binding are also involved in the van der Waals interactions in the
domain D dimer (Fig. 2A). The surface areas buried in these two
types of interactions are also comparable. Although only two
hydrogen bonds are formed in the dimer, there are four hydro-
gen bonds in the interaction of domain B and Fc␥ (2), suggesting
that under certain conditions the latter interaction may be
favored. However, we believe that this dimer form is unlikely to
Fig. 1. Schematic representation of the complex between SpA domain D and
IMMUNOLOGY

Fab 2A2 from a human IgM. (A) Side view showing SpA domain D (red) bound
to the framework region of the Fab heavy chain (cyan). The VL domain, which SpA domain D and Fab 2A2 involved in the interaction. Kabat numbering is
is not involved in this interaction, is shown in dark blue. The CDR loops as used for the VH residues (blue); domain D is numbered (in brown) with the
defined by Chothia and Lesk (33) are highlighted in magenta. (B) Ribbon convention used for SpA domains (34). Contact residues are identified if 20 Å2
representation of the VH region of Fab showing the positions of the residues or more of their surface are buried in the interface and if they make at least
that interact with domain D. (C) Schematic diagram detailing the residues of one van der Waals contact. All figures were generated by using MOLMOL (35).

Graille et al. PNAS 兩 May 9, 2000 兩 vol. 97 兩 no. 10 兩 5401

 A

B C
Downloaded from https://www.pnas.org by 41.254.43.121 on November 27, 2023 from IP address 41.254.43.121.

Fig. 2. Interactions of individual SpA domains. (A) Alignment of the amino acid sequences of the five SpA domains. Domain D residues involved in interaction
with Fab 2A2 are highlighted in cyan. With the exception of Gln-32 (pink), there is no overlap between the residues involved in Fab interaction and those
mediating Fc␥ binding (2) (gray highlight). The engineered domain Z differs by the key mutation Gly-29 in Ala and does not bind Fab. The residues involved in
the dimer of domain D observed in the asymmetric unit are indicated by red and green boxes. (B) Cross-linking of a VH3 Fab (cyan surface) and a Fc␥ (gray surface)
by a single domain of SpA (red ribbon). This model is based on the superposition of helix I and II of SpA domains in the Fab-domain D complex reported here
and in the previously determined Fc␥-domain B complex (2) (rmsd of 0.73 Å for 140 backbone atoms). (C) Interface between domain D monomers. Schematic
view of the interaction between the two domains D observed in the asymmetric unit, dom-D1 (red ribbon) and dom-D2 (green ribbon). Contact residues from
both domains are shown in stick representation.

represent a random relationship, as the dimensions of the two of the VH residues involved in the interaction, Lys-H57 and
associated interface falls in the upper 3% of what can be Tyr-H59, originally were assigned to CDR2 based on primary
considered random crystal contacts (40). Recent NMR studies sequence hypervariability comparisons, these residues are not
strengthen our hypothesis, as weak homodimer interactions have part of loop H2 and are in fact part of the C⬙ strand (33) with
been noted for domain B with the involvement of residues from side chains pointing away from the combining site (Fig. 1 A).
helix I and II (41), which is in good agreement with the Moreover, in the complex, domain D is more than 10 Å away
crystallographic dimer. Considering that the residues involved in from VH CDR3, the loop most commonly contributing contacts
this dimerization are all conserved in each of the five extracel- for antigen recognition and specificity, which is consistent with
lular domains of SpA, should such an interaction occur between the lack of correlation between SpA binding activity and CDR3
the proximal domains of membrane-associated molecules of sequence (7, 8). Based on superposition of the Fab-domain D
native SpA, the functional valency for binding of lymphocyte complex onto reported Fab-antigen complexes, and based on the
membrane-associated B-cell antigen receptor potentially could observation that complexed and uncomplexed Fab molecules in
be enhanced. The biological relevance of this postulated inter- the crystal have identical conformations, it is unlikely that the
action merits further investigation. binding of domain D would alter the combining site or result in
steric hindrance so that it would prevent the binding of an
Accessibility of the Antigen Combining Site in the Fab-Domain D antigen, including very large ones (data not shown). Further-
Complex. The crystal structure of the Fab-domain D complex more, because of the spatial relationships between the termini of
accounts for functional evidence that SpA does not compete each of the domains (Fig. 1 A) and their associated VH, even
with antigen binding (42), as the VH binding surface does not binding of intact five-domain SpA would appear unlikely to
involve the complementarity determining region (CDR) loops interfere with antigen binding. We therefore conclude that the
that mediate binding of conventional antigens (33, 43). Although manner by which SpA is bound to the VH does not disturb Fab

5402 兩 www.pnas.org Graille et al.

 Table 1. Sequences of human VH gene families at the positions involved in the interface with domain D
H15 H17 H19 H57 H59 H64 H65 H66 H68 H70 H81 H82a H82b

VH1 (11) G S, T K T, A Y Q G, D, E R T T E S, R S, R
VH2 (3) T T T K Y K S, T R T T, S T T N
VH3 (22) G S R, K K, I, T Y K G R T S Q N, G S
VH4 (11) S T S T Y K S R T S K S S
VH5 (2) G S K, R T Y Q G Q, H T S Q S S
VH6 (1) S T S N Y K S R T N Q N S
VH7 (1) G S K P Y T G R V S Q C S

The numbers in parentheses for each VH group are the numbers of human germ-line V gene sequences reported in the V Base database (Medical Research
Council, Centre for Protein Engineering). The VH3 family represents nearly half of all inherited human VH gene segments. Amino acids in italics are represented
in less than 10% of functional inherited human genes.

conformation and does not appear to affect accessibility to the of adult VH3 IgM and a greater proportion of VH3 IgG lack this
antigen binding site. reactivity (5–7). In almost every case, the loss of binding activity
now can be correlated with identifiable nonconservative muta-
VH Specificity of the SpA Binding Interaction. The structure of the tions at one or more of the VH positions at the interface (data
VH site responsible for SpA binding explains the high frequency not shown). Although some somatic mutations can be tolerated
of these binding interactions in the human immune system. at position H57, the inversion of charge from Lys to Glu (46)
About half of a panel of human IgM, and 32–54% of human results in loss of activity because of electrostatic repulsion.
peripheral blood B cells, have been reported to interact with SpA Although the change from Asn to Ser at position H82a (47) can
(5, 8, 44, 45). Among human antibodies studied to date, SpA be considered conservative, this position is part of the core and
binding has been restricted to products of the VH3 gene family hence even such a minor change leads to reduced activity.
with no examples from any of the other six families (5–8). Among
the 13 VH residues implicated as SpA contacts, VH3 genes Structural Comparisons with the Interactions of T-Cell Superantigens.
include frequent germ-line sequence variations only at position The capacity of SpA to bind a large proportion of the human
H57 (Table 1 and Fig. 1B). This is not a core residue in the immune repertoire, expressed either as receptors on B-cells (i.e.,
interface, and Lys, Ile, and Thr are permissive at this position (7, B-cell antigen receptor) or in soluble Ig form, has parallels with
8). The other 12 VH residues in direct contact with domain D are the abilities of Staphylococcus aureus enterotoxins and other
highly conserved in members of this family. In fact, in the 22 T-cell superantigens to interact with a large repertoire of T
Downloaded from https://www.pnas.org by 41.254.43.121 on November 27, 2023 from IP address 41.254.43.121.

potentially functional human VH3 gene segments, there are only lymphocytes through certain TcR V␤. By virtue of their strong
two germ-line variations (the conservative change of Lys at H19 structural similarities, the VH of Fab 2A2 bound to domain D can
in V3–73 and the nonconservative change of Gly at H82a in be superposed onto the V␤ of TcR complexed with the Staph-
V3–64) but each of these genes are the source of less than 2% ylococcus aureus enterotoxin, SEC3, (48) with an rmsd of 1.6 Å
of adult expressed VH3 Ig repertoire (E. Milner, personal over 393 main-chain atoms (Fig. 3). Although in both cases they
communication). In the mouse, homologous genes in the S107, are required for antigen recognition, the partner chains in these
J606, 7183, and VH10兾DNA4 families also commonly encode for heterodimeric receptors, the VL of the B-cell antigen receptor
SpA binding (9, 10). With the exception of the conservative H19 and V␣ of the TcR, do not directly interact with domain D or
Lys substitution (15), contact residues also are conserved in the SEC3, respectively. Overall, SEC3 is positioned closer to the
murine homologues. Some of these related murine VH genes, antigen binding site of the TcR. Only about 25% of the V␤ chain
including certain 7183 genes, contain a Ser or Thr substitution contacts derive from the ␤-strands, whereas the remainder are
at H82a that correlates with weaker SpA binding activity. from the CDR1, CDR2, and HV4 loops. For characterized
VH10兾DNA4 genes also are associated with low binding activity
and include an Asp at H65. By comparison, inherited genes of the
other nonbinding human and murine VH families each contain
two or more residue differences at the 13 VH positions identified
as SpA contacts (15). In particular, for the two other large
human families, VH1 and VH4, there are nonconservative dif-
ferences at position H82a (Table 1). VH4 genes also have a
nonconservative change at H19. From a structural point of view,
a core of critical residues can be defined, consisting of seven
residues: Arg兾Lys-H19, Gly-H65, Arg-H66, Thr-H68, Ser-H70,
Gln-H81, and Asn-H82a (Fig. 1B). Apart from the essential salt
bridge formed at position H19, all other residues are virtually
inaccessible to solvent, making their replacement by a larger
residue highly disruptive. Replacement by a smaller residue, as
for Asn-H82a by Ser, is likely to lead to weaker binding. Hence
from such considerations, we conclude that this core of seven VH
residues constitutes the structural motif for SpA binding, and this
conveys the restricted specificity for VH3-encoded Ig and their
IMMUNOLOGY

Fig. 3. Comparison of the interactions of B-cell and T-cell superantigens. The V␤

homologues.
region (yellow) of the TcR superimposed well on VH (cyan) of Fab 2A2 with an
The structure offers the possibility of rationalizing the influ- rmsd of 1.6 Å over 393 main-chain atoms. The T-cell superantigen, enterotoxin
ence of somatic hypermutations in VH3-encoded IgM upon SEC3, (green) binds to the CDR1, CDR2, and HV4 loops of the V␤ region. SpA (red)
binding reactivity with SpA. Whereas all tested VH3-germ-line- binds at framework region 1 and 3, and the carboxyl-terminal portion of the
encoded IgM were found to be reactive with SpA (8), about 20% CDR2 (including positions H57 and H59 in the C⬙ strand) of the VH region.

Graille et al. PNAS 兩 May 9, 2000 兩 vol. 97 兩 no. 10 兩 5403

 staphylococcal T-cell superantigens (21, 48, 49), the interaction equivalent sites in B-cell and T-cell receptors. Therefore, despite
with the TcR also differs radically from the domain D-Fab prominent differences in the molecular basis of these interac-
complex as it is mediated primarily through hydrogen bonds with tions, they both act to stimulate large proportions of the host’s
the V␤ region backbone atoms that are contacted in a defined B cells or T cells, resulting in subversion of the immune system.
conformationally sensitive distribution. Because this interaction
is permissive of amino acid side-chain variations, the promiscuity
of these T-cell superantigens enables the targeting of the cellular We thank M. Léonetti for helpful discussions and help in the preparation
of the manuscript, M. Hamon and G. Konfortova for cell culture, D.
products of different TcR V␤ families, each of which generally
Beale for purification of 2A2 IgM and preparation of the Fab fragment,
include only 1–2 V␤ genes. In contrast, the SpA-Fab interaction
and L. Luo for protein expression and purification. The 2A2 cell line was
clearly relies on a sequence-restricted binding mode with a kindly provided by Prof. R. Maini (Kennedy Institute, London). We
structural motif presented by the side chains of highly conserved thank M.-H. LeDu and M. Knossow for reading of the manuscript, W.
VH residues. Because these residues are part of a large set of VH Shepard and A. Bentley for assistance at the Laboratoire pour
genes that are highly expressed in the immune system, SpA l’Utilisation du Rayonnement Electromagnétique synchrotron facility
targets a large proportion of the B-cell pool. (Orsay), and A. Ménez for continuous support. B.J.S. and M.J.T.
In conclusion, our studies demonstrate that by exploiting a acknowledge support from the Arthritis Research Campaign (U.K.) and
conformational surface on the antigen receptor well represented are also supported by the Biotechnology and Biological Sciences Re-
in the repertoire, a bacterial virulence factor targets host B search Council (U.K.). G.J.S. was supported by National Institutes of
lymphocytes and their soluble Ig products. Staphylococcus aureus Health Grants R01-AI40305, 5P60-AR40770, and R03-AI46637–01, and
has developed proteins that interact with analogous but not fully a Biomedical Sciences Award from the Arthritis Foundation.

1. Boyle, M. D. P. (1990) in Bacterial Immunoglobulin-Binding Proteins, ed. Boyle, 24. Stura, E. A. & Wilson, I. A. (1991) J. Crystallogr. Growth 110, 270–282.
M. P. D. (Academic, San Diego), Vol. 1, pp. 17–28. 25. Otwinowski, Z. & Minor, W. (1997) Methods Enzymol. 276, 307–326.
2. Deisenhofer, J. (1981) Biochemistry 20, 2361–2370. 26. Navaza, J. (1994) Acta Crystallogr. A 50, 157–163.
3. Tashiro, M. & Montelione, G. T. (1995) Curr. Opin. Struct. Biol. 5, 471–481. 27. McRee, D. E. (1999) J. Struct. Biol. 125, 156–165.
4. Vidal, M. A. & Conde, F. P. (1985) J. Immunol. 135, 1232–1238. 28. Starovasnik, M. A., Skelton, N. J., O’Connell, M. P., Kelley, R. F., Reilly, D.
5. Sasso, E. H., Silverman, G. J. & Mannik, M. (1989) J. Immunol. 142, 2778–2783. & Fairbrother, W. J. (1996) Biochemistry 35, 15558–15569.
6. Sasso, E. H., Silverman, G. J. & Mannik, M. (1991) J. Immunol. 147, 1877–1883. 29. Brünger, A. T. (1992) X-PLOR Manual (Yale Univ. Press, New Haven, CT),
7. Sasano, M., Burton, D. R. & Silverman, G. J. (1993) J. Immunol. 151, Version 3.1.
5822–5839. 30. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993)
8. Hillson, J. L., Karr, N. S., Oppliger, I. R., Mannik, M. & Sasso, E. H. (1993) J. Appl. Crystallogr. 26, 238–291.
J. Exp. Med. 178, 331–336. 31. Gouda, H., Torigoe, H., Saito, A., Sato, M., Arata, Y. & Shimada, I. (1992)
9. Seppala, I., Kaartinen, M., Ibrahim, S. & Makela, O. (1990) J. Immunol. 145, Biochemistry 31, 9665–9672.
2989–2993. 32. Conte, L. L., Chothia, C. & Janin, J. (1999) J. Mol. Biol. 285, 2177–2198.
10. Cary, S., Krishnan, M. R., Marion, T. & Silverman, G. J. (1999) Mol. Immunol. 33. Chothia, C. & Lesk, A. M. (1987) J. Mol. Biol. 196, 901–917.
Downloaded from https://www.pnas.org by 41.254.43.121 on November 27, 2023 from IP address 41.254.43.121.

36, 769–776. 34. Jansson, B., Uhlen, M. & Nygren, P. A. (1998) FEMS Immunol. Med. Microbiol.
11. Romagnani, S., Giudizi, M. G., del Prete, G., Maggi, E., Biagiotti, R., 20, 69–78.
Almerigogna, F. & Ricci, M. (1982) J. Immunol. 129, 596–602. 35. Koradi, R., Billeter, M. & Wuthrich, K. (1996) J. Mol. Graphics 14, 29–32.
12. Kristiansen, S. V., Pascual, V. & Lipsky, P. E. (1994) J. Immunol. 153, 36. Nilsson, B., Moks, T., Jansson, B., Abrahmsen, L., Elmblad, A., Holmgren, E.,
2974–2982. Henrichson, C., Jones, T. A. & Uhlen, M. (1987) Protein Eng. 1, 107–113.
13. Kozlowski, L. M., Kunning, S. R., Zheng, Y., Wheatley, L. M. & Levinson, A. I. 37. Silverman, G. J., Pires, R. & Bouvet, J. P. (1996) J. Immunol. 157, 4496–4502.
(1995) J. Clin. Immunol. 15, 145–151. 38. Gouda, H., Shiraishi, M., Takahashi, H., Kato, K., Torigoe, H., Arata, Y. &
14. Silverman, G. J., Nayak, J. V., Warnatz, K., Hajjar, F. F., Cary, S., Tighe, H. Shimada, I. (1998) Biochemistry 37, 129–136.
& Curtiss, V. E. (1998) J. Immunol. 161, 5720–5732. 39. Starovasnik, M. A., O’Connell, M. P., Fairbrother, W. J. & Kelley, R. F. (1999)
15. Cary, S., Lee, J., Wagenknecht, R. & Silverman, G. J. (2000) J. Immunol. 164, Protein Sci. 8, 1423–1431.
4730–4741. 40. Janin, J. (1997) Nat. Struct. Biol. 12, 673–674.
16. Foster, T. J., O’Reilly, M., Patel, A. H. & Bramley, A. J. (1988) Antonie Van 41. Karimi, A., Matsumura, M., Wright, P. E. & Dyson, H. J. (1999) J. Pept. Res.
Leeuwenhoek 54, 475–482. 54, 344–352.
17. Patel, A. H., Nowlan, P., Weavers, E. D. & Foster, T. (1987) Infect. Immun. 55, 42. Young, W. W., Jr., Tamura, Y., Wolock, D. M. & Fox, J. W. (1984) J. Immunol.
3103–3110. 133, 3163–3166.
18. Kotzin, B. L., Leung, D. Y., Kappler, J. & Marrack, P. (1993) Adv. Immunol. 43. Wilson, I. A. & Stanfield, R. L. (1994) Curr. Opin. Struct. Biol. 4, 857–867.
54, 99–166. 44. Hakoda, M., Kamatani, N., Hayashimoto-Kurumada, S., Silverman, G. J.,
19. Bohach, G. A., Fast, D. J., Nelson, R. D. & Schlievert, P. M. (1990) Crit. Rev. Yamanaka, H., Terai, C. & Kashiwazaki, S. (1996) J. Immunol. 157, 2976–2981.
Microbiol. 17, 251–272. 45. Silverman, G. J., Sasano, M. & Wormsley, S. B. (1993) J. Immunol. 151,
20. Papageorgiou, A. C., Tranter, H. S. & Acharya, K. R. (1998) J. Mol. Biol. 277, 5840–5855.
61–79. 46. Randen, I., Potter, K. N., Li, Y., Thompson, K. M., Pascual, V., Forre, O.,
21. Li, H., Llera, A., Malchiodi, E. L. & Mariuzza, R. A. (1999) Annu. Rev. Natvig, J. B. & Capra, J. D. (1993) Eur. J. Immunol. 23, 2682–2686.
Immunol. 17, 435–466. 47. Zhang, M., Majid, A., Bardwell, P., Vee, C. & Davidson, A. (1998) J. Immunol.
22. Sohi, M. K., Sutton, B. J., Corper, A. L., Wan, T., Maini, R. N., Brown, C., 161, 2284–2289.
Rijnders, T., Beale, D., Feinstein, A., Humphreys, A. S. & Taussig, M. J. (1994) 48. Fields, B. A., Malchiodi, E. L., Li, H., Ysern, X., Stauffacher, C. V., Schlievert,
J. Mol. Biol. 242, 706–708. P. M., Karjalainen, K. & Mariuzza, R. A. (1996) Nature (London) 384, 188–192.
23. Roben, P. W., Salem, A. N. & Silverman, G. J. (1995) J. Immunol. 154, 49. Li, H., Llera, A., Tsuchiya, D., Leder, L., Ysern, X., Schlievert, P. M.,
6437–6445. Karjalainen, K. & Mariuzza, R. A. (1998) Immunity 9, 807–816.

5404 兩 www.pnas.org Graille et al.