Disulfide bond-mediated dimerization of HLA-G on
the cell surface
Jonathan E. Boyson*, Robert Erskine*, Mary C. Whitman*, Michael Chiu*, Julie M. Lau*, Louise A. Koopman*,
Markus M. Valter*†, Pavla Angelisova‡, Va ´clav Horejsı ´‡, and Jack L. Strominger*§
*Department of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138; and‡Institute of Molecular Genetics, 142 20 Prague,
Contributed by Jack L. Strominger, October 22, 2002
HLA-G is a nonclassical class I MHC molecule with an unknown
function and with unusual characteristics that distinguish it from
other class I MHC molecules. Here, we demonstrate that HLA-G
forms disulfide-linked dimers that are present on the cell surface.
Immunoprecipitation of HLA-G from surface biotinylated transfec-
tants using the anti-?2-microglobulin mAb BBM.1 revealed the
presence of an ?78-kDa form of HLA-G heavy chain that was
reduced by using DTT to a 39-kDa form. Mutation of Cys-42 to a
serine completely abrogated dimerization of HLA-G, suggesting
that the disulfide linkage formed exclusively through this residue.
A possible interaction between the HLA-G monomer or dimer and
the KIR2DL4 receptor was also investigated, but no interaction
between these molecules could be detected through several ap-
proaches. The cell-surface expression of dimerized HLA-G mole-
cules may have implications for HLA-G?receptor interactions and
for the search for specific receptors that bind HLA-G.
molecules. Its expression is largely restricted to the invasive
extravillous trophoblasts of the decidua (1, 2), it exhibits limited
polymorphism (3, 4), and it possesses a truncated cytoplasmic
domain (5). Based on its pattern of expression, HLA-G is
assumed to play some role at the maternal–fetal interface, but its
specific function remains unclear. HLA-G expression has also
been reported on some tumors, suggesting a possible role in
tumor evasion (6, 7).
Several roles for HLA-G in the decidua are possible. One role
may be to provide a signal sequence-derived peptide for binding
by HLA-E (also expressed on trophoblasts), a high-affinity
ligand for the inhibitory CD94?NKG2A receptor (8). Another
role may be to bind KIR2DL4, an activating receptor expressed
on NK cells (9–12); lastly, HLA-G may bind the inhibitory ILT2
and ILT4 receptors (13, 14).
Certain aspects of HLA-G suggest that it may function in a
manner distinct from other class I MHC molecules. The HLA-G
promoter, for example, is divergent from other class I MHC
promoters, which may help to explain its pattern of expression
(15). HLA-G also has an unusually long half-life on the cell
surface, the result of the absence of an endocytosis motif in its
truncated cytoplasmic domain (16, 17). Interestingly, this trun-
cation also reveals a dilysine motif that acts to recycle HLA-G
from the Golgi back to the endoplasmic reticulum, thus provid-
ing an apparent means for the binding by HLA-G of high-affinity
Here, another unusual aspect of HLA-G is demonstrated: its
ability to form disulfide-linked dimers both in vitro and on the
cell surface. By using site-directed mutagenesis, it is shown that
the dimers form via Cys-42 of the HLA-G ?1 domain. The
formation of a dimerized form of HLA-G may have implications
in identifying its functional receptors.
LA-G is an unusual class I MHC molecule possessing
several characteristics that distinguish it from other MHC
Materials and Methods
Cell Lines and Antibodies. The 721.221 B-LCLs2transfected with
various class I MHC cDNAs have been described (18, 19). The
BW5147.3 mouse thymoma (BW?), and BBM.1 [anti-?2-
microglobulin (?2m)] and W6?32 (conformation-sensitive pan-
HLA) hybridomas were obtained from American Type Culture
Collection. The anti-KIR2DL1 mAb EB6 was obtained from
Beckman Coulter. Monoclonal antibodies to HLA-G were pre-
pared by standard procedures from mice immunized with bac-
terially produced heavy chain, either fully denatured (MEM-
G?1, IgG1) or refolded to produce native complexes with ?2m
and peptide (MEM-G?11, IgG1; see below). KIR2DL4?
KIR2DL5-reactive rabbit antiserum was generated by injecting
rabbits s.c. with 100 ?g of KIR2DL5-Ig fusion protein in
complete Freund’s adjuvant, followed 14 days later by a second
injection s.c. in incomplete Freund’s adjuvant (Serasource, Roy-
alston, MA). This polyclonal antibody recognizes both
KIR2DL4 and KIR2DL5.
Protein Expression. Soluble HLA-G was PCR-amplified from
JEG-3 choriocarcinoma cell line cDNA and subcloned into the
pGMT7 expression vector. The HLA-Cw6 and ?2m expression
constructs have been described (20). Soluble recombinant pro-
teins were produced in Escherichia coli BL21 (DE3) pLysS
(Novagen) and refolded as described (20). HLA-G was refolded
with the peptide, VLPKLYVKL, and HLA-Cw6 was refolded
with the peptide, YQFTGIKKY.
KIR-Ig fusion constructs were constructed by fusing the signal
sequence and extracellular domains of KIRs to the Fc portion of
human IgG1 (derived from the CD51neg1, a gift from B. Seed,
Massachusetts General Hospital, Boston) and by subcloning into
the pCDNA3 expression plasmid (Invitrogen). KIR-Ig plasmids
were transfected into COS-7 cells using Fugene 6 (Roche
Molecular Biochemicals) and purified over a Protein G?Poros20
column using perfusion chromatography (PerSeptive Biosys-
tems, Framingham, MA). Fusion proteins were eluted in 2%
acetic acid and immediately neutralized with 1 M Tris, pH 8.0,
followed by dialysis against PBS, pH 7.4.
SDS?PAGE and Western Blotting. Cells were lysed directly in SDS?
PAGE sample buffer containing freshly prepared 50 mM
iodoacetamide (Sigma). Purified inclusion bodies in 8 M urea
were resuspended in SDS?PAGE sample buffer containing 50
mM DTT. Samples were boiled for 5 min, run on SDS?PAGE,
and blotted to nitrocellulose. Heavy chain was detected by using
the MEM-G?1 mAb, followed by horseradish peroxidase-
conjugated goat anti-mouse IgG mAb (Jackson ImmunoRe-
search). Blots were visualized by using enhanced chemilumines-
cence (ECL; Amersham Pharmacia).
Immunohistochemistry. Decidual tissue was identified macroscop-
ically and washed in PBS, pH 7.2. Decidual pieces were fixed
Abbreviations: ?2m, ?2-microglobulin; MM, molecular mass.
†Present address: Clinics for Obstetrics and Gynecology, University of Cologne Medical
School, 50931 Cologne, Germany.
§To whom correspondence should be addressed at: Harvard University, Fairchild Biochem-
istry Building, 7 Divinity Avenue, Cambridge, MA 02138. E-mail: firstname.lastname@example.org.
December 10, 2002 ?
vol. 99 ?
no. 25 www.pnas.org?cgi?doi?10.1073?pnas.212643199
overnight at 4°C in 4% (wt?vol) paraformaldehyde. Fixed tissue
was dehydrated in an ethanol?xylene series and embedded in
paraffin. Then, 6-?m sections were rehydrated and incubated in
3% (vol?vol) H2O2in methanol to block endogenous peroxidase
activity. MEM-G?1 mAb staining was detected by using a
biotinylated secondary mAb and horseradish peroxidase-
conjugated streptavidin. Staining was visualized by using diami-
Cell-Surface Labeling and Immunoprecipitation. Cells were surface
biotinylated by first washing with PBS, pH 7.2, supplemented
with 0.1 mM CaCl2?1 mM MgCl2and then incubated for 30 min
at 4°C in 0.5 mg?ml sulfo-NHS-biotin (Pierce). Cells then were
washed and lysed in lysis buffer [50 mM Tris, pH 7.6?150 mM
NaCl?5 mM EDTA?1% (vol?vol) Nonidet P-40 and protease
inhibitors (Roche Molecular Biochemicals)]. When iodoacet-
amide was used, biotinylated cells were split into two aliquots.
One aliquot was lysed with lysis buffer alone and one with lysis
buffer supplemented with 50 mM iodoacetamide (freshly pre-
pared). Cell lysates were precleared with protein A-Sepharose
beads coated with normal rabbit serum and immunoprecipitated
overnight at 4°C with BBM.1-coated beads. After washing, the
beads were divided into two samples, and one was boiled in
SDS?PAGE buffer and one in SDS?PAGE buffer containing 50
mM DTT. Eluted protein was run on SDS?PAGE gels and
blotted onto nitrocellulose. Biotinylated proteins were detected
by using HRP-conjugated streptavidin (Roche Molecular Bio-
chemicals) and visualized by ECL.
Mutagenesis. The HLA-A2 crystal structure was visualized by
using RASMOL. Cys-42 was mutagenized to serine by using
site-directed mutagenesis according to the manufacturer’s in-
structions (Stratagene). The PAGE-purified primer Gmut1:
G-3? and its reverse complement were used for mutagenesis.
KIR-? Fusion Constructs and Transfectants. A KIR2DL4-? fusion
protein in which the extracellular domain of KIR2DL4 was fused
to the transmembrane and cytoplasmic domains of CD3-? was
generated by overlapping PCR. The KIR2DL4 extracellular
domain was amplified by using 5?K103Kpn, 5?-GGCGGTAC-
CATGTCCATGTCACCCACG-3?, and 3?K103-?, 5?-CAGGC-
CAAAGCTCTGATGCAGGTGTCTGGC-3?. ?-chain was
amplified by using 5?K103-? 5?-GCCAGACACCTGCATCA-
GAGCTTTGGCCTG-3? and 3??, 5?-CGGAATTCCGTTAGC-
GAGGGGCAGG-3?. The gel-purified templates were reampli-
fied in a second PCR by using 5?K103Kpn and 3?? and cloned
into the pEF6 expression vector (Invitrogen). KIR-? constructs
were electroporated into mouse BW cells (250 V, 500 ?F), and
stable transfectants were established.
BW?KIR-? transfectants were cocultured in 96-well round-
bottom plates with ?-irradiated (3,000 rad) stimulators for 72 h.
As positive controls, BW?KIR-? transfectants were incubated in
wells coated with either 10 ?g?ml EB6 (for KIR2DL1-?) or with
20 ?g?ml purified anti-KIR2DL4?KIR2DL5 antiserum (for
KIR2DL4-?). Supernatants were tested for mouse IL-2 by using
ELISA (BioSource International, Camarillo, CA).
Surface Plasmon Resonance. Goat anti-human Fc receptor anti-
body (Jackson ImmunoResearch) was coupled to CM5 chips by
using standard amine coupling according to the manufacturer’s
instructions (Biacore, Piscataway, NJ). KIR2DL1-Ig and
KIR2DL4-Ig fusion proteins (30 ?g?ml) were then bound to the
surface at a flow rate of 5 ?l?min. Recombinant class I MHC
molecules were purified by gel filtration just before use, diluted
in HBS-EP running buffer, and flowed over the chip at a flow
rate of 5 ?l?min. Both mock-coupled and anti-human Fc-
coupled surfaces were used as controls. Binding data were
evaluated by using BIAEVALUATION V.3.0 (Biacore).
Generation and Characterization of Anti-HLA-G mAbs. To facilitate
the examination of HLA-G in vitro and in vivo, HLA-G-specific
mAbs were generated against both denatured heavy chain
(MEM-G?1 mAb) and HLA-G that had been refolded with ?2m
and peptide (MEM-G?11 mAb). MEM-G?1 was specific for
HLA-G heavy chain in Western blots (Fig. 1A), and it stained
cytospin preparations of JEG-3 (HLA-G?) but not JAR (HLA-
G?) trophoblast cell lines (Fig. 1B). Furthermore, MEM-G?1
was specific for the extravillous trophoblast population when
used in immunohistochemical staining of paraffin-embedded
placental sections (Fig. 1C). Immunoprecipitation of a panel of
cell lines and HLA transfectants demonstrated that MEM-G?11
recognized HLA-G specifically (Fig. 1D). The HLA-G specific-
ity of MEM-G?11 was confirmed by using fluorescence activated
cell sorter (FACS) analysis of a panel of HLA-transfected cell
lines (data not shown).
In Vitro Dimerization of HLA-G. Repurification by gel filtration of
refolded HLA-G??2m?peptide heterotrimers (designated
HLA molecules detected by MEM-G?1 mAb. (B) Cytospin preparations of the
HLA-G?choriocarcinoma cell line JEG-3 and the HLA-G?cell line JAR were
stained with the MEM-G?1 mAb. Whereas JAR is completely negative, HLA-G
staining (brown) can be seen clearly in the JEG-3 sample. Note that not all of
the JEG-3 cells stain positively, which is consistent with the heterogeneous
nature of this tumor line. (C) Immunohistochemistry of paraffin-embedded
detected on the cell islands (denoted by arrows) that had migrated from the
villi. (D) The MEM-G?11 mAb was used to immunoprecipitate HLA-G from
surface-biotinylated B-lymphoblastoid cell lines and from various class I MHC
transfected into the MHC-deficient 721.221 B-lymphoblastoid cell line. ECL
visualization demonstrated that only the two different clones of HLA-G?
transfectants were recognized.
Characterization of HLA-G-specific mAbs. (A) Western blot of bacte-
Boyson et al.
December 10, 2002 ?
vol. 99 ?
no. 25 ?
throughout as HLA-G monomer), resulted in the serendipitous
discovery of two peaks, one corresponding to the expected
molecular mass (MM) of 43 kDa and one approximately twice
that. These data suggested that HLA-G dimerization had oc-
curred. To confirm this, HLA-G was refolded with ?2m and
peptide and purified to homogeneity by using gel filtration (Fig.
2A). The purified HLA-G monomer was kept at 4°C for 21 days,
and the sample was analyzed again by gel filtration. After the
incubation, the majority (?75%) of the refolded HLA-G eluted
earlier than the HLA-G monomer. The early eluting peak
corresponded to a MM of approximately twice that of the
monomeric form of HLA-G (Fig. 2B). Analysis of the fractions
by SDS?PAGE under nonreducing conditions confirmed that
this new peak contained dimerized (?66 kDa) and monomeric
(?33 kDa) HLA-G heavy chain. The HLA-G dimers could be
the dimers were disulfide-linked. Thus, HLA-G was capable of
forming disulfide-linked dimers in vitro.
HLA-G Dimerization on the Cell Surface. To investigate whether
HLA-G dimerization occurred naturally, HLA-G transfectants
of 721.221 B-LCLs were lysed, run on SDS?PAGE gels under
both nonreducing and reducing conditions, and blotted to
nitrocellulose. Detection of HLA-G heavy chains using the
HLA-G-specific mAb MEM-G?1 demonstrated the presence of
two bands under nonreducing conditions, one of ?39 kDa, the
expected MM of the HLA-G heavy chain, and the other of ?78
kDa, the expected MM of dimerized HLA-G heavy chain (Fig.
3A). Under reducing conditions, this higher MM band was not
present, suggesting that it was disulfide-bonded HLA-G heavy
chain. However, only a small fraction of the entire pool of
HLA-G heavy chains existed in the dimerized form.
10?30. Arrows depict MM calibration standards. (Inset) A nonreducing SDS?PAGE gel of the fractions before pooling. Because it is a soluble truncated form of
HLA-G, the heavy chain has an expected MM of 32 kDa, and the refolded HLA-G monomer has an expected MM of ?44 kDa. (B) Chromatogram of HLA-G after
a 21-day incubation at 4°C run on a Superdex 200 10?30. Approximately 25% of the sample eluted in a volume corresponding to a MM of ?43 kDa and 75%
Soluble HLA-G dimerizes in vitro. (A) Chromatogram of recombinant soluble HLA-G analyzed by size-exclusion chromatography using a Superdex 200
iodoacetamide and run under nonreducing and reducing conditions. After blotting to nitrocellulose, HLA-G heavy chains were detected by using the
HLA-G-specific mAb, MEM-G?1. (B) Immunoprecipitation of cell surface-biotinylated HLA-A2 molecules from 721.221?HLA-A2 transfectants. Inclusion of
iodoacetamide abrogated the formation of HLA-A2 dimer artifacts, because under nonreducing conditions the ?85-kDa MM band does not appear in samples
were lysed in the presence or absence of iodoacetamide. Class I MHC molecules were immunoprecipitated with BBM.1 and run under both nonreducing and
reducing conditions. Even in the presence of iodoacetamide, HLA-G dimers were detected.
Cell-surface dimerization of HLA-G. (A) 721.221?HLA-G transfectants and the 721.221 parental cell line were lysed in SDS?PAGE buffer containing
www.pnas.org?cgi?doi?10.1073?pnas.212643199Boyson et al.
To determine whether HLA-G??2m?peptide heterotrimers
existed in a dimerized form before cell lysis, cells were lysed in
the presence and absence of iodoacetamide, which carboxy-
methylates free cysteines and blocks formation of disulfide
bonds. First, in a control experiment, HLA-A2 was immuno-
precipitated by using anti-?2m mAb (BBM.1) beads from surface-
biotinylated HLA-A2 transfectants in the presence and absence
of iodoacetamide. As reported (21), dimerized HLA-A2 heavy
chains were observed under nonreducing conditions in the
absence of iodoacetamide, but this dimerization was completely
abrogated by the addition of iodoacetamide (Fig. 3B). Thus, the
observed disulfide-bonded HLA-A2 dimers were an artifact of
cell lysis and immunoprecipitation, and the inclusion of iodoac-
etamide in the lysis buffer allowed us to distinguish between
preexisting and artifactual class I MHC dimers. Similarly, when
HLA-G was immunoprecipitated from surface-biotinylated
HLA-G transfectants, only HLA-G dimers were detected under
nonreducing conditions even in the presence of iodoacetamide
(Fig. 3C), suggesting that they were preexisting and did not form
after cell lysis, and that HLA-G is present on the surface of these
transfectants only as the dimer. Immunoprecipitations with the
conformation-specific W6?32 mAb and the HLA-G-specific
MEM-G?11 mAb yielded similar results (data not shown). Thus,
disulfide-linked dimers of HLA-G??2m?peptide are expressed
on the cell surface of 721.221?HLA-G transfectants. When this
experiment was repeated with the HLA-G?JEG-3 choriocar-
cinoma cell line, no dimers were observed (data not shown).
However, the level of HLA-G expression was much (one log)
lower than on the transfectant (data not shown), suggesting that
dimerization may depend on the cell-surface density of HLA-G.
Cell-Surface Dimerization of HLA-G Is Mediated Through Disulfide
Bonds of Cys-42 of the Heavy Chain. Examination of the predicted
amino acid sequence of HLA-G revealed the presence of two
free cysteines, Cys-42 and Cys-147, not normally found in class
I MHC molecules. Superimposition of these residues onto the
between the third and fourth ? strands of the ?1 domain,
normally occupied in HLA-A2 by a tryptophan pointing into the
groove. Therefore, because of its accessibility, Cys-42 was the
most likely participant in an intermolecular disulfide linkage.
Site-directed mutagenesis was performed, mutating Cys-42 to
Ser-42 (the amino acid occurring at position 42 in HLA-A2 and
in most other class I MHC molecules; Fig. 4A), and the resulting
HLA-G?C42S construct was transfected into 721.221 B-LCLs.
FACS analysis demonstrated that HLA-G?C42S was ex-
pressed at the cell surface and could be detected by using both
the conformation-specific pan-class I MHC mAb W6?32 and the
HLA-G-specific mAb MEM-G?11 (Fig. 4B). HLA-G?C42S
transfectants were FACS-sorted to ensure that they expressed
similar levels of class I MHC at the cell surface as the HLA-G
transfectants used in the previous experiments. Interestingly,
mutation of Cys-42 to Ser-42 completely abrogated dimerization
of HLA-G in the absence of iodoacetamide (Fig. 4C). Thus,
dimerized HLA-G exists on the cell surface linked by means of
a Cys-42-mediated disulfide bond.
HLA-G–KIR2DL4 Interactions. The existence of a dimerized form of
HLA-G on the cell surface may have implications for the
identification of HLA-G receptors. HLA-G has been reported to
bind ILT-2 and KIR2DL4. Attempts by us, however, to detect an
interaction between HLA-G and KIR2DL4-Ig fusion proteins as
reported (9, 10) have been unsuccessful (unpublished results).
To explore the HLA-G–KIR2DL4 interaction further, two
additional approaches were taken. First, KIR2DL4-? and
KIR2DL1-? chain fusion proteins were constructed in which the
extracellular domains of the KIRs were fused to the transmem-
brane and cytoplasmic domain of the CD3-? chain. These
constructs then were stably transfected into the BW?mouse
thymoma line. Anti-KIR2DL1 (mAb EB6)–coated and anti-
KIR2DL4?KIR2DL5-coated wells were used as positive con-
trols for the ability of the BW transfectants to signal. Whereas
BW??KIR2DL1 cells secreted IL-2 upon coculture with HLA-
Cw6 but not –Cw7 transfectants, BW??KIR2DL4 transfectants
did not signal upon coculture with the HLA-G transfectant or
any of the other class I MHC protein-expressing targets (Fig. 5).
A direct examination of HLA-G–KIR2DL4 interaction was
performed with surface plasmon resonance. KIR2DL4-Ig and
KIR2DL1-Ig were bound to CM5 sensor chips coupled with goat
anti-human Fc. Whereas HLA-Cw6 bound the KIR2DL1-Ig
fusion protein with the expected affinity, neither HLA-G mono-
mer nor HLA-G dimer bound KIR2DL4-Ig fusion proteins at
the highest concentrations used (5 mg?ml; Fig. 6).
HLA-G is an unusual class I MHC molecule whose function is
unknown. The data presented here indicate an additional unique
aspect of HLA-G: its ability to form disulfide-linked cell surface-
residues superimposed. Shown below is a portion of the ?1 domain HLA-G sequence containing Cys-42 which was chosen for mutagenesis to a serine. Dots (?)
the conformation-specific mAb W6?32 and the HLA-G-specific mAb, MEM-G?11. After FACS-sorting, HLA-G?C42S and the wild-type HLA-G transfectants
expressed similar levels of protein. (C) Mutagenesis of Cys-42 to Ser-42 completely abrogates HLA-G dimerization. HLA-G?C42S transfectants were cell
surface-biotinylated and lysed in the presence or absence of iodoacetamide. Class I MHC molecules were immunoprecipitated and run under both nonreducing
and reducing conditions. Even in the absence of iodoacetamide, HLA-G dimers could not be detected.
Identification and mutagenesis of extracellular cysteines in HLA-G. (A) A ribbon diagram of the crystal structure of HLA-A2 with Cys-42 and Cys-147
Boyson et al.
December 10, 2002 ?
vol. 99 ?
no. 25 ?
expressed dimers. Numerous other class I MHC molecules
possess unpaired cysteines, but the majority of these are located
in the transmembrane and cytoplasmic domains. However,
consistent with the reducing environment inside the cell, these
cysteines do not form disulfide linkages in vivo (21). Only a few
HLA molecules possess unpaired free cysteines in their extra-
cellular domains, (i.e., exposed to the oxidizing extracellular
environment), and even fewer possess unpaired free cysteines in
regions capable of intermolecular contact. An example of such
an unpaired free cysteine is Cys-67 that exists in some HLA-B27,
HLA-B15, and HLA-B39 allotypes. The question of whether
Cys-67 can mediate formation of disulfide-linked HLA-B27
heavy chain homodimers is controversial (22, 23).
HLA-G dimers could be formed either through oxidation at the
cell surface or through an intracellular enzymatic redox reaction.
The bulk of the evidence, however, points to the former hypothesis
only a relatively small proportion of the total cellular pool of
HLA-G heavy chain, yet they comprised a large proportion of the
cell-surface HLA-G. This finding suggests that most of the dimer is
on the cell surface. Second, soluble recombinant HLA-G was able
to spontaneously dimerize. The presence of monomeric heavy
contains disulfide-linked dimers suggests that HLA-G dimer for-
mation is a two-step process: first, a stable, noncovalent dimer is
formed, followed afterward by Cys-42-mediated disulfide bond
17), may promote dimer formation on the cell surface. Therefore,
a likely hypothesis is that HLA-G arrives at the cell surface as a
monomer, where it subsequently dimerizes. HLA-G dimerization
may depend on a high cell-surface density of HLA-G because
dimers were detected on high-expressing HLA-G transfectants but
not on the low-expressing JEG-3 choriocarcinoma cell line. Thus,
conditions favoring high local concentrations of class I MHC
protein such as lipid raft formation, or clustering with a ligand at a
cell–cell synapse, may promote HLA-G dimer formation.
The existence of class I MHC protein dimers is especially
interesting in light of the fact that KIR inhibitory receptors also
dimerize by means of divalent metal cations such as cobalt and zinc
(24, 25). KIR dimerization has been demonstrated to increase the
affinity of KIRs for their MHC ligands (25, 26), which helps to
explain the initial observation that KIR-mediated inhibition of NK
cells is zinc-dependent (27, 28). The data presented here suggest
that the converse may also be true; HLA-G dimerization could
influence its interactions with putative receptors such as KIR2DL4
or ILT-2 and -4. However, an interaction between KIR2DL4 and
HLA-G (monomer or dimer) could not be detected by using either
a cell-based system or surface plasmon resonance. The failure to
detect a HLA-G–KIR2DL4 interaction suggests that a specific
interaction between these two proteins does not occur. Alterna-
human) protein critical for the interaction, or it may reflect an
inherent low-affinity interaction as is seen for other KIR-activating
receptors with their MHC ligands (29). Thus, the nature of the
The expression of HLA-G in a dimerized form on the cell
surface may have implications for its interaction with putative
receptors. Dimer formation may affect the specificity of HLA-G
for its receptors, either positively (in which the dimerized
HLA-G is the functional receptor ligand) or negatively (in which
the dimerized HLA-G cannot function as a receptor ligand).
Alternatively, dimer formation may have more subtle effects,
such as modulating the affinity (again, either positively or
possibilities may yield insights into the function of HLA-G.
stably transfected with KIR-? constructs were cocultured with various class I
MHC transfectants. After 72 h, supernatants were collected and IL-2 secre-
tion was measured by ELISA. As positive controls, some wells were coated
with anti-KIR2DL1-specific mAb EB6 and anti-KIR2DL4?KIR2DL5 polyclonal
between HLA-Cw6 and KIR2DL1-Ig (A), HLA-G monomer and KIR2DL4- and KIR2DL1-Ig (B), and HLA-G dimer and KIR2DL4- and KIR2DL1-Ig (C). The varying
concentrations of HLA-Cw6 that flowed over the surface are indicated. Neither HLA-G monomer nor HLA-G dimer bound to the KIR-Ig fusion proteins when run
at a concentration of 5 mg?ml (?100 ?M).
www.pnas.org?cgi?doi?10.1073?pnas.212643199 Boyson et al.
We thank Dr. Hugh Reyburn for providing the KIR2DL1-? construct, Download full-text
and we thank Drs. Sumati Rajagopalan and Eric Long for providing their
KIR2DL4-Ig fusion construct. This work was supported by a National
Research Service Award (to J.E.B.), by National Institutes of Health
Grants CA-47554 and AI-50207-02 (to J.L.S.), and by Center of
Molecular and Cellular Immunology Grant LN00A026 (to P.A.).
1. Kovats, S., Main, E. K., Librach, C., Stubblebine, M., Fisher, S. J. & DeMars,
R. (1990) Science 248, 220–223.
2. McMaster, M. T., Librach, C. L., Zhou, Y., Lim, K. H., Janatpour, M. J.,
DeMars, R., Kovats, S., Damsky, C. & Fisher, S. J. (1995) J. Immunol. 154,
3. Alizadeh, M., Legras, C., Semana, G., Le Bouteiller, P., Genetet, B. & Fauchet,
R. (1993) Hum. Immunol. 38, 206–212.
4. Kirszenbaum, M., Djoulah, S., Hors, J., Le Gall, I., de Oliveira, E. B., Prost,
S., Dausset, J. & Carosella, E. D. (1997) Hum. Immunol. 53, 140–147.
5. Geraghty, D. E., Koller, B. H. & Orr, H. T. (1987) Proc. Natl. Acad. Sci. USA
6. Lefebvre, S., Antoine, M., Uzan, S., McMaster, M., Dausset, J., Carosella, E. D.
& Paul, P. (2002) J. Pathol. 196, 266–274.
7. Davies, B., Hiby, S., Gardner, L., Loke, Y. W. & King, A. (2001) Am. J. Reprod.
Immunol. 45, 103–107.
McMichael, A. J., Loke, Y. W. & Braud, V. M. (2000) Eur. J. Immunol. 30,
9. Cantoni, C., Verdiani, S., Falco, M., Pessino, A., Cilli, M., Conte, R., Pende,
D., Ponte, M., Mikaelsson, M. S., Moretta, L. & Biassoni, R. (1998) Eur.
J. Immunol. 28, 1980–1990.
10. Rajagopalan, S. & Long, E. O. (1999) J. Exp. Med. 189, 1093–1100.
11. Rajagopalan, S., Fu, J. & Long, E. O. (2001) J. Immunol. 167, 1877–1881.
12. Faure, M. & Long, E. O. (2002) J. Immunol. 168, 6208–6214.
13. Navarro, F., Llano, M., Bellon, T., Colonna, M., Geraghty, D. E. & Lopez-
Botet, M. (1999) Eur. J. Immunol. 29, 277–283.
14. Allan, D. S., Colonna, M., Lanier, L. L., Churakova, T. D., Abrams, J. S., Ellis,
S. A., McMichael, A. J. & Braud, V. M. (1999) J. Exp. Med. 189, 1149–1156.
15. Gobin, S. J., Biesta, P., De Steenwinkel, J. E., Datema, G. & Van Den Elsen,
P. J. (2002) J. Biol. Chem. 277, 39525–39531.
16. Davis, D. M., Reyburn, H. T., Pazmany, L., Chiu, I., Mandelboim, O. &
Strominger, J. L. (1997) Eur. J. Immunol. 27, 2714–2719.
17. Park, B., Lee, S., Kim, E., Chang, S., Jin, M. & Ahn, K. (2001) Immunity 15,
18. Colonna, M., Borsellino, G., Falco, M., Ferrara, G. B. & Strominger, J. L.
(1993) Proc. Natl. Acad. Sci. USA 90, 12000–12004.
19. Mandelboim, O., Pazmany, L., Davis, D. M., Vales-Gomez, M., Reyburn, H. T.,
Rybalov, B. & Strominger, J. L. (1997) Proc. Natl. Acad. Sci. USA 94,
20. Vales-Gomez, M., Reyburn, H. T., Mandelboim, M. & Strominger, J. L. (1998)
Immunity 9, 337–344.
21. Springer, T. A., Robb, R. J., Terhorst, C. & Strominger, J. L. (1977) J. Biol.
Chem. 252, 4694–4700.
22. Allen, R. L., O’Callaghan, C. A., McMichael, A. J. & Bowness, P. (1999)
J. Immunol. 162, 5045–5048.
23. Malik, P., Klimovitsky, P., Deng, L., Boyson, J. E. & Strominger, J. L. (2002)
J. Immunol. 169, 4379–4387.
24. Fan, Q. R., Long, E. O. & Wiley, D. C. (2000) J. Biol. Chem. 275, 23700–23706.
25. Vales-Gomez, M., Erskine, R. A., Deacon, M. P., Strominger, J. L. & Reyburn,
H. T. (2001) Proc. Natl. Acad. Sci. USA 98, 1734–1739.
26. Fan, Q. R., Long, E. O. & Wiley, D. C. (2000) Eur. J. Immunol. 30, 2692–2697.
28. Rajagopalan, S. & Long, E. O. (1998) J. Immunol. 161, 1299–1305.
29. Vales-Gomez, M., Reyburn, H. T., Erskine, R. A. & Strominger, J. L. (1998)
Proc. Natl. Acad. Sci. USA 95, 14326–14331.
Boyson et al.
December 10, 2002 ?
vol. 99 ?
no. 25 ?