(Received for publication, November 14, 1996, and in revised form, February 17, 1997)
From the Ludwig Institute for Cancer Research, S-751 24 Uppsala, Sweden
Platelet-derived growth factor (PDGF) is a
dimeric growth factor that activates its tyrosine kinase receptor by
inducing receptor dimerization. In this study, we investigated if
receptor-receptor interactions, in addition to ligand-receptor
interactions, contribute to the ligand-induced dimerization of the PDGF
receptors. Analysis of two deletion mutants of the PDGF -receptor
indicated a role for Ig-like domain 4 in ligand-receptor or
receptor-receptor interactions. When the fourth Ig-like domain of the
PDGF
-receptor instead was replaced with the corresponding sequence
of the stem cell factor receptor, the binding of PDGF-AA and -BB was
not affected, nor was the ability to form homodimeric receptor
complexes. This indicates that Ig-like domain 4 does not participate in
ligand-receptor interactions. However, the chimeras did not form
heterodimers with wild-type PDGF
- or
-receptors. Together, these
findings suggest that Ig-like domain 4 mediates specific
receptor-receptor interactions. This notion was also supported by the
finding that a soluble form of Ig-like domain 4 of the PDGF
-receptor acted as a PDGF
-receptor antagonist. We conclude that
specific receptor-receptor interactions contribute to PDGF receptor
dimerization in vivo and that complementary epitopes in
Ig-like domain 4 mediate these interactions. Our experiments also
identify Ig-like domain 4 as a target for PDGF antagonists.
Platelet-derived growth factor (PDGF)1
constitutes a family of disulfide-bonded dimeric isoforms of A- and
B-chains, which exert their effects on cells by binding to two
structurally similar tyrosine kinase receptors, denoted - and
-receptors (reviewed in Refs. 1 and 2). Each of the receptors
consists of an extracellular domain composed of five Ig-like domains, a
single transmembrane region, and an intracellular split tyrosine kinase domain (3-5). The receptors are activated by ligand-induced
dimerization (6, 7). Since the A-chain binds only
-receptors,
whereas the B-chain binds both
- and
-receptors with high
affinity, the different isoforms will induce different dimeric receptor complexes.
Structurally, the PDGF B-chain consists of a tight cystine knot motif
from which two loops (loops 1 and 3) point in one direction and one
(loop 2) points in the other direction (8). The two subunits of the
PDGF-BB molecule are arranged in an antiparallel manner so that loops 1 and 3 from one subunit will be close to loop 2 of the other. The
receptor-binding epitopes reside mainly in loops 1 and 3, but loop 2 also contributes to some extent (9-12). It is likely that other PDGF
isoforms are structurally similar. The dimeric PDGF molecule thus
causes dimerization by simultaneously binding two receptor molecules.
The ligand-binding region of the PDGF -receptor has been mapped to
Ig-like domains 1-3 (13, 14).
Several types of receptors are activated following ligand-induced dimerization (15, 16). In addition to the interactions between ligands and receptors, receptor-receptor interactions contribute to the dimerization of the growth hormone receptor (17) and possibly also to the ligand-induced dimerization of the SCF receptor, which is structurally related to the PDGF receptors, since deletion of Ig-like domain 4 of the SCF receptor was found to prevent dimerization, but not ligand binding (18). In this study, we have explored whether, in addition to ligand-receptor interactions, direct receptor-receptor interactions contribute to ligand-induced PDGF receptor dimerization.
Previously described
cDNAs encoding the PDGF -receptor (5) and the SCF receptor (19)
were used to generate the mutants used in this study. Site-directed
mutagenesis was performed with the Altered SitesTM system
(Promega). To generate the deletion mutant
5-R, XhoI sites (CTCGAG) were introduced into the
-receptor cDNA at
positions corresponding to those encoding the ends of Ig-like domains 4 (amino acids 418-419) and 5 (amino acids 521-522), and the
XhoI fragment was subsequently excised. In a similar manner,
4,5-R was generated by introducing XhoI sites at
positions encoding the ends of Ig-like domains 3 (amino acids 315-316)
and 5, followed by excision of the XhoI fragment. In both
deletion mutants, an NcoI site was also introduced at the
position of the stop codon, enabling ligation to a vector encoding
three tandem HA epitopes followed by a stop codon (a kind gift of Dr.
Y. Xiong). To generate the chimeric receptors, a cDNA fragment
encoding SCF receptor Ig-like domain 4 (amino acids 315-416) was
obtained by polymerase chain reaction using primers introducing
XhoI sites at the ends of the polymerase chain reaction
product. This fragment was ligated to a PDGF
-receptor cDNA in
which an XhoI fragment encoding Ig-like domain 4 had been
cut out. To make the truncated forms, a BglII site (AGATCT)
was introduced into the sequence encoding the juxtamembrane portion
(amino acids 560-561) of the
-receptor and used to ligate to the
vector encoding three tandem HA epitopes followed by a stop codon. All
constructs were inserted into the pSV7d vector (20) and used for
transient expression in COS cells.
Plasmids were transfected into COS cells by the
calcium phosphate method and metabolically labeled with
[35S]methionine and [35S]cysteine for
3 h. After extraction with lysis buffer (20 mM Tris,
pH 7.5, 150 mM NaCl, 0.5% Triton X-100, 0.5% deoxycholic acid, 10 mM EDTA, 1 mM phenylmethylsulfonyl
fluoride, and 1% Trasylol) and centrifugation at 10,000 × g for 15 min, the receptors were immunoprecipitated with
rabbit antisera against the C-terminal part (amino acids 1066-1084) of
the human PDGF -receptor (21) or the HA epitope (a kind gift of Dr.
Lars Rönnstrand). After SDS-gel electrophoresis using 7%
polyacrylamide gels, the labeled proteins were analyzed using a
phosphoimager (FUJIX BAS 2000, Fuji).
PDGF-AA and -BB were 125I-labeled with the chloramine-T (22) and Bolton and Hunter (23) methods, respectively, to specific activities of ~50,000 cpm/ng. Binding experiments were performed as described (24) using transfected COS cells in 12-well dishes.
Analysis of Receptor Homo- and HeterodimerizationCross-linking and immunoprecipitation after
binding of 125I-PDGF-AA in the absence or presence of
excess unlabeled PDGF-AA were performed using transfected COS cells in
60-mm dishes. Cells were washed twice with PBS supplemented with 1 mg/ml bovine serum albumin and incubated for 90 min on ice with 10 ng/ml 125I-PDGF-AA in the absence or presence of 1 µg/ml
unlabeled PDGF-AA. After three washes with PBS, ligand-receptor
complexes were cross-linked by incubation in 1 mM
bis(sulfosuccinimidyl) suberate for 30 min at room temperature.
After incubation in 70 mM methylammonium chloride for 10 min, cell lysates were prepared as described above. Immunoprecipitations were performed with rabbit antisera against the
PDGF -receptor or the HA epitope. After SDS-gel electrophoresis, the
labeled proteins were detected with a phosphoimager.
Cross-linking after PDGF-BB binding was performed as described above
using unlabeled PDGF-BB (200 ng/ml in PBS/bovine serum albumin). For
immunoprecipitations, PDGF -receptor and HA epitope antisera were
used. After SDS-gel electrophoresis using 4-12% gradient gels,
electrophoretic transfer was carried out onto a nitrocellulose membrane
(Hybond-C Extra, Amersham Corp.). For immunoblotting, the filter was
probed with 10 µg/ml monoclonal anti-HA epitope antibody (12CA5,
Boehringer Mannheim) or a rabbit PDGF
-receptor antiserum (PR-4, a
kind gift from Dr. Sara Courtneidge); precipitated components were
detected using horseradish peroxidase-linked secondary antibodies and
enhanced chemiluminescence (ECL, Amersham Corp.).
A polymerase
chain reaction-generated BamHI-EcoRI fragment
encoding Ig-like domain 4 (amino acids 317-417) of the PDGF
-receptor was cloned into the vector pGEX-3X (Pharmacia Biotech
Inc.). Expression of GST-
RIg4 protein was induced by growing
transformed JM109 bacteria in the presence of 0.1 mM
isopropyl-
-D-thiogalactopyranoside at 30 °C for
3 h. Bacteria were lysed in PBS supplemented with 1% Triton
X-100, 1% Trasylol, and 1 mM phenylmethylsulfonyl fluoride and subjected to three pulses of sonication. After clearing of the
lysate by centrifugation at 20,000 × g for 20 min,
GST-
RIg4 was affinity-purified with glutathione-Sepharose
(Pharmacia) and eluted in 2% SDS and 0.1 M Tris, pH 7.4. Before using GST-
RIg4 in experiments on cultured cells, SDS was
removed by four consecutive dilution and concentration cycles in 0.1%
Chaps and 0.1 M Tris, pH 7.4, using a Centriplus spin
column (Amersham Corp.). The final concentration of GST-
RIg4 in the
stock solution was determined by staining with Coomassie Brilliant Blue
after SDS-gel electrophoresis, using known amounts of marker proteins
as standards.
For analysis of the effect of GST-RIg4 on
PDGF
-receptor tyrosine phosphorylation, serum-starved PAE cells
stably transfected with the PDGF
-receptor (21) were used.
Unstimulated cells were incubated in PBS supplemented with 1 mg/ml
bovine serum albumin on ice for 60 min with or without 10 µg/ml GST
control protein or GST-
RIg4. Identical incubations were also
performed in the presence of 3 ng/ml PDGF-AA and on cells that had been
pretreated for 10 min at 37 °C with a 50 mM
concentration of the protein-tyrosine phosphatase inhibitor
pervanadate. After incubation on ice, cells were lysed in lysis buffer,
and PDGF
-receptors were immunoprecipitated from cleared lysates
with rabbit antisera against the human PDGF
-receptor. After SDS-gel
electrophoresis, precipitated receptors were transferred to
nitrocellulose filters, and the filters were probed with
anti-phosphotyrosine antibody PY20 (Transduction Laboratories). After
incubation with horseradish peroxidase-conjugated anti-mouse antibodies, tyrosine-phosphorylated receptors were detected by enhanced
chemiluminescence. The filters were subsequently stripped and reprobed
with the PDGF
-receptor rabbit antisera, and the amount of receptors
was determined by incubation with horseradish peroxidase-conjugated
anti-rabbit antibodies followed by ECL. Control experiments performed
on serum-starved PAE cells stably transfected with the SCF receptor
(25) were performed in a similar fashion using a rabbit antiserum
against the SCF receptor (25) for analysis by immunoprecipitation and
immunoblotting.
The region of the PDGF -receptor
primarily involved in direct PDGF binding has been mapped to Ig-like
domains 1-3 (13, 14). In an initial attempt to determine if regions
outside the ligand-binding domain are of any importance for high
affinity binding of ligand, two deletion mutants were generated in
which either Ig-like domain 5 (
5-R) or Ig-like domain 4 and 5 (
4,5-R) were deleted (Fig. 1A).
Analysis by immunoprecipitations of the two deletion mutants after
expression in metabolically labeled COS cells revealed precursor and
mature forms of the expected sizes (Fig. 1B).
To determine the functional effects of the deletions, COS cells
expressing the wild-type PDGF -receptor or the deletion mutants were
subjected to a 125I-PDGF-AA binding assay (Fig.
1C). Whereas deletion of Ig-like domain 5 had only minor
effects on PDGF-AA binding, deletion of both Ig-like domains 4 and 5 decreased binding dramatically (Fig. 1C). When the results
from the binding data were subjected to Scatchard analysis (26), a
Kd value of 0.5 nM was obtained for both
the wild-type PDGF
-receptor and
5-R. The two deletion mutants
were also compared with the wild-type PDGF
-receptor using
cross-linking of 125I-PDGF-AA to the receptor by
bis(sulfosuccinimidyl) suberate, followed by immunoprecipitations with
PDGF
-receptor antiserum. As shown in Fig. 1D, deletion
of Ig-like domain 5 affected neither the total amount of recovered
ligand-receptor complex nor the ratio of monomeric and dimeric receptor
forms, as compared with the wild-type
-receptor. In contrast,
deletion of both Ig-like domains 4 and 5 reduced dramatically the
formation of both monomeric and dimeric receptor complexes.
Together, these experiments suggest that Ig-like domain 4 is required for high affinity binding of PDGF-AA to its receptor, either by mediating ligand-receptor interactions or by mediating receptor-receptor interactions required for high affinity ligand binding. The following experiments were designed to distinguish between these two possibilities.
The Fourth Ig-like Domain of the PDGFA cDNA encoding a chimeric PDGF -receptor
in which the fourth Ig-like domain of the receptor was replaced with
the corresponding sequence of the SCF receptor (
/kit-R) was
generated (Fig. 1A). In addition, cDNAs encoding
carboxyl-terminally truncated forms of the chimera (T
/kit-R) as well
as of the wild-type PDGF
-receptor (T
-R) were generated (Fig.
1A). Both truncated forms contain three tandem HA epitopes
at their carboxyl termini for immunodetection. When expressed
transiently in COS cells, all four types of receptors were detected as
precursor and mature forms of the expected sizes after
immunoprecipitations from lysates of metabolically labeled cells (Fig.
2A).
The four receptors were expressed in COS cells and analyzed with regard
to binding of PDGF-AA. As shown in Fig. 2B, no major differences were observed between the full-length and truncated receptors on one hand and between the wild-type and chimeric receptors on the other (Fig. 2B). Scatchard analysis of the results
from the binding experiments yielded Kd values
between 0.3 and 0.5 nM. Furthermore, when analyzed in
cross-linking experiments, the full-length chimeric receptor formed
cross-linked dimers in a manner indistinguishable from that of the
wild-type PDGF -receptor (Fig. 3, A and
B, lanes 1-4). The finding that Ig-like domain 4 of the PDGF
-receptor can be substituted with the corresponding domain of the SCF receptor, which does not bind PDGF, suggests that
Ig-like domain 4 is not involved in ligand-receptor interactions.
To explore if Ig-like domain 4 is
involved in direct and specific receptor-receptor interactions, we
designed cross-linking experiments in which the ability of the receptor
chimera to form homodimers could be compared with the ability to form
heterodimers with wild-type PDGF -receptors. The full-length
-receptor, either alone or together with the truncated wild-type or
truncated chimeric receptor, was transiently expressed in COS cells.
After binding of 125I-PDGF-AA, with or without an excess of
unlabeled PDGF-AA, ligand-receptor complexes were cross-linked with
bis(sulfosuccinimidyl) suberate, immunoprecipitated, and subjected to
analyses by SDS-gel electrophoresis, followed by exposure using a
phosphoimager (Fig. 3A). In the immunoprecipitations with
PDGF
-receptor antiserum, dimers of the full-length wild-type
-receptor were recovered (lanes 3, 5, and
7). In addition, cross-linked dimers of full-length and
truncated wild-type receptors could be demonstrated (lane
5). However, no heterodimers composed of the full-length wild-type
PDGF
-receptor and truncated chimeric receptors were detected
(lane 7). In the HA tag precipitations, homodimeric forms of
both truncated versions were easily identified as broad tailing
components at positions that partially overlap with the positions of
the heterodimeric complexes (lanes 9-14).
In an analogous experiment in which the full-length chimeric receptor was expressed alone or together with either of the two truncated forms, similar findings were obtained; all possible homodimeric receptor forms were identified (Fig. 3B, lanes 3, 5, 7, 11, and 13). Also, dimers of full-length and truncated chimeric receptors were detected (lane 7). In contrast, no heterodimers between full-length chimeric and truncated wild-type receptors were observed (lane 5).
Thus, in the immunoprecipitations with PDGF -receptor antiserum,
dimers composed of one truncated and one full-length form of the PDGF
-receptor could be demonstrated as well as dimers composed of one
truncated and one full-length SCF receptor/PDGF
-receptor chimera.
However, no heterodimers between wild-type and chimeric receptors were
seen. These experiments thus suggest that PDGF
-receptor
dimerization is not exclusively a consequence of ligand-receptor
interactions, but also requires specific receptor-receptor interactions. Since the chimera's ability to form homodimers was unaffected, we also conclude that the receptor-receptor interactions occur between complementary epitopes in Ig-like domain 4.
PDGF-AA binds exclusively to PDGF -receptors. In
contrast, PDGF-BB binds with equal affinity to
- and
-receptors
and will thus, on cells expressing both receptor types, induce the
formation of heterodimeric
- and
-receptor complexes in addition
to homodimeric complexes. We therefore investigated whether Ig-like
domain 4 of the
-receptor also is involved in interactions with the
-receptor. The truncated wild-type and chimeric
-receptors
expressed in COS cells showed similar affinities for binding of PDGF-BB
(Fig. 4A). A Kd value of
0.7 nM was obtained when binding data were subjected to
Scatchard analysis. The PDGF
-receptor was coexpressed with either
the truncated wild-type or the truncated chimeric receptor; after
ligand binding, cells were incubated with or without cross-linker, and
lysates were subjected to immunoprecipitations with a PDGF
-receptor
antiserum and analysis by immunoblotting with a monoclonal anti-HA tag
antibody. A dimeric complex between the PDGF
-receptor and the
truncated
-receptor complex could be demonstrated, but no
heterodimeric complexes between the truncated chimera and the
-receptor (Fig. 4B, upper panel). Control
experiments showed that equal amounts of the different receptor forms
were expressed (Fig. 4B, middle and lower
panels). The finding that substitution of Ig-like domain 4 altered
the ability of the chimera to dimerize with wild-type PDGF
-receptors, without affecting the ability to bind PDGF-BB, extends
the role for Ig-like domain 4 in homodimerization of PDGF
-receptors
to include also heterodimerization with PDGF
-receptors.
A Soluble Form of the Fourth Ig-like Domain of the PDGF
We
next used a receptor autophosphorylation assay to investigate whether a
soluble Ig-like domain 4 could interfere with receptor activation. PDGF
-receptor Ig-like domain 4 was expressed as a GST fusion protein
(GST-
RIg4) in bacteria and purified through glutathione-Sepharose
affinity purification. To investigate its ability to interfere with
PDGF
-receptor activation, PAE cells stably expressing the PDGF
-receptor were used (Fig. 5). The potential
inhibitory effect of GST-
RIg4 on PDGF
-receptor tyrosine phosphorylation was analyzed under three different conditions: using
unstimulated cells, cells stimulated with 3 ng/ml PDGF-AA, and cells
pretreated with pervanadate, a potent inhibitor of protein-tyrosine phosphatases. After cell lysis, the extent of tyrosine phosphorylation of PDGF
-receptors was determined by immunoprecipitations of the
PDGF
-receptor followed by immunoblotting with anti-phosphotyrosine antibodies. As shown in Fig. 5A (upper panel),
the presence of 10 µg/ml GST-
RIg4 led to a reduction in background
as well as ligand-stimulated tyrosine phosphorylation of the PDGF
-receptor. No such effect was seen when cells were incubated with
GST (Fig. 5A, upper panel). The total amount of
receptors was the same under the different experimental conditions
(Fig. 5A, lower panel). To investigate the
specificity of the antagonistic action of GST-
RIg4, PAE cells stably
expressing the SCF receptor were subjected to similar experiments. As
shown in Fig. 5B, GST-
RIg4 did not show any effect on
ligand-independent tyrosine phosphorylation of the SCF receptor, nor
was any effect seen on SCF-stimulated tyrosine phosphorylation of the
SCF receptor (data not shown). The finding that GST-
RIg4 functions
as an antagonist to PDGF
-receptor tyrosine phosphorylation supports
the notion that Ig-like domain 4 is involved in direct
receptor-receptor interactions.
In this study, we provide evidence that PDGF receptor dimerization
involves, in addition to the interaction between ligand and receptor, a
direct interaction between the receptors. Whereas the ligand binds to
epitopes in Ig-like domains 1-3 (13, 14), Ig-like domain 4 mediates
direct receptor-receptor interactions (Fig. 6).
PDGF receptor dimerization can thus occur via two different routes:
either the ligand binds first to one receptor and subsequently to
another, or alternatively, the ligand binds to and stabilizes preformed
transient receptor dimers (Fig. 6). It is noteworthy that in PAE cells
stably expressing the PDGF -receptor, which express receptor levels
comparable to those in cells expressing endogenous receptors, the
antagonistic effect of GST-
RIg4 was observed not only on
ligand-stimulated cells, but also on the background phosphorylation of
the receptor in unstimulated cells (Fig. 5, A and
B). These findings suggest that the background receptor
phosphorylation is due to the formation of transient ligand-independent
receptor dimers, rather than to receptor monomers displaying a basal
kinase activity. Thus, an equilibrium between monomeric and dimeric
receptors appears to exist in the absence of ligand. It follows that
the receptor density on the target cell is likely to influence which of
the two routes in Fig. 6 are preferentially used.
The structurally best characterized example of ligand-receptor interaction is the binding of growth hormone (GH) to the GH receptor (17), for which the crystal structure is known. In this case GH, binds with high affinity to one receptor and, through a different epitope, with lower affinity to another receptor, thus forming a 1:2 complex. The binding of the second receptor is stabilized by direct interactions between the receptors. Thus, the the PDGF ligand-receptor complex and the GH ligand-receptor complex are similar in that they both are stabilized by a combination of ligand-receptor and receptor-receptor interactions, but differ in that the two ligand-receptor interactions in the case of PDGF are of similar strength, but differ in the case of GH.
The PDGF receptors are structurally most related to the SCF and colony-stimulating factor-1 receptors (19, 27). These receptors also bind dimeric ligands. In the case of the SCF receptor, an involvement of Ig-like domain 4-mediated receptor-receptor interactions in dimerization was recently postulated based on the finding that deletion of Ig-like domain 4 prevented dimerization and, to a lesser extent, ligand binding (18). Interestingly, mutations in the colony-stimulating factor-1 receptor that lead to its ligand-independent activation are localized to Ig-like domain 4 (27). It is possible that the activating effects of these mutations are mediated by causing increases in the affinity of receptor-receptor interactions. Together with the findings presented in this paper, these observations suggest that direct interactions between receptors may be common in ligand-induced dimeric receptor complexes.
PDGF antagonists are highly warranted given the overactivity of PDGF receptors in certain pathological situations, including tumorigenesis and atherosclerosis (28-31). Previous attempts to design antagonists acting on PDGF receptors have focused mainly on blocking direct ligand-receptor interactions or on inhibiting the tyrosine kinase activity of the receptors (13, 28, 32-34). In addition, dominant-negative forms of both PDGF itself and the receptors have been described (35-37). Our results suggest that blocking receptor-receptor interaction by targeting Ig-like domain 4 may serve as an additional strategy to interfere with signaling through PDGF receptors.
We thank Ingegärd Schiller for
secretarial assistance; Sara Courtneidge and Lars Rönnstrand for
the PDGF -receptor and HA epitope antisera, respectively; Y. Xiong
for the HA epitope vector; and Christer Wernstedt for synthesis of
oligonucleotides.