Contacts between membrane proximal regions of thePDGF receptor ectodomain are required for receptoractivation but not for receptor dimerizationYan Yang, Satoru Yuzawa, and Joseph Schlessinger*

Department of Pharmacology, Yale University School of Medicine, 333 Cedar Street, New Haven, CT 06520

Contributed by Joseph Schlessinger, March 24, 2008 (sent for review March 2, 2008)

The mechanism of PDGF-receptor � (PDGFR�) activation was ex-plored by analyzing the properties of mutant receptors designedbased on the crystal structure of the extracellular region of therelated receptor tyrosine kinase KIT/stem cell factor receptor. Here,we demonstrate that PDGF-induced activation of a PDGFR� mu-tated in Arg-385 or Glu-390 in D4 (the fourth Ig-like domain of theextracellular region) was compromised, resulting in impairment ofa variety of PDGF-induced cellular responses. These experimentsdemonstrate that homotypic D4 interactions probably mediated bysalt bridges between Arg-385 and Glu-390 play an important rolein activation of PDGFR� and all type III receptor tyrosine kinases.We also used a chemical cross-linking agent to covalently cross-linkPDGF-stimulated cells to demonstrate that a Glu390Ala mutant ofPDGFR� undergoes typical PDGF-induced receptor dimerization.However, unlike WT PDGFR that is expressed on the surface ofligand-stimulated cells in an active state, PDGF-induced Glu390Aladimers are inactive. Although the conserved amino acids that arerequired for mediating D4 homotypic interactions are crucial forPDGFR� activation, these interactions are dispensable for PDGFR�dimerization. Moreover, PDGFR� dimerization is necessary but notsufficient for tyrosine kinase activation.

cell proliferation � cell signaling � phosphorylation � surface receptors �tyrosine kinases

The generally accepted mechanism of receptor tyrosine kinase(RTK) activation is that ligand-induced receptor dimeriza-

tion facilitates transautophosphorylation of critical regulatorytyrosine residues in the activation loop of the catalytic core, astep essential for tyrosine kinase activation. This is followed byautophosphorylation of multiple tyrosine residues in the cyto-plasmic domain that serve as binding sites for Src homology 2(SH2) or phosphotyrosine-binding (PTB) domains of a variety ofsignaling proteins, which upon recruitment and/or tyrosinephosphorylation transmit signals to a variety of intracellularcompartments in a regulated manner (1–3).

Although nearly all RTKs are activated by dimerization,different RTK families have evolved to use different molecularstrategies for ligand-induced receptor dimerization and activa-tion. Dimerization and activation of members of the EGFreceptor (EGFR) family is mediated by interactions betweenEGF or TGF� with the extracellular region (ectodomain) ofEGFR, which exposes a buried dimerization interface thatfacilitates receptor-mediated EGFR dimerization and formationof an activated EGF/EGFR 2:2 complex (4). Dimerization of theFGF receptors (FGFRs), however, is mediated by tripartiteinteractions among monomeric FGF molecules, heparan sulfateproteoglycans (HSPG), and FGFR molecules to stabilize theformation of an active ternary FGF/HSPG/FGFR 2:2:2 complex(5). By contrast to EGF or FGF, all ligands of type III RTKs,including PDGFs, stem cell factor (SCF), colony-stimulatingfactor (CSF), and Flt3-ligand (Flt3L), are dimeric moleculescapable of cross-linking their cognate receptors by bivalentbinding to equivalent sites of two neighboring receptor mole-cules. Each PDGF protomer is composed of a central four-

stranded �-sheet with the characteristic cystine knot at one end ofthe molecule. Two PDGF protomers are arranged in an antiparallelmanner and linked to each other by two interchain disulfide bridges(6). By contrast each SCF, CSF, or Flt3L protomer is composed ofa short helical fold, and they are connected to each other bynoncovalent interactions (7–10). Despite their diverse folds, thetwo growth factor subtypes bind to and activate their cognateRTKs in a virtually identical manner, resulting in formation ofactivated ligand/RTK 2:2 complexes (10). All type III RTKs arecomposed of an extracellular ligand-binding region containingfive tandem Ig-like domains followed by a single transmembranehelix and a cytoplasmic tyrosine kinase domain with a largekinase-insert region flanked by regulatory regions that aresubject to autophosphorylation and to phosphorylation by het-erologous protein kinases (11).

The elucidation of the x-ray crystal structure of the entireectodomain of KIT/stem cell factor receptor before and afterSCF stimulation provided valuable insights concerning themechanism of SCF-induced KIT dimerization and activation(12). The structure shows that the first three Ig-like domains ofKIT, designated D1, D2, and D3, are responsible for SCFbinding. The main role of SCF binding is to cross-link two KITmolecules to increase the local concentration of KIT on the cellmembrane. This facilitates a large conformational change in themembrane-proximal regions of KIT, resulting in a homotypicinteraction between D4 or D5 of neighboring KIT molecules.The lateral interactions between D4 of two neighboring KITmolecules occur via direct contacts through two pairs of saltbridges from EF loops of each D4 protomer. The membraneproximal D5 domain provides additional indirect interactionsbetween neighboring KIT molecules to further stabilize andposition the membrane proximal part of the ectodomain at adistance and orientation that enables the activation of cytoplas-mic tyrosine kinase.

On the basis of the structure-based sequence alignment of typeIII RTKs ectodomains and a homology model of the PDGFreceptor (PDGFR) D4 structure, we identified amino acids inthe PDGFR� D4 domain that may form salt bridges similar tothose shown to mediate homotypic D4 interactions essential forSCF-induced KIT activation. In this report, we demonstrate thatPDGF-induced activation of PDGFR� is compromised whenArg-385 and Glu-390 in D4 were mutated to alanine residues.Furthermore, a variety of cellular responses that depend onPDGFR� activation either are reduced or their kinetics stronglyattenuated. We also apply a chemical cross-linking agent tocovalently cross-link intact unstimulated or PDGF-stimulated

Author contributions: Y.Y., S.Y., and J.S. designed research; Y.Y. and S.Y. performedresearch; Y.Y., S.Y., and J.S. analyzed data; and Y.Y. and J.S. wrote the paper.

The authors declare no conflict of interest.

*To whom correspondence should be addressed. E-mail: joseph.schlessinger@yale.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0802896105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0802896105 PNAS � June 3, 2008 � vol. 105 � no. 22 � 7681–7686

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

1

cells to demonstrate that an E390A point mutation in D4 doesnot interfere with PDGF-induced receptor dimerization. How-ever, unlike the covalently cross-linked WT PDGFR� dimersthat are displayed on the cell surface in an activated state, thecovalently cross-linked dimers of the E390A mutant are inactive.These experiments demonstrate that the conserved amino acidsthat take part in formation of D4 homotypic interactions play acritical role in PDGFR� activation. However, the D4 homotypicinteractions are dispensable for PDGFR� dimerization. Al-though bivalent PDGF binding is the driving force for PDGFR�dimerization, dimerization itself is necessary but not sufficientfor tyrosine kinase activation.

ResultsSimilar to the D4 domain of KIT, D4 of PDGFR� and PDGFR�lack a characteristic disulfide bond that bridges cysteine residueslocated in B5 and F5 in Ig-like domains. The amino acidsequence alignment presented in supporting information (SI)Fig. S1 shows that 13 of 20 fingerprint residues of the I-set IgSFfold are conserved in D4 of PDGFRs, and that the number andlength of strands corresponding to the fingerprint residues arehighly conserved in D4 of KIT, PDGFR�, PDGFR�, andCSF1R (12).

D4 of KIT is composed of two � sheets, each containing fourstrands, with the arrangement ABED/A�GFC, and the homo-typic D4 contacts are mediated by the EF loop of D4 projectingfrom two neighboring KIT molecules (12). The KIT structureshows that Arg-381 and Glu-386 in the EF loop form salt bridgesand van der Waals contacts across a twofold axis of KIT dimer.In addition, the side chains of Arg-381 of each protomer formhydrogen bonds with the main chain carbonyl of the correspond-ing residue of neighboring KIT molecules (12). Structure-basedsequence alignment has shown that the size of the EF loop andthe critical amino acids comprising the D4–D4 interface areconserved in KIT, PDGFR�, PDGFR�, and CSF1R. InPDGFR�, Glu-386 is replaced by an aspartic acid, a residue thatmay also function as a salt-bridge partner. In addition, a pair ofbasic and acidic (Glu/Asp) residues is strictly conserved inPDGFR� and PDGFR� of different species ranging fromTakifugu rubripes to Homo sapiens (Fig. S1), providing furthersupport for the functional importance of this region.

PDGF-Induced PDGFR Activation Is Compromised by Mutations in D4.The amino acid sequence alignment presented in Fig. S1 dem-onstrates that Arg-385 and Glu-390 in the EF loop of PDGFR�may mediate homotypic D4 interactions similar to the saltbridges formed between Arg-381 and Glu-386 of KIT that areresponsible for mediating homotypic D4 interactions between

neighboring KIT receptors (Fig. 1). To investigate whether asimilar mechanism is used by PDGFR�, Arg-385, and Glu-390,each alone (R385A, E390A) or in combination (RE/AA), weresubstituted by alanine residues. An additional conserved Lys-387residue in the loop region was also substituted by an alanine(RKE/AAA) residue to examine its potential role in control ofPDGF-induced PDGFR� activation. WT and mutant PDGFR�s were stably expressed in fibroblasts derived from mouseembryos (MEFs) deficient in both PDGFR� and PDGFR�(13–15). MEFs expressing WT or mutant PDGFR� s matchedfor expression level were used in the experiments describedbelow. The experiment presented in Fig. 2A shows that PDGF-induced tyrosine autophosphorylation of PDGFR� is stronglycompromised in cells expressing the E390A, R385A, RE/AA,and RKE/AAA mutants of PDGFR�; both the magnitude (Fig.2A) and kinetics (Fig. S2) of tyrosine autophosphorylation werereduced and attenuated, respectively. These experiments dem-onstrate that Arg-385 and Glu-390 in the EF loop of D4 play animportant role in PDGF-induced stimulation of PDGFR�,which suggests that a similar pair of salt bridge to those identifiedin KIT structure may exist in activated PDGFRs. Direct inter-action between D4 of neighboring receptor within the ligand–receptor complex may represent a common mechanism used forligand-induced activation of type III RTKs. We have consistentlyand reproducibly observed that PDGF-induced receptor auto-phosphorylation is more strongly compromised in cells express-ing the E390A compared with cells expressing the R385A,RE/AA, or the RKE/AAA mutants. Although the precisemechanism responsible for the difference between these mutantsis not clear it is possible that the positive local surface charge atthe D4 interface may cause electrostatic repulsion to maintainD4 of neighboring receptors apart before ligand stimulation.

Fig. 1. Homology modeling of membrane proximal region of PDGFRs. Themembrane proximal region of PDGFR� ectodomain is shown as ribbons withtransparent molecular surface (D4 colored in gold and D5 in magenta) (Left).A closer view (Right) of the D4–D4 interface of two neighboring PDGFR�

molecules demonstrates that interactions between D4 are mediated by resi-dues Arg-385 and Glu-390 projected from two adjacent EF loop. Key aminoacids are labeled and shown as a stick model. Fig. 2. PDGF-induced PDGFR activation is compromised by mutations in D4.

(A) PDGFR�/��/� MEFs expressing WT PDGFR� and various PDGFR� D4 mu-tants (R385A, E390A, RE/AA, and RKE/AAA) were serum-starved overnight andstimulated with indicated PDGF concentration for 5 min at 37°C. Cell lysateswere immunoprecipitated with anti-PDGFR antibodies, followed by immuno-blotting with antiphosphotyrosine antibodies 4G10. Membranes werestripped off and reblotted with anti-flag tag antibodies. (B) PDGFR�/��/�MEFs expressing WT (■ ), R385A (Œ), E390A (�), and RE/AA (}) PDGFR� wereincubated with 5 ng/ml 125I-PDGF at 4°C for 90 min in the presence ofincreasing concentration of native PDGF. Cell-associated 125I-PDGF were col-lected with 0.5 M NaOH solution and quantitated with a scintillation counter.The IC50 values were determined by curve fitting with Prism4 (GraphPad). (C)MEFs expressing WT, R385A, E390A, RE/AA, and RKE/AAA PDGFR were serum-starved overnight and lysed. Cell lysates were immunoprecipitated with anti-PDGFR antibodies, and immunopellets were subjected to in vitro autophos-phorylation assay in the absence (�) or presence (�) of 1 mM ATP and 10 mMMg2� for 10 min at room temperature. Pellets were resolved with SDS/PAGEfollowed by immunoblotting with antiphosphotyrosine antibodies and anti-flag antibodies.

7682 � www.pnas.org�cgi�doi�10.1073�pnas.0802896105 Yang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

1

Whereas substitution of Arg-385 by an alanine residue willprevent salt bridge formation this change may also decrease thenet positive charge in the D4-D4 interface resulting in weakerinhibition of PDGFR activation.

To examine the possibility of whether mutation in D4 ofPDGFR may have affected cell membrane expression andligand-binding affinity of mutant PDGFR�s, we next performedquantitative PDGF-binding experiments to cells expressing WTor mutant PDGFR�s. Cells expressing WT, R385A, E390A, orthe RE/AA PDGFR� mutants were incubated with a buffersolution containing 125I-PDGF for 90 min at 4°C in the presenceof increasing concentration of native PDGF. Cell-bound radio-activity was measured by using a scintillation counter. The EC50values of the displacement curves of WT and mutant PDGFR�swere analyzed by curve fitting with Prism4 (Fig. 2B). We havealso compared the amount of WT and mutant PDGFR�sexpressed in the transfected MEFs by immunoblotting of totalcell lysates with antibodies against PDGFR or antitag antibodies(Fig. 2 A and C). Taken together, these experiments demonstratethat similar amount of WT or mutant PDGFR�s are expressedon the cell surface of the transfected cells. Moreover, similar IC50values (PDGF concentration that displaces 50% of 125I-PDGFbinding) were obtained for cells expressing WT (3.7 nM), R385A(6.0 nM), E390A (2.8 nM), or the RE/AA (3.0 nM) mutants. Wealso examined the possibility of whether the intrinsic tyrosinekinase activity of mutant PDGFR�s was adversely affected bycomparing the in vitro tyrosine kinase activities of WT andmutant receptors. In this experiment, cell lysates from serum-starved cells were subjected to immunoprecipitation with anti-PDGFR antibodies, and the immobilized PDGFRs were sub-jected to in vitro kinase assays in the presence of 1 mM ATP and10 mM magnesium chloride. After incubation, the samples wereanalyzed by immunoblotting with antiphosphotyrosine antibod-ies. The experiment presented in Fig. 2C demonstrates that theR385A, E390A, or RE/AA mutations do not influence theintrinsic tyrosine kinase activity of PDGFR. Together, theseexperiments demonstrate that the mutations in D4 that affectPDGF-induced stimulation of PDGFR� do not alter the expres-sion of PDFGR� on the cell surface, influence the ligand bindingaffinity of PDFGR�, or alter the intrinsic tyrosine kinaseactivities of mutant PDGFR�.

PDGFR D4 Point Mutants Are Expressed on the Surface of PDGF-Stimulated Cells in the Form of Inactive Dimers. Because receptordimerization has been established as critical mechanism under-lying RTK activation, we investigated whether reduced tyrosineautophosphorylation of mutant PDGFR� in response to PDGFstimulation is caused by deficiency in receptor dimerization. Wehave previously applied chemical cross-linking agents to monitorand follow ligand-induced dimerization of several cell membranereceptors, including WT, and a variety of EGF receptor mutantson the cell surface of living cells (16). In this experiment, cellsexpressing WT PDGFR� or the E390A mutant were serum-starved overnight, followed by PDGF incubation for 90 min at4°C. Several washes were used to remove unbound PDGF, andthe cells were incubated with 0.5 mM disuccinimidyl suberate(DSS) in PBS for 30 min at 25°C. Cell lysates from unstimulatedor PDGF-stimulated cells were subjected to immunoprecipita-tion with anti-PDGFR antibodies followed by SDS/PAGE andimmunoblotting with either antif lag antibodies to monitor thestatus of PDGFR dimerization or with antiphosphotyrosineantibodies to monitor the status of PDGFR activation (Fig. 3).

The experiment depicted in Fig. 3 demonstrates that, in lysatesof unstimulated cells, a band that migrates in SDS gel with anapparent molecular mass of 180 kDa corresponding to PDGFRmonomers was detected in lysates from cells expressing eitherWT PDGFR� or the E390A mutant. Upon PDGF stimulation,an additional band that migrates in SDS gel with an apparent

molecular mass of 360 kDa corresponding to PDGFR dimers wasdetected in cells expressing both WT PDGFR� and the E390Amutant. However, immunoblotting of the samples with antiphos-photyrosine antibodies demonstrates that, whereas the bandcorresponding to dimers of WT PDGFR is strongly tyrosine-phosphorylated, very weak tyrosine phosphorylation of the bandcorresponding to the dimers of E390A mutant is detected (Fig.3). This experiment shows that impaired ligand-induced tyrosineautophosphorylation of the E390A mutant is not caused bydeficiency in ligand-induced receptor dimerization. It also dem-onstrates that the covalently cross-linked WT PDGFR� aredisplayed on the cell surface of PDGF-stimulated cells in theform of active dimers, whereas the E390A mutant is displayed onthe surface of PDGF-stimulated cells in the form of inactivedimers. We conclude that the D4 homotypic interactions inPDGFR are dispensable for receptor dimerization, and thatPDGF-induced receptor dimerization is necessary (17) but notsufficient for tyrosine kinase activation (Fig. 3).

Impaired Stimulation of Cells Signaling in Cells Expressing D4 PDGFRMutants. We next examined the impact of PDGFR D4 mutationson cell signaling in response to PDGF stimulation. Lysates fromunstimulated or PDGF-stimulated cells expressing either WT orPDGFR D4 mutants were subjected to immunoprecipitationwith antiphospholipase C� (anti- PLC�) antibodies followed bySDS/PAGE and immunoblotting with either anti-PLC� or anti-pTyr antibodies. The experiment presented in Fig. S3 shows thattyrosine phosphorylation of PLC� is severely compromised incells expressing the R385A, E390A, RE/AA, or the RKE/AAAPDGFR mutants. Impaired stimulation of additional PDGF-induced cellular responses are observed in cells expressingPDGFR D4 mutants. The experiment presented in Fig. 4A showsthat MAPK response and Akt stimulation were strongly com-promised in cells expressing the R385A, E390A, RE/AA, orRKE/AAA PDGFR mutants, as compared with similar re-sponses induced by PDGF in MEFs expressing WT PDGFRs.Overall, �10-fold higher concentrations of PDGF were requiredfor a similar level of MAPK response and Akt stimulation in cellsexpressing the E390A, RE/AA, or RKE/AAA PDGFR mutants.

One of the hallmarks of PDGF stimulation of cultured fibro-blasts is a typical formation of membrane ruffles and circularactin ring structures on the dorsal surface of PDGF-stimulatedcells. The experiment presented in Fig. 4 B and C shows thatPDGF stimulation of actin ring formation is compromised inMEFs expressing PDGFR D4 mutants. Although �83% ofMEFs expressing WT PDGFR exhibited circular actin ringformation, only 5% of PDGFR D4 mutant cells showed similarcircular actin ring formation after 2-min stimulation with 50

Fig. 3. PDGF-stimulated PDGFR� mutated in D4 are expressed on the cellsurface in the form of inactive dimers. PDGFR�/��/� MEFs expressing WTPDGFR� or E390A mutant were serum-starved overnight, followed by incu-bation with the indicated amount of PDGF at 4°C for 90 min. After removingthe unbound ligand, cells were incubated with 0.5 mM DSS in PBS for 30 min.Cell lysates were immunoprecipitated with anti-PDGFR antibodies, and im-munopellets were analyzed by SDS/PAGE and immunoblotted with anti-flagantibodies (Left) and antiphosphotyrosine antibodies (Right), respectively.

Yang et al. PNAS � June 3, 2008 � vol. 105 � no. 22 � 7683

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

1

ng/ml of PDGF. Furthermore, the transient circular actin ringformation that peaks in MEFs expressing WT PDGFR after 2–5min of PDGF stimulation was weakly detected in cells expressingthe R385A, E390A, or the RE/AA PDGFR mutants.

Reduced Internalization and Degradation of D4 PDGFR Mutants. Wenext examined the effect of PDGFR D4 mutations on PDGFinternalization, PDGFR degradation, and PDGFR ubiquitina-tion. MEFs expressing WT PDGFR or the PDGFR D4 mutantswere treated with 5 ng/ml of 125I-labeled PDGF for 90 min at 4°Cfollowed by brief washes with PBS (pH 7.4) to remove the excessligand in the medium. Prelabeled cells were warmed to 37°C toinitiate the endocytosis of ligand–receptor complex for varioustime intervals up to 4 h. Cell surface-bound, intracellular, anddegraded 125I-PDGF in medium were collected, quantitated byusing a scintillation counter, and presented as percentage of totalcell-associated 125I-PDGF radioactivity after 90-min incubation(t � 0) at 4°C (mean � SD). The experiment presented in Fig.5A shows that the kinetics of internalization of 125I-labeledPDGF bound to MEFs expressing WT PDGFR is much fasterthan the kinetics of internalization of 125I-labeled PDGF boundto cells expressing the E390A, R385A, or the RE/AA (data notshown) PDGFR mutants. After 30 min, �75–80% of 125I PDGFwas removed from the cell surface and accumulated inside thecells expressing WT receptors compared with �50% in cells

expressing mutant receptors. The low molecular mass degrada-tion product of 125I-PDGF became detectable after 30 min. Therelease of degraded 125I-PDGF was much slower in E390Amutant cells than in WT cells (Fig. 5A). Reduced PDGFinternalization and degradation were reflected in reduced deg-radation of PDGFR D4 mutants. Cells expressing WT or theR385A, E390A, or RE/AA PDGFR mutants were first incu-bated 30 min with cycloheximide, to prevent the biosynthesis ofnew PDGFR molecules during the degradation experiment. Theexperiment presented in Fig. 5B shows that the kinetics ofdegradation of R385A, E390A, or the RE/AA PDGFR mutantswas strongly attenuated, whereas half of WT PDGFRs weredegraded within 1.5 h of PDGF stimulation, the half-life forPDGFR D4 mutants was extended to �4 to 6 h. The experimentpresented in Fig. S4 shows that PDGF-induced stimulation ofubiquitination of the E390A PDGFR was also strongly reducedas compared with WT PDGFR under similar conditions. Takentogether, these experiments demonstrate that PDGFR internal-ization and ubiquitin-mediated PDGFR degradation are com-promised by mutations in D4 of PDGFR.

DiscussionThe extracellular domains of all members of type III RTKs,including PDGFR�, PDGFR�, CSF1R, Flt3, and KIT, arecomposed of five Ig-like domains, of which the first three

Fig. 4. PDGF-induced cellular responses are compromised by mutations in PDGFR� D4 mutant. (A) Cells were stimulated with 10 ng/ml PDGF for 5 min, asdescribed above. Total lysates were subjected to SDS/PAGE and analyzed by immunoblotting with antiphospho-MAPK, MAPK, phospho-Akt, and Akt antibodies.(B) Cells seeded on coverslips were serum-starved for 16 h and either left untreated or stimulated with 50 ng/ml PDGF for 2, 5, 10, or 30 min. Coverslips were stainedwith FITC-phalloidin, and the percentage of cells showing dorsal actin rings were quantitated and presented linearly in C. WT (■ ), R385A (�), E390A (Œ), andRE/AA (�).

7684 � www.pnas.org�cgi�doi�10.1073�pnas.0802896105 Yang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

1

function as binding sites for the dimeric ligand molecule, whichupon binding stimulates receptor dimerization and activation.Because the molecular architecture, ligand-binding characteris-tics, and mechanism of receptor dimerization of type III RTKsare highly conserved, the mechanism of SCF-induced KITactivation revealed by the crystal structures of the completeextracellular domain of KIT before and after SCF stimulationmay represent a general mechanism of activation of all type IIIRTKs. Moreover, phylogenic analysis of RTKs containing Ig-like domains in their extracellular domains suggests a commonevolutionary origin for types III and IV RTK, a family includingVEGFR1 (Flt1), VEGFR2 (KDR), and VEGFR3 (Flt4) (18).Moreover, both VEGF and PDGF belong to the same cystine-knot family, homodimeric growth factors, sharing similar topol-ogy, size, and receptor-binding strategy. The salient features ofKIT activation revealed by the x-ray structural analysis of itsextracellular domain may, therefore, also apply for ligand-induced activation of type IV RTKs.

The structural analysis of KIT has shown that a pair of saltbridges formed between Glu-386 and Arg-381 of two neighbor-ing D4 domains are responsible for mediating homotypic D4interactions that are essential for SCF-induced KIT activation.Comparison of the amino acid sequences of type III RTKsdemonstrates that an identical sequence motif exists in the EFloop region of D4 of PDGFR�, PDGFR�, and CSF1R (Fig. 1and Fig. S1), suggesting that a similar salt bridge may also beformed between D4 of type III RTKs. Indeed, substitution ofArg-385 or Asp-390 in D4 of PDGFR� by alanines has com-promised PDGF stimulation of PDGFR� activation resulting inimpairment of a variety of cellular responses that are stimulatedby PDGF in cells expressing WT PDGFR�. The mechanism ofligand-induced KIT activation revealed by analysis of KIT struc-ture may therefore apply for the activation of all type III RTKs.

Studies exploring a variety of receptor mutants or usingmonoclonal antibodies that bind specifically to individual Ig-like

domains of KIT (19), PDGFRs (20), and other type III RTKshave proposed that D4 plays a role in mediating receptordimerization even when KIT is stimulated by monovalent SCFligands (21). However, quantitative analysis using microcalorim-etry of SCF binding and SCF stoichiometry toward a purifiedextracellular domain of KIT composed of either the first threeIg-like domains (D1–D3) or all five Ig-like domains (D1–D5)have shown that D4 and D5 are dispensable for SCF stimulationof KIT dimerization. In other words, KIT dimerization is pri-marily driven by the dimeric nature of SCF binding to KIT (22).Rather than playing a role in receptor dimerization, the homo-typic D4 and presumably also homotypic D5 interactions be-tween neighboring receptors are required for precise positioningof the membrane proximal regions of two receptors at a distanceand orientation that enable interactions between their cytoplas-mic domains resulting in tyrosine kinase activation. Therefore,rather than interfering with receptor dimerization, monoclonalantibodies that bind to D4 of PDGFRs, KIT, or other type IIIRTK most likely exert their inhibitory effect on receptor acti-vation by preventing critical homotypic interactions betweenmembrane proximal regions of type III RTK that are essentialfor positioning the cytoplasmic domain at a distance and orien-tation essential for tyrosine kinase activation.

The experiments presented in this report, together with earlierbiophysical (22) and structural (12) studies, demonstrate thatdimerization of PDGFR�, KIT, and other type III RTKs isentirely driven by ligand binding, and that the sole role of ligandbinding is to cross-link two receptor molecules to increase theirlocal concentration in the cell membrane. The two salt bridges(with interface of a buried surface area of 360 Å2) responsiblefor mediating homotypic D4 interactions are too weak to supportreceptor interactions without the support of ligand-mediatedreceptor dimerization, which in the case of KIT is mediated bya variety of strong interactions with a total buried surface areaof 2,060 Å2 for each SCF protomer. We have used an approachbased on ‘‘average distance to nearest neighbor calculation’’ (23,24) to estimate that the apparent concentration of a receptor inthe cell membrane of an unstimulated cell expressing 20,000receptors per cell is �1–3 �M. Upon binding a dimeric ligandsuch as SCF, two occupied receptors are held together at adistance of 75 Å (12). Under this condition, the apparentreceptor concentration in the cell membrane calculated by usingthe average distance to nearest-neighbor approach is increasedby 2 orders of magnitude to 4–6 10�4 M. This calculationshows that even weak interactions with a dissociation constant inthe range of 10�4-10�5 M, such as those mediated by the two saltbridges, could mediate association and direct contacts betweenmembrane proximal regions of two neighboring receptors. Thehigh local concentration in the cell membrane together withthe flexibility of the joints connecting D4 and D5 to the restof the receptor molecule enable movement and formation ofhomotypic D4 and homotypic D5 contacts that position themembrane proximal region of the receptor at a precise orien-tation and distance (15 Å in the case of KIT) that enableinteractions between neighboring cytoplasmic domains, tyrosineautophosphorylation, and stimulation of tyrosine kinase activity.

Finally, applying a chemical cross-linking agent to covalentlycross-link WT or mutant receptors on unstimulated or PDGF-stimulated cells, we demonstrate that an E390A PDGFR�mutant undergoes PDGF-induced dimerization similar toPDGF-induced dimerization of WT receptors. However, bycontrast to WT PDGFR� that is expressed on the cell surface ofPDGF-stimulated cells in the form of activated dimers, theE390A mutant is expressed on the surface of PDGF-stimulatedcells in the form of inactive dimers. This experiment demon-strates that homotypic D4–D4 interactions are dispensable forPDGFR� dimerization, and that PDGFR� dimerization is nec-essary but not sufficient for receptor activation.

Fig. 5. Altered kinetics of the ligand–receptor complex internalization andreceptor degradation in cells expressing PDGFR D4 mutants. (A) Cells wereincubated with 5 ng/ml 125I-PDGF for 90 min, and unbound ligand was re-moved. Cells were transferred to 37°C for the indicated time intervals. Cellsurface receptor-associated (■ ), internalized (Œ), and degradation product (�)of 125I-PDGF were determined and expressed as percentage of total binding att � 0 min. Each point was performed in triplicates and presented as mean �standard error. (B) Cells were pretreated with 10 �g/ml cycloheximide for 30min before PDGF stimulation. PDGF (20 ng/ml) was added for the indicatedtime. Cell lysates were immunoprecipitated with PDGFR antibodies and im-munoblotted with anti-flag antibodies. Total cell lysates were immunoblottedwith anti-actin antibodies as control.

Yang et al. PNAS � June 3, 2008 � vol. 105 � no. 22 � 7685

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

1

Experimental ProceduresCell Lines and Retroviral Infection. MEFs deficient in both PDGFR� and PDGFR�

(PDGFR�/�) were provided by Philippe Soriano (Fred Hutchinson Cancer Re-search Center, Seattle) and Andrius Kazlauskas (Harvard Medical School,Boston). PDGFR� cDNA was provided by Daniel DeMaio (Yale University, NewHaven, CT). PDGFR� cDNA was subcloned into pLXSHD retroviral vector, anda flag-tag was added to the C terminus of the receptor. All mutants in D4 weregenerated by site-directed mutagenesis according to manufacturer’s instruc-tion (Stratagene). Retrovirus encoding WT and mutant PDGFR� was producedin 293GPG cells (25). After infection, cells were selected with L-histidinol, andpools of selected cells were used in the experiments.

In Vitro Phosphorylation Assay for PDGFR. Cells were serum-starved for 16 hand solubilized in lysis buffer containing 150 mM NaCl, 50 mM Hepes (pH 7.4),1 mM EDTA, 25 mM NaF, 0.1 mM sodium orthovanadate, 5 �g/ml leupeptinand aprotinin, 1 mM PMSF, and 1% Nonidet P-40. Lysates were immunopre-cipitated with anti-PDGFR� antibodies, and immunopellets were incubated inreaction buffer containing 50 mM Hepes (pH 7.4), 1 mM ATP, and 10 mMMgCl2 at room temperature for 5 min. After incubation, pellets were analysisby SDS/PAGE followed by immunoblotting with antiphosphotyrosine anti-bodies. Membrane was stripped off and reblotted with anti-Flag tag antibod-ies for determination of total PDGFR� level.

Chemical Cross-Linking of Receptor Dimers. Cells were serum-starved for 16 hbefore incubation with the indicated concentration of PDGF in DMEM con-taining 50 mM Hepes (pH 7) at 4°C. After 90 min, the cells were extensivelywashed with PBS (pH 7.4). Plates were transferred to room temperature, anddisuccinimidyl suberate (DSS) was added to a final concentration of 0.5 mM.The cross-linking reaction was quenched after 30 min by incubation with 10mM Tris buffer. Cell lysates were immunoprecipitated with anti-PDGFR anti-bodies and resolved by SDS/PAGE. Nitrocellulose membrane was immunoblot-ted with antibodies against flag-tag or antiphosphotyrosine (4G10) antibod-ies to detect the total receptor and phosphorylated receptor level,respectively.

PDGF-Induced Actin Cytoskeletal Reorganization. Followed by overnight serumstarvation, cells were either treated with 50 ng/ml PDGF for 2, 5, 10, or 30 min

or left untreated. Cells were fixed in 4% paraformaldehyde in PBS, perme-abilized with 0.1% Triton and stained with FITC-phalloidin (Sigma). Coverslipswere mounted with Prolong Antifade mounting medium (Invitrogen), andimages were acquired with Nikon fluorescence microscope. Approximately400 cells on each coverslip were analyzed, and the percentage of cells showingactin ring formation was calculated and presented linearly.

PDGF Binding and Internalization Experiments. PDGF was labeled by usingBolton–Hunter reagent (Pierce) before iodination by using Iodo-gen iodina-tion tubes (Pierce), according to the manufacturer’s instructions. Cells werewashed twice in cold DMEM containing 20 mM Hepes (pH 7.4) and 0.1% BSA.Triplicate wells were incubated with 5 ng/ml of 125I-PDGF in the presence ofincreasing amounts of native PDGF. Binding was allowed to proceed at 25°Cfor 1 h. Cells were then washed in cold PBS and solubilized in 0.5 M NaOH. Theradioactive content of the samples was determined by using a LS6500 scintil-lation counter (Beckman Coulter), and the data were analyzed by using PRISMsoftware (GraphPad).

For the internalization experiment, cells were incubated with 5 ng/ml125I-PDGF in DMEM/0.1% BSA/50 mM Hepes, pH 7.4, for 90 min at 4°C.Unbound ligand was removed by washing with ice-cold PBS (pH 7.4). Pre-warmed DMEM/0.1% BSA/50 mM Hepes was added to the cells and incubatedat 37°C for the time indicated. Cell surface-associated ligand was collectedwith ice-cold acidic buffer containing PBS (pH 3) and 0.1% BSA for 10 min.Internalized ligands were collected by solubilization with 0.5 M NaOH. Theamount of degraded 125I-PDGF was determined by precipitation of the incu-bation medium with 10% trichloroacetic acid (TCA) and counting the super-natant for the TCA soluble fraction. Radioactive content of the samples wasdetermined by using a LS6500 scintillation counter (Beckman Coulter). Theamount of surface-bound, intracellular, and degraded PDGF was expressed asa percentage of total cell-associated radioactivity after 90-min incubation onice (t � 0 min). Each time point was performed in triplicate, and the resultswere expressed as mean � SE.

For additional experimental procedures, see SI Experimental Procedures.

ACKNOWLEDGMENTS. This work was supported by National Institutes ofHealth Grants AR 051448, AR 051886, and P50 AR054086. Satoru Yuzawa wassupported by a fellowship from the Uehara Memorial Foundation.

1. Schlessinger J (2000) Cell signaling by receptor tyrosine kinases. Cell 103:211–225.2. Pawson T, Nash P (2003) Assembly of cell regulatory systems through protein interac-

tion domains. Science 300:445–452.3. Hunter T (2000) Signaling-2000 and beyond. Cell 100:113–127.4. Burgess AW, et al. (2003) An open-and-shut case? Recent insights into the activation

of EGF/ErbB receptors. Mol Cell 12:541–552.5. Schlessinger J, et al. (2000) Crystal structure of a ternary FGF-FGFR-heparin complex

reveals a dual role for heparin in FGFR binding and dimerization. Mol Cell 6:743–750.6. Oefner C, D’Arcy A, Winkler F, Eggimann B, Hosang M (1992) Crystal structure of

human platelet-derived growth factor BB. EMBO J 11:3921–3926.7. Jiang X, et al. (2000) Structure of the active core of human stem cell factor and analysis

of binding to its receptor kit. EMBO J 19:3192–3203.8. Zhang Z, Zhang R, Joachimiak A, Schlessinger J, Kong XP (2000) Crystal structure of

human stem cell factor: implication for stem cell factor receptor dimerization andactivation. Proc Natl Acad Sci USA 97:7732–7737.

9. Pandit J, Bohm A, Jancarik J, Halenbeck R, Koths K, Kim SH (1992) Three-dimensionalstructure of dimeric human recombinant macrophage colony-stimulating factor. Sci-ence 258:1358–1362.

10. Savvides SN, Boone T, Andrew Karplus P (2000) Flt3 ligand structure and unexpectedcommonalities of helical bundles and cystine knots. Nat Struct Mol Biol 7:486–491.

11. Hubbard SR (1999) Structural analysis of receptor tyrosine kinases. Prog Biophys MolBiol 71:343–358.

12. Yuzawa S, et al. (2007) Structural basis for activation of the receptor tyrosine kinase KITby stem cell factor. Cell 130:323–334.

13. Soriano P (1994) Abnormal kidney development and hematological disorders in PDGFbeta-receptor mutant mice. Genes Dev 8:1888–1896.

14. Soriano P (1997) The PDGF alpha receptor is required for neural crest cell developmentand for normal patterning of the somites. Development 124:2691–2700.

15. Andrews A, et al. (1999) Platelet-derived growth factor plays a key role in proliferativevitreoretinopathy. Invest Ophthalmol Vis Sci 40:2683–2689.

16. Cochet C, et al. (1988) Demonstration of epidermal growth factor-induced receptordimerization in living cells using a chemical covalent cross-linking agent. J Biol Chem263:3290–3295.

17. Heldin CH, Ernlund A, Rorsman C, Ronnstrand L (1989) Dimerization of B-type platelet-derived growth factor receptors occurs after ligand binding and is closely associatedwith receptor kinase activation. J Biol Chem 264:8905–8912.

18. Grassot J, Gouy M, Perriere G, Mouchiroud G (2006) Origin and molecular evolution ofreceptor tyrosine kinases with immunoglobulin-like domains. Mol Biol Evol 23:1232–1241.

19. Blechman JM, et al. (1995) The fourth immunoglobulin domain of the stem cell factorreceptor couples ligand binding to signal transduction. Cell 80:103–113.

20. Miyazawa K, et al. (1998) Role of immunoglobulin-like domains 2–4 of the platelet-derived growth factor alpha-receptor in ligand-receptor complex assembly. J BiolChem 273:25495–25502.

21. Lev S, Yarden Y, Givol D (1992) Dimerization and activation of the kit receptor bymonovalent and bivalent binding of the stem cell factor. J Biol Chem 267:15970–15977.

22. Lemmon MA, Pinchasi D, Zhou M, Lax I, Schlessinger J (1997) Kit receptor dimerizationis driven by bivalent binding of stem cell factor. J Biol Chem 272:6311–6317.

23. Klein P, Mattoon D, Lemmon MA, Schlessinger J (2004) A structure-based model forligand binding and dimerization of EGF receptors. Proc Natl Acad Sci USA 101:929–934.

24. Chandrasekhar S (1943) Stochastic problems in physics and astronomy. Rev Mod Phys15:1.

25. Ory DS, Neugeboren BA, Mulligan RC (1996) Stable human-derived packaging cell linefor production of high titer retrovirus/vesicular stomatitis virus G pseudotypes. ProcNatl Acad Sci USA 93:11400–11406.

7686 � www.pnas.org�cgi�doi�10.1073�pnas.0802896105 Yang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

1