Structural basis for fibroblast growth factor receptor2 activation in Apert syndromeOmar A. Ibrahimi*, Anna V. Eliseenkova*, Alexander N. Plotnikov*, Kai Yu†, David M. Ornitz†,and Moosa Mohammadi*‡

*Department of Pharmacology, New York University School of Medicine, New York, NY 10016; and †Department of Molecular Biology and Pharmacology,Washington University Medical School, St. Louis, MO 63110

Communicated by Joseph Schlessinger, Yale University School of Medicine, New Haven, CT, April 12, 2001 (received for review March 1, 2001)

Apert syndrome (AS) is characterized by craniosynostosis (prema-ture fusion of cranial sutures) and severe syndactyly of the handsand feet. Two activating mutations, Ser-2523 Trp and Pro-2533Arg, in fibroblast growth factor receptor 2 (FGFR2) account fornearly all known cases of AS. To elucidate the mechanism by whichthese substitutions cause AS, we determined the crystal structuresof these two FGFR2 mutants in complex with fibroblast growthfactor 2 (FGF2) . These structures demonstrate that both mutationsintroduce additional interactions between FGFR2 and FGF2,thereby augmenting FGFR2–FGF2 affinity. Moreover, based onthese structures and sequence alignment of the FGF family, wepropose that the Pro-253 3 Arg mutation will indiscriminatelyincrease the affinity of FGFR2 toward any FGF. In contrast, theSer-252 3 Trp mutation will selectively enhance the affinity ofFGFR2 toward a limited subset of FGFs. These predictions areconsistent with previous biochemical data describing the effects ofAS mutations on FGF binding. Alterations in FGFR2 ligand affinityand specificity may allow inappropriate autocrine or paracrineactivation of FGFR2. Furthermore, the distinct gain-of-functioninteractions observed in each crystal structure provide a model toexplain the phenotypic variability among AS patients.

The fibroblast growth factor receptor (FGFR) signaling path-way is involved in a variety of critical physiological and

pathological processes (1). Activating mutations in the extracel-lular domains of FGFR1–3 are responsible for many humanskeletal disorders (2), including the craniosynostosis syndromes(3). The craniosynostosis syndromes are mostly caused by gain-of-function mutations in FGFR2 and include Apert syndrome(AS), Crouzon syndrome, Crouzon syndrome with acanthosisnigricans, coronal craniosynostosis, Pfeiffer syndrome, Jackson–Weiss syndrome, Antley–Bixler syndrome, and Beare–Stevensoncutis gyrata. AS [ref. 4; Online Mendelian Inheritance in Mandatabase at http:yywww3.ncbi.nlm.nih.govyomimy], the mostsevere of the craniosynostosis syndromes, is unique among thecraniosynostosis syndromes in that it is also characterized bysevere syndactyly (bony and cutaneous) of the hands and feet.Numerous other anomalies, including central nervous systemabnormalities, cardiovascular defects, urogenital anomalies, anddermatologic manifestations, also have been observed in ASpatients. Interestingly, statistically significant differences weredescribed for severity of syndactyly and presence of cleft palatebetween two subgroups of AS patients with Ser-252 3 Trp(Ser252Trp) or Pro-253 3 Arg (Pro253Arg) mutations (5, 6),which account for '67% and 32% of AS patients (6–9),respectively.

Most mutations in the extracellular domain of FGFR2 causereceptor activation by facilitating dimerization. For example,Crouzon syndrome is caused by mutations that create an un-paired cysteine residue either by gain or loss of an ectodomaincysteine residue or by destabilizing disulfide bond formation inIg-like domain 3 (D3). Free cysteines then form intermoleculardisulfide bridges between two FGFRs, leading to ligand-independent receptor dimerization and activation (10–12). Incontrast, AS is caused by substitution of one of two adjacent

residues, Ser252Trp or Pro253Arg, in the highly conserved linkerregion between Ig-like domain 2 (D2) and D3 (7). Because of thenature and location of these mutations, it is unlikely thatconstitutive activation via intermolecular disulfide bond forma-tion is the etiology of AS. Nevertheless, it is still probable thatAS results from FGFR2 gain-of-function because these muta-tions decrease the dissociation rate of FGFR2 from FGFs in vitro(13, 14) and cause ligand-dependent FGFR2 activation in vivo(14, 15). A common theme in models put forth for FGFR2hyperactivation by AS mutations suggests that AS mutationscause a conformational change in FGFR2 that increases ligand-binding affinity (13, 14, 16, 17). However, based on the proximityof Ser-252 and Pro-253 to FGF2 in the crystal structure ofFGF2–FGFR2 complex, we recently proposed that AS muta-tions could create additional interactions between FGFR andFGF that may lead to increased ligand-binding affinity (18). Inthis report, we describe the crystal structures of these mutantFGFR2s in complex with FGF2. Based on these structures andsequence alignment analysis of the FGF family, we propose amodel that accounts for the phenotypic differences observedbetween the two subgroups of AS patients.

Materials and MethodsProtein Expression and Purification. The point mutationsSer252Trp and Pro253Arg were introduced into the ligand-binding domain of human FGFR2(IIIc) (residues 147–366) byusing the Quik Change site-directed mutagenesis kit (Strat-agene) and using a previously reported bacterial expressionconstruct of wild-type FGFR2(IIIc) (residues 147–366) as tem-plate (18). The wild-type and mutant FGFR2(IIIc) proteins wereexpressed in Escherichia coli, refolded in vitro, and complexedwith full-length FGF2 immobilized on a heparin-Sepharosecolumn by using a previously reported protocol (18). Theresulting complexes were further purified by size exclusionchromatography on Superdex 200 column (Amersham Pharma-cia), using previously reported conditions (18).

DNA fragments generated by PCR of full-length human FGF2cDNA (residues 1–155) were subcloned into the pET-28a bac-terial expression vector by using NcoI and XhoI cloning sites. Toavoid intermolecular disulfide-bridge formation that could ham-per crystallization, two solvent-exposed cysteines in FGF2(Cys-78 and Cys-96) were replaced by serines. After transfor-mation of the BL21 (DE3) E. coli strain, cells containing theFGF2-expression plasmid were induced with 1 mM isopropyl1-thio-b-D-galactopyranoside for 5 h. The bacteria were thencentrifuged and subsequently lysed in a 25 mM Hepes–NaOH

Abbreviations: AS, Apert syndrome; FGF, fibroblast growth factor; FGFR, FGF receptor;D2 and D3, Ig-like domains 2 and 3.

Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,www.rcsb.org (PDB ID codes 1II4 and 1IIL).

‡To whom reprint requests should be addressed. E-mail: mohammad@saturn.med.nyu.edu.

The publication costs of this article were defrayed in part by page charge payment. Thisarticle must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.§1734 solely to indicate this fact.

7182–7187 u PNAS u June 19, 2001 u vol. 98 u no. 13 www.pnas.orgycgiydoiy10.1073ypnas.121183798

Dow

nloa

ded

by g

uest

on

Apr

il 8,

202

1

buffer (pH 7.5) containing 250 mM NaCl, by using a French cellpress (Sim Aminco, Rochester, NY). After centrifugation, sol-uble FGF2 was loaded onto a Source S column (AmershamPharmacia). Bound FGF2 was eluted by a linear gradient ofNaCl to 1 M in a 25 mM Hepes–NaOH buffer (pH 7.5). Furtherpurification of FGF2 was achieved by size exclusion chromatog-raphy on Superdex 75 (Amersham Pharmacia) equilibrated with25 mM Hepes–NaOH buffer (pH 7.5) containing 1 M NaCl.

Crystallization and Data Collection. Crystals of wild-type and ASmutant FGFR2(IIIc) in complex with full-length FGF2 (residues1–155) were generated by using previously reported crystalliza-tion conditions (18) for wild-type FGFR2(IIIc) complexed withN-terminal truncated FGF2 (residues 24–155). Two microlitersof protein solution (10 mgyml in 25 mM Hepes–NaOH, pH7.5y150 mM NaCl) were mixed with 2 ml of the crystallizationbuffer, consisting of 10–15% polyethylene glycol 4000 and 10%isopropyl alcohol in 0.1 M Hepes–NaOH (pH 7.5). All threecomplexes formed triclinic crystals, space group P1, with fourmolecules of FGF2 and four molecules of FGFR2 in the unit celland a solvent content of '58%. The unit cell dimensions are asfollows: a 5 71.681 Å, b 5 73.567 Å, c 5 90.828 Å, a 5 90.009°,b 5 90.164°, and g 5 90.005° for the wild-type complex; a 570.970 Å, b 5 72.657 Å, c 5 89.739 Å, a 5 89.985°, b 5 89.696°,and g 5 89.999° for the Ser252Trp mutant complex, and a 570.796 Å, b 5 72.490 Å, c 5 89.919 Å, a 5 90.011°, b 5 89.816°,and g 5 89.996° for the Pro253Arg mutant complex.

Crystals were flash-frozen in dry nitrogen stream by usingmother liquor containing 10% glycerol as cryoprotectant. Dif-fraction data for the Ser252Trp mutant FGFR2–FGF2 complexwere collected on an R-Axis IIC image plate detector at homeinstitution x-ray generator [Rigaku RU-200 rotating anodesource (Cu Ka) equipped with double-focusing mirrors] oper-ated at 50 kV and 100 mA. Diffraction data for the wild-type andPro253Arg mutant FGFR2(IIIc)–FGF2 complexes were col-lected on a charge-coupled device detector at beamline X4A atthe National Synchrotron Light Source, Brookhaven NationalLaboratory. All data were processed by using DENZO andSCALEPACK (19).

Structure Determination and Refinement. Molecular replacementsolutions for the four copies of FGFR2(IIIc)–FGF2 complex inthe unit cells of wild-type and mutant FGFR2(IIIc)–FGF2crystals were found with AMORE (20) by using the structure ofwild-type FGFR2(IIIc) complexed with the N-terminal trun-cated FGF2 (Protein Data Bank identification code 1EV2) (18)as the search model. Tight noncrystallographic symmetry re-straints were imposed throughout the refinement for the back-bone atoms of FGF2, D2 and D3. Simulated annealing andpositionalyB factor refinement were performed by using CNS(21). Model building into 2Fo 2 Fc and Fo 2 Fc electron density

maps was performed with program O (22). The average B factorsfor all of the protein atoms are 33.4 Å2 for the wild-type, 39.2 Å2

for the Ser252Trp mutant, and 43.3 Å2 for the Pro253Arg mutantcomplexes. The refined model for the wild-type FGF2–FGFR2(IIIc) structure is composed of four FGF2 molecules(residues 25–154) and four FGFR2(IIIc) molecules (residues150–361). The bC–bC9 loop in D3 (residues 295–306) in all fourcopies of FGFR2 has a poor electron-density map and is notincluded in the final model. The refined model for the Ser252Trpmutant FGFR2(IIIc)–FGF2 structure contains four FGF2 mol-ecules (residues 21–154) and four FGFR2(IIIc) molecules (res-idues 150–361). As in the wild-type structure, the bC-bC9 loopin D3 (residues 295–306) in all of the four copies of theSer252Trp FGFR2(IIIc) has poor electron density map and arenot included in the final model. The refined model for thePro253Arg mutant FGFR2(IIIc)–FGF2 structure is composedof four FGF2 molecules (residues 25–154), four FGFR2(IIIc)molecules (residues 150–361), and 77 water molecules. In thisstructure, residues 295–306 (bC-bC9 loop in D3) in FGFR2(IIIc)and residues 22–24 at the N terminus of FGF2 have sufficient2Fo 2 Fc electron density in two copies of FGFR2(IIIc) andFGF2, respectively, and are modeled. However, the B values forthese residues are high (above 80.0 Å2).

ResultsWild-type and AS mutant FGFR2 ectodomains were producedin E. coli, refolded in vitro, and complexed with full-length FGF2(residues 1–155). The resulting complexes were then purified andcrystallized by using previously reported crystallization condi-tions for wild-type FGFR2 complexed with an N-terminaltruncated FGF2 (residues 24–155) (18). All three FGFR2–FGF2 complexes formed triclinic crystals with unit cell dimen-sions similar to those of previously reported FGFR2–FGF2(24–155) crystals (18). Crystal structures of the wild-type and mutantcomplexes were solved by molecular replacement by using thestructure of the FGFR2–FGF2(24–155) complex as the searchmodel and were refined to 2.6 (wild-type), 2.7 (Ser252Trp), and2.3 (Pro253Arg) Å. The atomic model for each complex consistsof four FGF2 and four FGFR2 molecules. In addition, thePro253Arg mutant FGFR2–FGF2 structure contains 77 orderedwater molecules. Data collection and refinement statistics aregiven in Table 1.

Overall Structures. The overall structure of each AS mutantFGFR2–FGF2 complex is similar to the wild-type FGF2–FGFR2structure. Superimposition of the Ca traces of mutant FGFR2s withthat of wild-type FGFR2 gives a root-mean-square deviation of only0.39 Å (for Ser252Trp) and 0.31 Å (for Pro253Arg), demonstratingthat the relative disposition of the D2 and D3 domains is notaffected by the mutations (Fig. 1). This observation contradicts aprevious model proposing that AS mutations augment FGF-

Table 1. Summary of crystallographic analysis

Structure

Data collection statistics Refinement statistics*

Resolution,Å

Reflections(totalyunique)

Completeness,%

Rsym,†

%Signal,^IysI&

Resolution,Å Reflections

RcrystyRfree,‡

%

Root-mean square deviations

Bonds, Å Angles, ° B factors,§ Å2

Wild type 30.0–2.5 150,523y65,847 96.4 (78.9)¶ 4.9 (18.4)¶ 18.6 25.0–2.6 54,637 24.8y27.0 0.008 1.4 1.1Ser252Trp 30.0–2.7 74,426y46,770 91.6 (79.7)¶ 5.0 (24.5)¶ 12.1 25.0–2.7 46,161 23.8y26.7 0.008 1.4 1.2Pro253Arg 30.0–2.3 321,836y85,442 97.8 (95.9)¶ 4.1 (13.7)¶ 26.1 25.0–2.3 75,473 23.7y26.0 0.008 1.5 1.4

*Atomic model: 10,059 protein atoms (wild type); 10,280 protein atoms (Ser252Trp); 10,669 protein atoms and 77 water molecules (Pro253Arg).†Rsym 5 100 3 ShklSiuIi(hkl ) 2 ^I(hkl )&uyShklSiIi(hkl ).‡RcrystyRfree 5 100 3 ShkliFo(hkl )u 2 uFc(hkl )iyShkluFo(hkl )u, where Fo (.0s) and Fc are the observed and calculated structure factors, respectively. Five percentof the reflections were used for calculation of Rfree.

§For bonded protein atoms.¶Value in parentheses is for the highest resolution shell: 2.59–2.50 Å (wild type), 2.80–2.70 Å (Ser252Trp), and 2.38–2.30 Å (Pro253Arg).

Ibrahimi et al. PNAS u June 19, 2001 u vol. 98 u no. 13 u 7183

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 8,

202

1

binding affinity by increasing the rigidity of the D2-D3 linker regionand accentuating a conformational change in FGFR2 that occursupon FGF binding (13–14).

Of interest, a recent report of a ternary FGF1–FGFR2-heparin complex suggested that the Ser252Trp mutation acti-vates FGFR2 by facilitating a trans- to cis-isomerization ofPro-253 that occurs upon heparin binding (17). However, in theSer252Trp mutant structure, as in the wild-type structure,Pro-253 clearly remains in a trans configuration. Moreover, in allpreviously reported binary FGF–FGFR structures (18, 25–26) orternary FGF2–FGFR1-heparin structure (27), Pro-253 is foundonly in a trans configuration. The trans- to cis-isomerization ofPro-253 observed in the FGF1–FGFR2-heparin structure (17)may be the result of partial refolding of FGFR2 rather than aconsequence of a heparin-catalyzed conformational change.Instead, the crystal structures presented here unequivocallydemonstrate that AS mutations lead to aberrant FGFR2 acti-vation solely by altering ligand affinity and specificity.

FGF2 in Complex with Ser252Trp Mutant FGFR2. In the crystalstructure of wild-type FGF2–FGFR2, the N-terminal residuesbefore His-25 in FGF2 are disordered and not included in theatomic model. However, in the Ser252Trp mutant FGF2–FGFR2 structure, the difference Fourier electron density map

allows for the modeling of four additional N-terminal residues(Phe21-Pro-Pro-Gly24) of FGF2. The favorable ordering of theseresidues is due to the introduction of a unique set of interactionsbetween the Ser252Trp mutant FGFR2 and FGF2 (Fig. 2A).These interactions can be attributed to the formation of ahydrophobic patch composed of Trp-252, Tyr-281, and Ile-257 ofFGFR2 that is engaged by Phe-21 of FGF2. Although Tyr-281and Ile-257 are inherently present in wild-type FGFR2, thishydrophobic patch cannot form efficiently in the absence of ahydrophobic residue at position 252. This is further evident bythe wild-type FGF2–FGFR2 structure, where the side chain ofTyr-281 is observed to exist in two alternating rotamer positions(data not shown). The stabilization of this hydrophobic patch byTrp-252 allows the side chain hydroxyl group of Tyr-281 to forma hydrogen bond with the FGF backbone residue Pro-22 (Fig.2A). In addition, the FGF backbone carbonyl oxygen of Gly-24engages in an oxygen-aromatic interaction with phenyl ring ofTyr-281. The presence of these additional interactions will likelystrengthen FGF2–FGFR2 affinity and is consistent with datademonstrating that, compared to wild-type FGFR2, theSer252Trp mutant has an approximate 6.5-fold decrease in therate of dissociation (koff) from FGF2 (13). Of interest, there aretwo instances of AS patients with a Ser252Phe mutation (6, 8)that is likely to activate FGFR2 through a similar mechanism.

FGF2 in Complex with Pro253Arg Mutant FGFR2. Activation ofFGFR2 by the Pro253Arg mutation is due to the presence of aset of FGFR2–FGF2 interactions that is not observed in eitherthe Ser252Trp mutant or in the wild-type FGFR2–FGF2 struc-tures (Fig. 2B). The guanidinium group of Arg-253 in FGFR2makes three hydrogen bonds with FGF2, two with backbonecarbonyl oxygens of Leu-107 and Glu-108 and a third with theside-chain amide group of Asn-111. These additional hydrogenbonds are also predicted to increase ligand-receptor affinity.These additional interactions will also likely strengthen FGF2–FGFR2 affinity and are consistent with data demonstrating thatthe Pro253Arg mutant has an almost 1.5-fold decrease in kofffrom FGF2, relative to wild-type FGFR2. It is also noteworthythat homologous mutations in FGFR1 (Pro252Arg) and FGFR3(Pro250Arg) are responsible for Pfeiffer syndrome (28) andMuenke craniosynostosis (29), respectively. Hence, these mutantreceptors are also likely to engage in similar additional contactswith FGF.

DiscussionImplications for Ligand Binding. Receptor-binding specificity is anessential mechanism in the regulation of FGF responses and isprimarily achieved through alternative splicing of two exons, IIIband IIIc, in the second half of D3 in FGFRs (30). The critical roleof second half of D3 in conferring specificity is best exemplifiedby the FGFR2(IIIc)yFGFR2(IIIb) system. FGFR2(IIIc) bindsFGF1 and FGF2 but does not bind FGF7. In contrast,FGFR2(IIIb) binds FGF1 and FGF7 but binds FGF2 withnegligible affinity. Because AS mutations are located in theD2-D3 linker region, they affect both the FGFR2(IIIb) andFGFR2(IIIc) splice variants.

The basis for increased affinity of FGF2 for the Ser252Trpmutant FGFR2(IIIc) is the hydrophobic contact between Phe-21of FGF2 and Trp-252, Tyr-281, and Ile-257 of FGFR2(IIIc).Therefore, the Ser252Trp mutant FGFR2(IIIc) is predicted toexhibit increased affinity toward FGFs with a hydrophobicresidue at a position corresponding to Phe-21 of FGF2. A surveyof the 22 human FGF sequences reported to date (18, 31) revealsthat the N-terminal region outside the b-trefoil core of FGFs ishighly divergent and that only a limited subset of FGFs, mostnotably FGF7 (KGF), fulfill this criterion (Fig. 3A). Interest-ingly, we have also detected an increased interaction betweenFGF10 and the Ser252Trp mutant FGFR2 by size-exclusion

Fig. 1. AS mutations do not affect the relative disposition of D2 and D3 inFGFR2. Ca traces of wild-type (green), mutant Ser252Trp (blue), andPro253Arg (red) FGFR2s are superimposed. The N and C termini are denotedby the letters NT and CT. This figure was made by using programs MOLSCRIPT (23)and RASTER3D (24). The Ca traces of Ser252Trp and Pro253Arg mutant FGFR2sdeviate by only 0.39 Å and 0.31 Å, respectively, from the Ca trace of wild-typeFGFR2.

7184 u www.pnas.orgycgiydoiy10.1073ypnas.121183798 Ibrahimi et al.

Dow

nloa

ded

by g

uest

on

Apr

il 8,

202

1

chromatography (M.M., unpublished data). FGF10 has an as-paragine (Asn-71) at a position corresponding to Phe-21 ofFGF2 (Fig. 3A) and is therefore not predicted to engage inadditional hydrophobic interactions with the Ser252Trp mutantFGFR2. It is possible that the inherent flexibility of the Nterminus of FGFs preceding the rigid b-trefoil core may permitadditional interactions to take place between Tyr-70 (adjacent toAsn-71) of FGF10 and the Ser252Trp mutant FGFR2.

In contrast, the three hydrogen bonds between Arg-253 andFGF2 in the Pro253Arg mutant FGFR2(IIIc)–FGF2 involveresidues within the conserved b-trefoil core of FGFs. Becausetwo of these hydrogen bonds are made with the backbone atomsof FGF2, we predict the Pro253Arg mutant FGFR2(IIIc) shoulddisplay a general increase in affinity toward all 22 known human

FGFs. Interestingly, the third hydrogen bond with FGF2 ismediated by the side chain of Asn-111, which is also conservedin FGF4, FGF6, FGF8, FGF17, and FGF18 (Fig. 3B). TheseFGFs are predicted to show a more marked increase in affinityfor the Pro253Arg mutant FGFR2(IIIc) in comparison toother FGFs.

Of interest, each AS mutation is not only predicted to increasethe affinity of FGFR2(IIIc) for its specific ligands but also mayenable FGFR2(IIIc) to bind ligands that it does not normallybind. For example, the presence of a tyrosine residue (Tyr-58) inFGF7 in a homologous position to Phe-21 of FGF2 (Fig. 3A) mayallow FGF7 to bind to the Ser252Trp mutant FGFR2(IIIc).Similarly, the Pro253Arg mutant FGFR2(IIIc) may bind toFGF7 via two additional nonspecific hydrogen bonds as ob-

Fig. 2. Gain-of-function interactions between the AS mutant FGFR2s and FGF2. (A) Additional contacts between the N terminus of FGF2 and the Ser252Trpmutant FGFR2. (B) Hydrogen bonds between FGF2 and Arg-253 of the Pro253Arg mutant FGFR2. D2 and D3 of FGFR2 are shown in green and cyan, respectively.The short linker that connects D2 and D3 is colored gray. FGF2 is shown in orange. In addition, the FGF2 N-terminal region (Phe21-Pro-Pro-Gly24), which is orderedonly in the Ser252Trp mutant FGFR2–FGF2 structure, is colored purple. Oxygen atoms are red, nitrogen atoms blue, and carbon atoms have the same coloringas the molecules to which they belong. Dotted lines represent hydrogen bonds. The hydrogen-bonding distances are indicated. (Right) Views of whole structurein the exact orientation as in the detailed views are shown, and the region of interest is boxed. This figure was created by using the programs MOLSCRIPT and RASTER3D.

Ibrahimi et al. PNAS u June 19, 2001 u vol. 98 u no. 13 u 7185

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 8,

202

1

served in the crystal structure of Pro253Arg FGFR2(IIIc) incomplex with FGF2 (Fig. 2B). Indeed, our structural predictionsare supported by recent biochemical experiments that showedthat both AS mutant FGFR2s can bind to and be activated byFGF7 (14). These data strongly support our structure- andsequence-based prediction that AS mutations can affect FGFR2specificity.

Analysis of the crystal structures provides a potential expla-nation for the differential effects of AS mutations on FGF-binding kinetics. Anderson et al. (13) demonstrated that both ASmutant FGFR2s exhibit a decrease in koff from FGFs. Therelative decreases in koff values of the wild-type and Ser252Trpmutant FGFR2 for FGF1, FGF2, and FGF4 are entirely con-sistent with structure- and sequence-based predictions. Theprofound decrease in koff (6.5-fold) of FGF2 for the Ser252Trpmutant can be attributed to the hydrophobic interactions be-tween Phe-21 of FGF2 and Trp-252 and Tyr-281 of the mutantreceptor (Fig. 2 A). The aromatic ring of Phe-21 packs optimallyagainst Trp-252 and Tyr-281 of the Ser252Trp mutant FGFR2.Sequence alignment of FGFs (Fig. 3A) shows that the aliphaticside chain of Leu-18 in FGF1 can also engage in hydrophobicinteractions with Trp-252 and Tyr-281, although this interactionwould be expected to be much weaker than the interactionbetween FGF2 and the Ser252Trp FGFR2 mutant. This isconcurrently reflected by the significantly smaller decrease inkoff (1.65-fold) of FGF1 for the Ser252Trp FGFR2 mutantrelative to wild-type FGFR2.

Unlike the profound decrease in koff of FGF2 for theSer252Trp FGFR2 mutant, the reduction in koff of FGF2 for thePro253Arg FGFR2 mutant is modest. Nevertheless, this de-crease is relevant, because a similar decrease was found in theoff-rate of FGF1 binding to the same mutant receptor (13). Theuniversal decrease in koff values is consistent with the structural

observations and reflects the nature of conserved interactionspredicted to occur between the Pro253Arg FGFR2 mutant andFGFs.

A Model for Clinical Variability. Changes in ligand affinity andspecificity of AS mutants present an attractive model for thephenotypic variability observed between the two AS mutants.Two reports have documented that AS patients with theSer252Trp mutation present more frequently with cleft palate,whereas patients with the Pro253Arg mutations exhibit a moresevere syndactyly (5, 6). Normal human limb development iscontrolled by an intricate epithelial–mesenchyme paracrine sig-naling loop, in which the mesenchymally expressed FGFR2splice variant, FGFR2(IIIc), is activated by epithelially expressedFGFs (FGF2, FGF4, FGF6, FGF8, and FGF9). Similarly, theepithelially expressed splice form, FGFR2(IIIb)ykeratinocytegrowth factor receptor, is normally activated by mesenchymallyexpressed FGFs (FGF7 and FGF10). Recently, two rare cases ofAS have been reported in which de novo Alu-element insertionsresult in the ectopic expression of wild-type FGFR2(IIIb) inmesenchymal cells (8). Consequently, it may be possible thatsyndactyly results from FGF7- or FGF10-mediated autocrineactivation of ectopic mesenchymal FGFR2(IIIb)ykeratinocytegrowth factor receptor.

Based on the structural data presented in this report, wepropose that both AS mutations can allow mesenchymalFGFR2(IIIc) to bind mesenchymally expressed FGF7 andFGF10. Because of loss of FGFR2(IIIc)-binding specificity, thepathology of AS may result from inappropriate autocrine acti-vation of FGFR2(IIIc). Moreover, relative to Ser252Trp ASpatients, the phenotypically more severe syndactyly inPro253Arg AS patients can be accounted for by the higher

Fig. 3. Sequence alignment of all 22 known FGFs at regions that are involved in additional contacts with the Ser252Trp (A) or Pro253Arg (B) mutant FGFR2s.The secondary structure assignments were obtained by using the program PROCHECK (32). The location and length of the b-strands are shown on the top ofsequence alignments. The numbering of b-strands is according to the published nomenclature (33). A period represents sequence identity to FGF2. FGF2 residueswhose side-chain or main-chain atoms are engaged in additional contacts with AS FGFR2s are highlighted with red and green, respectively. The correspondingidentical residues in other FGFs also are indicated by using the same coloring scheme. In addition, hydrophobic residues in other FGFs at the position homologousto Phe-21 of FGF2 are highlighted with purple and yellow.

7186 u www.pnas.orgycgiydoiy10.1073ypnas.121183798 Ibrahimi et al.

Dow

nloa

ded

by g

uest

on

Apr

il 8,

202

1

affinity of the latter mutant FGFR2(IIIc) for FGF7 and FGF10(ref. 14 and M.M., unpublished data).

Conversely, an increase in specific ligand affinity may alsoaccount for phenotypic differences among AS patients. FGF2,which is expressed in both facial ectoderm and mesenchyme, hasbeen shown to increase outgrowth of facial mesenchyme (34).Cleft palate formation, which is more common in AS patientswith the Ser252Trp mutation (5, 6), may result from an increasein FGFR2(IIIc) affinity for a specific ligand such as FGF2. Thisis consistent with the observation that the Ser252Trp mutantFGFR2(IIIc) has a more profound decrease in koff ('4.5-foldgreater) for FGF2 than the Pro253Arg mutant FGFR2(IIIc)

(13). In addition to furnishing a structural basis for AS and amodel for phenotypic variability, these crystal structures alsoestablish a framework for the engineering of high-affinity FGFsthat may have therapeutic value for a variety of physiological andpathological conditions.

M.M. acknowledges X.-P. Kong and S. R. Hubbard for comments andhelpful discussions and C. Ogata for synchrotron beamline assistance.We also thank B. K. Yeh for critically reading the manuscript. BeamlineX4A at the National Synchrotron Light Source, a Department of Energyfacility, is supported by the Howard Hughes Medical Institute. This workwas supported by the National Institutes of Health Grants R01 DE13686(to M.M.) and R01 HD35692 (to D.M.O.).

1. McIntosh, I., Bellus, G. A. & Jabs, E. W. (2000) Cell Struct. Funct. 25,85–96.

2. Gorlin, R. J. (1997) J. Craniomaxillofac. Surg. 25, 69–79.3. Hehr, U. & Muenke, M. (1999) Mol. Genet. 68, 139–151.4. Apert, M. E. (1906) Bull. Soc. Med. Hop. Paris 23, 1310–1313.5. Slaney, S. F., Oldridge, M., Hurst, J. A., Moriss-Kay, G. M., Hall, C. M., Poole,

M. D. & Wilkie, A. O. (1996) Am. J. Hum. Genet. 58, 923–932.6. Lajeunie, E., Cameron, R., El-Ghouzzi, V., de Parseval, N., Journeau, P.,

Gonzales, M., Delezoide, A. L., Bonaventure, J., Le Merrer, M. & Renier, D.(1999) J. Neurosurg. 90, 443–447.

7. Wilkie, A. O. M., Slaney, S. F., Oldridge, M., Poole, M. D., Ashworth, G. J.,Hockley, A. D., Hayward, R. D., David, D. J., Pulleyn, L. J., Rutland, P., et al.(1995) Nat. Genet. 9, 165–172.

8. Oldridge, M., Zackai, E. H., McDonald-McGinn, D. M., Iseki, S., Morriss-Kay,G. M., Twigg, S. R., Johnson, D., Wall, S. A., Jiang, W., Theda, C., et al. (1999)Am. J. Hum. Genet. 64, 446–461.

9. Tsai, F. J., Tsai, C. H., Peng, C. T., Lin, S. P., Hwu, W. L., Wang, T. R., Lee,C. C. & Wu, J. Y. (1999) Acta Paediatr. Taiwan 40, 31–33.

10. Neilson, K. M. & Friesel, R. E. (1995) J. Biol. Chem. 270, 26037–26040.11. Neilson, K. M. & Friesel, R. (1996) J. Biol. Chem. 271, 25049–25057.12. Robertson, S. C., Meyer, A. N., Hart, K. C., Galvin, B. D., Webster, M. K. &

Donoghue, D. J. (1998) Proc. Natl. Acad. Sci. USA 95, 4567–4572.13. Anderson, J., Burns, H. D., Enriquez-Harris, P., Wilkie, A. O. M. & Heath, J. K.

(1998) Hum. Mol. Genet. 7, 1475–1479.14. Yu, K., Herr, A. B., Waksman, G. & Ornitz, D. M. (2000) Proc. Natl. Acad. Sci.

USA 97, 14536–14541.15. Mansukhani, A., Bellosta, P., Sahni, M. & Basilico, C. (2000) J. Cell Biol. 149,

1297–1308.16. Wang, F., Lu, W., McKeehan, K., Mohamedali, K., Gabriel, J. L., Kan, M. &

McKeehan, W. L. (1999) Biochemistry 38, 160–171.17. Pellegrini, L., Burke, D. F., von Delft, F., Mulloy, B. & Blundell, T. L. (2000)

Nature (London) 407, 1029–1034.

18. Plotnikov, A. N., Hubbard, S. R., Schlessinger, J. & Mohammadi, M. (2000)Cell 101, 413–424.

19. Otwinowski, Z. & Minor, W. (1997) Methods Enzymol. 276, 307–326.20. Navaza, J. (1994) Acta Crystallogr. A 50, 157–163.21. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P.,

Grosse-Kunstleve, R. W., Jiang, J. S., Kuszewski, J., Nilges, M., Pannu, N. S.,et al. (1998) Acta Crystallogr. D 54, 905–921.

22. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. (1991) Acta Crystallogr.A 47, 110–119.

23. Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946–950.24. Merritt, E. A. & Bacon, D. J. (1997) Methods Enzymol. 277, 505–524.25. Plotnikov, A. N., Schlessinger, J., Hubbard, S. R. & Mohammadi, M. (1999)

Cell 98, 641–650.26. Stauber, D. J., DiGabriele, A. D. & Hendrickson, W. A. (2000) Proc. Natl.

Acad. Sci. USA 97, 49–54.27. Schlessinger, J., Plotnikov, A. N., Ibrahimi, O. A., Eliseenkova, A. V., Yeh,

B. K., Yayon, A., Linhardt, R. J. & Mohammadi, M. (2000) Mol. Cell 6,743–750.

28. Meunke, M., Schell, U., Hehr, A., Robin, N. H., Losken, H. W., Schinzel, A.,Pulleyn, L. J., Rutland, P., Reardon, W., Malcolm, S., et al. (1994) Nat. Genet.8, 269–274.

29. Bellus, G. A., Gaudenz, K., Zackai, E. H., Clarke, L. A., Szabo, J., Francomano,C. A. & Muenke, M. (1996) Nat. Genet. 14, 174–176.

30. Johnson, D. E. & Williams, L. T. (1993) Adv. Cancer Res. 60, 1–41.31. Venkataraman, G., Raman, R., Sasisekharan, V. & Sasisekharan, R. (1999)

Proc. Natl. Acad. Sci. USA 96, 3658–3663.32. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993)

J. Appl. Crystallogr. 26, 283–291.33. Faham, S., Linhardt, R. J. & Rees, D. C. (1998) Curr. Opin. Struct. Biol. 8,

578–586.34. Richman, J. M., Herbert, M., Matovinopvic, E. & Walin, J. (1997) Dev. Biol.

189, 135–147.

Ibrahimi et al. PNAS u June 19, 2001 u vol. 98 u no. 13 u 7187

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 8,

202

1