; Prepublished online 14 May 2010;J Immunol Bin GaoTang, Raymond J. Owens, David I. Stuart, Jingshan Ren andZhang, Xuekai Zhu, Ying Wu, Lucy R. Wedderburn, Peifu Changzhen Liu, Thomas S. Walter, Peng Huang, Shiqian RANK Interaction and Signaling
Receive free email-alerts when new articles cite this article. Sign up at
initial publication. Advance online articles must include the digital object identifier (DOIs) and date ofpublication priority; they are indexed by PubMed from initial publication. Citations to available prior to final publication). Advance online articles are citable and establishnot yet appeared in the paper journal (edited, typeset versions may be posted when Advance online articles have been peer reviewed and accepted for publication but have
Print ISSN: 0022-1767 Online ISSN: 1550-6606.Immunologists, Inc. All rights reserved.
Structural and Functional Insights of RANKL–RANKInteraction and Signaling
Changzhen Liu,*,†,1 Thomas S. Walter,‡,1 Peng Huang,x Shiqian Zhang,{ Xuekai Zhu,*,†
Ying Wu,*,† Lucy R. Wedderburn,‖ Peifu Tang,x Raymond J. Owens,‡ David I. Stuart,‡
Jingshan Ren,‡ and Bin Gao*,†,‖
Bone remodeling involves bone resorption by osteoclasts and synthesis by osteoblasts and is tightly regulated by the receptor activator
of the NF-kB ligand (RANKL)/receptor activator of the NF-kB (RANK)/osteoprotegerin molecular triad. RANKL, a member of the
TNF superfamily, induces osteoclast differentiation, activation and survival upon interaction with its receptor RANK. The decoy
receptor osteoprotegerin inhibits osteoclast formation by binding to RANKL. Imbalance in this molecular triad can result in
diseases, including osteoporosis and rheumatoid arthritis. In this study, we report the crystal structures of unliganded RANK
and its complex with RANKL and elucidation of critical residues for the function of the receptor pair. RANK represents the longest
TNFR with four full cysteine-rich domains (CRDs) in which the CRD4 is stabilized by a sodium ion and a rigid linkage with CRD3.
On association, RANK moves via a hinge region between the CRD2 and CRD3 to make close contact with RANKL; a significant
structural change previously unseen in the engagement of TNFR superfamily 1A with its ligand. The high-affinity interaction
between RANK and RANKL, maintained by continuous contact between the pair rather than the patched interaction commonly
observed, is necessary for the function because a slightly reduced affinity induced by mutation produces significant disruption of
osteoclast formation. The structures of RANK and RANKL–RANK complex and the biological data presented in the paper are
essential for not only our understanding of the specific nature of the signaling mechanism and of disease-related mutations found in
patients but also structure based drug design. The Journal of Immunology, 2010, 184: 000–000.
The receptor activator of the NF-kB (RANK; TNFR super-family [TNFRSF] 11A), and its cognate ligand, RANKL,play a pivotal role in bone remodeling, immune function
and mammary gland development in conjunction with various
cytokines and hormones (1–5). Recently the pair was also foundto be important for thermoregulation (6), demonstrating them tobe one of the most versatile physiological modulators in the body.RANK is a type I transmembrane protein, consisting of around
620 aas with ∼85% homology between mouse and humanhomologs (7). The extracellular region (residues 30–194) iscomprised of four tandem cysteine-rich pseudo-repeat domains(CRDs) that are characteristic of the TNFRSF (8). The C-terminal 383 residues form one of the largest cytoplasmic domainsin the TNFRSF. Like other members of the family this region ofthe protein lacks intrinsic enzymatic activity, therefore it trans-duces intracellular signals by the recruitment of various adaptorproteins, including TNFR-associated factors (TRAFs), leading tothe activation of JNK, ERK, p38, NFATc1, AKT, and NF-kBsignaling pathways (9–13). RANKL (TNF superfamily [TNFSF]11), the only ligand binding to the extracellular portion of RANK(14) was cloned respectively by four different groups (7, 15–17)and identified as a member of the TNFSF. It is a type II trans-membrane protein, primarily expressed on the surface of activatedT-cells, bone marrow stromal cells, and osteoblasts. Soluble formsof RANKL that arise from either proteolytic processing or alter-native mRNA splicing have also been observed (18). Both themembrane-spanning and soluble forms of RANKL are assembledinto functional homotrimers like other members of the TNFSF.The binding of RANKL to RANK causes trimerisation of thereceptor, which activates the signaling pathway and results inosteoclastogenesis from progenitor cells and the activation of ma-ture osteoclasts (19–21).Osteoprotegerin (OPG, TNFRSF11B), a soluble homolog of
RANK primarily secreted by bone marrow stromal cells andosteoblasts, acts as a decoy receptor of RANKL to block thebinding of RANKL to RANK (22). The RANKL/OPG ratio isdifferently regulated between physiological and pathological
*The Center for Molecular Immunology and †China-Japan Joint Laboratory forMolecular Immunology and Virology, Key Laboratory of Pathogenic Microbiologyand Immunology, Institute of Microbiology, Chinese Academy of Sciences; xDepart-ment of Orthopaedics, Chinese People’s Liberation Army General Hospital, Beijing;{Department of Orthopaedics, the First Clinical College of Harbin Medical Univer-sity, Harbin, China ‡Division of Structural Biology, Oxford Protein Production Fa-cility, The Wellcome Trust Centre for Human Genetics, University of Oxford,Oxford; and ‖Unit of Rheumatology, University College London Institute of ChildHealth, London, United Kingdom
1C.L. and T.S.W. contributed equally to this work.
Received for publication December 15, 2009. Accepted for publication April 1, 2010.
This work was supported by the 973 Scheme of Ministry of Science and Technology(Grant 2006CB504306), the National Natural Science Foundation of China (Grant30700749 and 30600623), the National S and T Major Project (Grant 2009ZX09503-007), the Ministry of Science and Technology Science Exchange Program (Grant2007DFC30240), the Beijing Municipal Natural Science Foundation (Grant7072072), and the United Kingdom Medical Research Council and BiotechnologyBiological Sciences Research Council.
The coordinates and structure factors presented in this article for the mouse RANKand RANKL–RANK complex have been deposited in the Protein Data Bank withaccess code 3ME4 and 3ME2, respectively, (www.rcsb.org/pdb/) for immediate re-lease on publication.
Address correspondence and reprint requests to Prof. Bin Gao and Prof. Jingshan Ren,Institute of Microbiology, Chinese Academy of Sciences, 1 Beichn Xilu Road, Beijing100101, China, or Division of Structural Bioloy, Oxford Protein Production Facility, TheWellcome Trust Centre for Human Genetics, The Henry Wellcome Building for GenomicMedicine, University of Oxford, Roosevelt Drive, Oxford OX3 7BN, U.K.. E-mail ad-dresses: [email protected] and [email protected]
Abbreviations used in this paper: CRD, cysteine-rich domain; GST, glutathione S-transferase; M-CSF, macrophage-CSF; OPG, osteoprotegerin; RANK, receptoractivator of the NF-kB; RANKL, RANK ligand; rmsd, root mean square deviation;TRAF, TNFR-associated factor; TNFRSF, TNFR superfamily; TNFSF, TNF super-family; TRAP, tartrate-resistant acid phosphatase.
Copyright� 2010 by The American Association of Immunologists, Inc. 0022-1767/10/$16.00
www.jimmunol.org/cgi/doi/10.4049/jimmunol.0904033
Published May 14, 2010, doi:10.4049/jimmunol.0904033 on January 24, 2011
conditions. Various pathological conditions characterized asderegulated bone remodeling are associated with an imbalancebetween OPG and RANKL. Thus, RANKL, RANK, and OPGprovide a ligand/receptor/receptor antagonist system for control-ling bone homeostasis and other related biological processes.OPG-deficient mice exhibit a decrease in total bone density anddevelop osteoporosis (23). Mice with a genetic mutation of rank,phenotypically exactly like rankl2/2 knock-out mice (24), haveseverely defective osteoclast development (25), which can be re-stored by the reintroduction of rank cDNA into bone marrowprogenitor cells (26). In humans, mutations in the genes encodingRANKL and RANK have been found to dramatically reduce thenumber of osteoclasts and cause osteopetrosis, a disease associ-ated with a high density of bone, resulting in blindness, facialparalysis, and deafness due to the increased pressure put on thenerves by the extra bone (27, 28). Presumably the mutations dis-rupt the association of RANKL to its receptor (29). A human mAbto RANKL has been developed for treatments of post-menopausal osteoporosis and rheumatoid arthritis (30, 31).Members of the TNFSF, although structurally related, show sig-
nificant sequence diversity. Structures of several ligands, receptors,and ligand–receptor complexes have been resolved, includingTNF-a (32), RANKL (33, 34), CD40L (35), CrmE (36), TNFRSF1A(37), TNF-b–TNFRSF1A (38), and TRAIL-DR5 (39, 40) com-plexes. Members of the TNFSF are homotrimeric with a core scaf-fold of b-sandwich jellyroll topology, whereas members of theTNFRSF consist of variable numbers of CRDs, the majority ofwhich comprise five irregular b-strands linked by three disulphides(41). Receptor molecules bind to the clefts between the subunits ofthe ligand trimer to form a heterohexamer. The published structuresof ligand–receptor complexes have provided detailed informationabout receptor–ligand interactions and the functional mechanismat atomic resolution for those pairs of molecules. Although thethree-dimensional structure of mouse RANKL shows an overall foldcharacteristic of TNFSF molecules, there are large structural andconformational differences in the loops that form the receptor bind-ing cleft (33, 34). In addition, the structures of TNFRmolecules haveshown great domain flexibility between the CRDs as well as struc-tural flexibility within each CRD, and there is little sequence homol-ogy among the members of the TNFRSF. The structures of RANKand the RANKL–RANK complex are therefore essential for ourunderstanding of the basis of ligand-receptor specificity in this sys-tem and the mechanism of molecular signaling.In this study, we report crystal structures of the extracellular
region ofmouseRANKalone and in complexwith the ectodomainof RANKL. The structure of RANK contains four full-lengthCRDs and folds into an elongated shape. There are distinct fea-tures in CRD disulphide topology and domain connectivity.The structure of the RANKL–RANK complex, when comparedwith the TNF-b–TNFRSF1A and TRAIL–DR5 complexes,reveals that both the position and orientation of the boundreceptor differ significantly, and there is little conservation inthe ligand–receptor interface contacts. A sodium ion boundbetween CRD3 and CRD4 of RANK may be crucial for main-taining the structural integrity of the receptor and explains someof the disease-related mutations. The affinity between RANKLand RANK has been determined using Biacore analysis, and theresults (Kd up to 10
211 M) indicate that the pair is bound stronglytogether. Structure-guided mutations of RANKL show that thecontribution of the individual residues tested to the binding ofRANKL to RANK is directly related to RANKL signaling-dependent osteoclast formation. A slight disruption of bindingbetween RANK and RANKL would have a dramatic effect onosteoclast formation.
Materials and MethodsCloning, protein expression, and purification
Oligonucleotides were prepared by Sangon Biotech (Shanghai, China).Restriction enzymes, T4 DNA ligase, and First Strand cDNA Synthesis kitwere purchased from Fermentas (Burlington, Ontario, Canada). Pfu DNApolymerase was purchased from Tiangen Biotech (Beijing, China). Glu-tathione (reduced and oxidized) was purchased from Sigma-Aldrich(St. Louis, MO).
The cDNA coding for the extracellular domain of murine RANK(residues 26–210) was obtained by RT-PCR from the mRNA of mouseRAW264.7 cells and cloned into pET28a vector (Novagen, Madison, WI).The expression plasmid for GST–RANKL (encoding residues 159–361 ofmouse RANKL as a fusion with Glutathione S-transferase) was a gift fromProf. Fremont (Washington University School of Medicine, St. Louis,MO). RANK was expressed with a His6 tag at each terminus of the protein.Site-directed mutagenesis of the rankl was performed using QuickChangeKit supplied by Stratagene (Agilent Technologies, Palo Alto, CA). Themutants were verified by DNA sequencing. Escherichia coli strain BL21-Gold (DE3) was used to express the recombinant proteins.
The recombinant RANK was produced as inclusion bodies that weredissolved by sonication in 6 M guanidine hydrochloride, 50 mM Tris (pH8.5), 1mMEDTA, 150mMNaCl, and10mMDTT to a protein concentrationof ∼30 mg/ml at room temperature. The refolding of recombinant RANKwas performed at 4˚C by diluting the solubilized protein in 20mMNa2HPO4
(pH 7.3), 1 M L-arginine, 20% glycerol, 10 mM reduced glutathione, and1 mM oxidized glutathione, followed by sequential dialysis against 20 mMNa2HPO4 (pH 7.3), 0.5 M L-arginine, and 10% glycerol for 12 h, 20 mMNa2HPO4 (pH 7.3), 0.2 M L-arginine, and 5% glycerol for 12 h, and finallytwice against 20 mM Na2HPO4 (pH 7.3) for 12 h. After centrifugation at20,000g for 10 min, the supernatant was further purified by size exclusionchromatography (Superdex 200, GE Healthcare) and the correctly refoldedRANK was collected and analyzed by SDS-PAGE.
The soluble extracellular domain of mouse RANKL was expressed asa GST fusion protein, purified by affinity purification with glutathione-Sepharose fast flow 4B beads (GE Healthcare) according to the manufac-turer’s protocol and the tag cleaved with PreScission protease (GE health-care). The cleaved RANKL was further purified by size exclusionchromatography (Superdex 200) in Tris pH 7.0.
Crystallization and data collection
Purified RANKL and RANK were concentrated to 10 mg/mL in with 0.1 MTris at pH 7.0. Crystallization screens of RANK and RANKL–RANKcomplex were performed at a temperature of 294 K using nano-liter sittingdrop vapor diffusion in the crystallization facility of the Oxford ProteinProduction Facility (42). The best RANK crystals were grown in 10%polyethylene glycol 3350, 15% polyethylene glycol 5000, 0.1 M ammo-nium sulfate, 0.1 M sodium tartrate, and 0.05 M MES at pH 6.5. Crystalsof RANKL–RANK complex were grown in 0.1 M sodium dihydrogenphosphate, 2 M sodium chloride, 0.1 M potassium dihydrogen phosphate,and 0.1 M MES (pH 6.5), using a 1:1 molar ratio of RANKL and RANK(10 mg/ml). Details of protein purification and crystallization have beenpublished elsewhere (43).
X-ray diffraction data for RANKwere collected at beamline BM14 at theESRF (Grenoble, France). A total of 180 images of 1.0˚ oscillation werecollected from a single crystal at a wavelength of 0.954 A. X-ray diffractiondata of the RANKL–RANK complex were collected at two ESRF beam-lines; 130 images of 1.0˚ oscillation from one crystal were collected atbeamline ID14-4 at a wavelength of 0.940 A, and 180 images of 1.0˚ oscil-lation were collected from two positions of a single crystal at ID23-2 oper-ated at a wavelength of 0.873 A. In all cases, 25% glycerol was added to thecrystallization drops as cryoprotectant, and crystals were frozen and main-tained at 100 K by a stream of nitrogen gas during data collections. Dataimages were indexed, integrated, and merged using HKL2000 (44). Thestatistics for x-ray data are given in Table I.
Structure solution and refinement
The space group of the RANK crystals is P212121 with unit cell dimensionsof a = 39.8 A, b = 94.3 A, and c = 102.4 A. The RANKL–RANK complexcrystals belong to a hexagonal space group of P63 with unit cell dimensionsof a and b = 121.2 A and c = 94.7 A. The structure of the complex wassolved first using RANKL monomer (33) as a search model for molecularreplacement with MOLREP (45) of the CCP4 program suite (46). There isone RANKL subunit and one RANK molecule in the crystal asymmetricunit, giving a solvent content of 74%. The 3-fold axis of the heterohexa-meric RANKL–RANK complex is aligned with the crystallographic 3-foldaxis. The initial difference electron density map calculated from this partial
model clearly showed the RANK polypeptide and was of sufficient qualityto allow the RANK molecule of four CRDs to be built. The structure ofRANK in the complex was then used to solve the crystal structure of RANKalone. However, structure solution using molecular replacement was notstraightforward because of the thin elongated shape and the flexibility of themolecule. The correct solution was only found by using the central twoCRDs with the program PHASER (47). The initial R factor was 0.54 fordata from 30 to 4.0 A. There are two molecules in one crystal asymmetricunit related by a local 2-fold rotation axis with the N terminus of onemolecule interacting with the C terminus of the other. Superposition ofthe RANK model from the complex onto the molecular replacement solu-tion resulted in a large number of clashes between the first and fourth CRDs,indicating large conformational changes. Nevertheless, the first and fourthCRDs were built after a round of refinement using all data to 2.0 A reso-lution. Both structures were refined with the crystallography and NMRsystem (48) using simulated annealing, conjugate gradient minimization,and individual isotropic B factor refinement, followed by model rebuildingand solvent molecule addition with COOT (49). Because of the large con-formational differences between the two molecules of the RANK, no non-crystallographic symmetry restraints were applied during refinement. Thefinal refined structures of both RANK and RANKL–RANK complex havegood crystallographic R factors and stereochemistry as shown in Table I.
Surface plasmon resonance
The affinities of RANKL and its mutants for the receptor, RANK weremeasured using Biacore 3000 (GE Heathcare) according to the publishedprotocol (50). Briefly, an NTA chip (GE Heathcare) was charged with 0.3M NiSO4, and then RANK (50 nM, 15 ml) was injected into the channel toload. Recombinant TNFRSF9 with two His-tags was injected into a differ-ent channel as a control. Different concentrations of RANKL or its mutants(0, 1.88, 3.75, 7.5, 15, and 30 nM, 30 ml) were injected into both channels.All steps were performed at 25˚C, and signals were recorded as sensor-grams. Sensorgrams were fitted into the 1:1 binding model using BIAevaluation software 4.1 (Biacore, GE Heathcare), and the equilibrium-dissociation constants (Kd) calculated.
Osteoclast formation and tartrate-resistant acid phosphatasestaining
The murine monocytic cell line RAW264.7 (American Type Culture Col-lection, Manassas, VA) was cultured in a humidified incubator (5% CO2 inair) at 37˚C, and maintained in a-MEM containing 10% (v/v) heat-inactivated FCS. For osteoclastogenesis experiments (20), cells were seededinto a 24-well tissue culture plate (23 103/well) in the presence or absenceof 50 ng/ml RANKL or its mutants for 4 d. The cells were then fixed andstained using the Acid Phosphatase, Leukocyte (tartrate-resistant acid phos-phatase [TRAP]) Kit (Sigma-Aldrich, 387A) according to the manufac-turer’s instructions. The numbers of TRAP-positive, multinucleated (.3)cells per well were counted under a light microscope as described (51).
Mouse bone marrow-derived monocytes were isolated from 7-wk-oldBALB/c mice, cultured in a-MEM containing 10% FCS, and plated ina 10-cm petri dish overnight (52). The following day, nonadherent cellswere collected, washed, and seeded into a 24-well tissue culture plate(5 3 105/well) with 20 ng/ml macrophage-CSF (M-CSF) in the presenceor absence of 50 ng/ml RANKL or its mutants. From the fourth day, themedium was changed daily with fresh a-MEM containing 10% FCS,20 ng/ml M-CSF, and 50 ng/ml RANKL or its mutants. Cells were thenfixed on the eighth day and stained using the TRAP staining kit as before.
ResultsStructure determinations
The extracellular domain of RANK has been crystallized bothalone and in complex with RANKL. The crystals of RANK and theRANK–RANKL complex diffracted to 2.0 A and 2.8 A, re-spectively, using synchrotron radiation. The structure of thecomplex was determined first using molecular replacement withthe published structure of RANKL as an initial model (33, 34).The crystallographic asymmetric unit contains one molecule ofRANK (residues 35–199) and one subunit of RANKL (residues161–316), which are assembled to form the biological hetero-hexameric complex through 3-fold crystallographic symmetry.The model has been refined to an R factor of 18.2% (Rfree of21.2%) with root mean square deviations (rmsds) from ideal val-ues of 0.007 A for bond lengths and 1.0˚ for bond angles (Table I).The unliganded structure of RANK was solved using the receptorfrom the complex as the search model and has been refined to an Rfactor of 20.7% (Rfree of 23.7%) with rmsds of 0.007 A for bondlengths and 1.1˚ for bond angles. There are two RANK monomersrelated by noncrystallographic 2-fold symmetry perpendicular tothe long axis of the molecules in the asymmetric unit. The finalmodel consists of residues 33–201 in one monomer and residues36–176 and 186–194 in the second monomer (Fig. 1).
The structure of RANK
The extracellular regions of members of the TNFRSF adopt elon-gated structures of variable numbers of pseudorepeats of CRDs. Atypical CRD, normally ∼40 residues, consists of five irregularb-strands linked typically by three interstrand disulphides and canbe further divided into two structural modules of various types de-fined by topology and number of disulphides (41). RANK contains
Refinement statisticsResolution range (A) 30.0–2.00 30.0–2.80No. of reflections (working/test) 24896/1332 18581/995R factorb (Rwork/Rfree) 0.207/0.237 0.182/0.212No. of atoms (protein/water/others) 2433/221/47 2487/48/5Rms bond length deviation (A) 0.007 0.007Rms bond angle deviation (˚) 1.1 1.0Mean B factorc (A2) 31/35/49 50/41/49
aNumbers in parentheses are for the highest resolution shell.bRwork and Rfree are defined by R = Shkl||Fobs |2 |Fcalc|| / Shkl |Fobs|, where h, k, and l are the indices of the reflections (used in
refinement for Rwork; 5%, not used in refinement for Rfree) and Fobs and Fcalc are the structure factors, deduced from measuredintensities and calculated from the model, respectively.
cMean B factors for protein, water, and others, including ions and glycerol molecules.
four such CRDs spanning a length of 100 A, the longest among thestructures of the TNFR family determined to date. RANK CRD1(residues 35–71) is comprised of so-called A1-B2 modules,whereas CRDs 2–4 (residues 72–114, 115–154, and 155–197, re-spectively) are all made of A1-B1 modules, where A and B definethe module topology and 1 and 2 the number of disulphides (41). Ofthe members of TNFR family with full-length CRD1, crystal struc-tures of TNFRSF1A, OX40, and CrmE (53, 36, 54), either alone orin complex with ligand, have been determined. CRD1 is the moststructurally conserved region among these receptors; .90% of Caatoms can be overlapped with rmsds ranging from 1.0 A to 1.4 Adespite the low sequence identity. Apart from the six conservedcysteines, Tyr41 and Gly54 of RANK are the only fully conservedresidues among the CRD1 domains of these four proteins. Tyr41,positioned in the middle of the second strand of the A1 module,makes hydrophobic interactions with the first disulphide of the B2module as well as a hydrogen bond to the highly conserved Ser67(threonines in the other three structures) from the fifth strand.Ser67, in turn, binds to the carboxyl group of Ser49 positionedbetween the third and fourth cysteines in the third strand, indicatingan important role of Tyr41 for stabilizing the relative position andorientation of the twomodules. In contrast, Gly54 acts to strengthenthe interactions between CRD1 and CRD2. It is the third residue ofa tight turn linking strands 3 and 4 and makes both hydrophobicinteractions with the first disulphide and main-chain hydrogenbonds to the amide group of Cys72 and the carbonyl group ofLeu78 from CRD2. These interactions are conserved between thesefour receptors, and also observed in DR5 (40).The A1-B1 modules of CRDs 2–4 in RANK do not have the
third and fifth cysteines (the 3–5 disulphides); in contrast to theonly previously experimentally observed A1-B1 module of CRD3in OX40 that lacks the 4–6 disulphides (53). The two missingcysteines are substituted by aromatic and glycine residues inCRD2 (His90 and Gly105) and CRD4 (Trp173 and Gly187),replacing the disulphide constraint by ring-stacking hydrophobicand hydrogen bond interactions. As a result, the B1 modules inthese two domains are structurally very similar to the B2 module(Fig. 2). In contrast, the B1 module in CRD3 of RANK is similarto the CRD3 B1 module of CrmE (36), in that the two cysteinesare not replaced by aromatic and glycine residues so that thetopological constraint by the disulphide is not compensated for,and the module adopts a much broader conformation (Fig. 2D). Inaddition, the b2b3 loop (residues 119–132) of CRD3 that makes
the key contacts with the ligand (corresponding to the 90S loop ofDR5 and residues 103–108 of TNFRSF1A) possesses an intra-strand disulphide formed by Cys125 and Cys127 (Fig. 2F). b2(residues 119–122) and b3 (residues127–130) form a regular an-tiparallel b-sheet linked by a four residue turn. Cys125, the thirdresidue of the turn, is so close to Cys127 that the plane of the turnis almost perpendicular to the b-sheet. This unusual conformationis stabilized by hydrogen bonds from the side chain of Asn122 tothe amide groups of residues 124 and 125, and a p-stacking ofTrp121 with one side of the b-sheet (Fig. 2F).Both the number and positions of cysteine residues in the ex-
tracellular regions of mouse and human RANKs are conserved, andit would be expected that all four CRDs in the human molecule arecomprised of the same structural modules as found in mouse. TheCRD3 domain of human RANK has, however, been wronglypredicted to contain A1-B1 modules lacking the 4–6 disulphidesbecause of sequence misalignment (53), highlighting the limita-tions of sequence alignment.There are significant conformational differences and rigid-body
movements apparent when the three independent copies of RANKare compared (two [A and B] from the unliganded crystal asym-metric unit and one [chain R] from the complex). The CXC motiflinking the ligand binding CRD2 and CRD3 in both TNFR1 andDR5 has previously been identified as a hinge region that allowsthe twoCRDs to orientate and position themselves onto the bindingregions of the ligands (40). This motif is also conserved in RANKand the hinge region appears to extend into the C-terminal halfof the CRD2 B1 module (Fig. 2A). The difference in relativeorientation between CRD2 and CRD3 is 20˚ between the twounliganded copies, and these differ by 49˚ and 32˚ from theliganded molecule (Fig. 2A). In contrast, there is little rigid-body movement between CRD1 and CRD2, and between CRD3and CRD4. It is interesting to note that the CRD1-CRD2 andCRD3-CRD4 junctions both have a CXXC motif, one residuelonger than the CRD2-CRD3 linker. The two cysteine residuesat the domain junctions are actually located in the same strand(b5) positioned on one side of the b3b4 loop. The longer CXXClinker enables the b2b3 loop of the second domain to contact theother side of the b3b4 loop and stabilize the two domains. Incontrast, the CRD3 and CRD4 of TNFRSF1A (like CRD2 andCRD3 in RANK) are linked by a CXC motif such that the b2b3loop is unable to closely interact with the b3b4 loop of CRD3(Fig. 2H, 2I).
FIGURE 1. Overall structures of RANK and the RANKL–RANK complex. A, The structure of RANK alone shown as coils with CRDs 1–4 colored red,
green, orange, and cyan, respectively. The disulphides are shown in yellow as ball and sticks. The magenta sphere represents the Na+ bound between CRD3
and CRD4. B, The heterodimer of RANKL–RANK complex in the crystal asymmetric unit with the receptor-free RANKL overlapped. RANK is drawn as
in A, the backbone of RANKL in the complex is shown as blue ribbons and coils with selected side chains shown as cyan sticks. The backbone of receptor-
free RANKL is colored in gray with regions that have large conformational changes because of receptor binding colored in red and side chains as gray
sticks. C, The biological heterohexameric complex as viewed perpendicular to the 3-fold axis.
Superposition of the first two RANKCRDs of chain Awith thoseof chain B results in an rsmd of 1.3 A for 77 equivalent Ca atoms(1.5 A for 75 Ca atoms and 1.2 A for 78 Ca atoms, respectively,when the two chains are compared with the first two domains ofchain R). Large conformational differences occur in the b4b5 loop(residues 60–67) of CRD1 and the hinge region of the B1 module(residues 90–114) of CRD2. Overlaps of CRDs 3–4 of chain R withthe corresponding regions of chain A and B show rmsds of 0.9 A for75 Ca atoms (of 83) and 0.7 A for 69 Ca atoms (of 73), re-spectively. The rigidity between CRD3 and CRD4 is further en-hanced by a metal ion bound between the two domains. This metalion exists in all three copies of RANK and has an octahedral co-ordination provided by the carbonyl oxygen atoms of Cys134,Ala135, and Phe138 from the b3b4 loop of CRD3, the hydroxylgroup of Ser161, and the carbonyl oxygen of Val163 from the b2b3loop of CRD4 and a water molecule. The bond distances range from2.35 to 3.00 A (Fig. 2G). This ion is assigned to be a Na+ based onthe average calcium bond-valence sum of 1.2 calculated from theaveraged bond lengths from the three binding sites (expected val-ues: 1.6 for Na+, K+ 0.6, Ca2+ 2.0, Mn2+ 3.2, and Mg2+ 4.2) (55).This assignment is supported by the anomalous difference mapcalculated using data collected at a wavelength of 1.698 A, whichshows electron density peaks for all ordered sulfur atoms but not forthis ion (f’’: S 0.7e2, Na 0.2e2, K 1.4e2, Mg 0.2e2, Mn 3.5e2, andCa 1.6e2). It is expected that this Na+ is conserved in human RANKbased on the sequence similarity.
The structure of RANKL
Each subunit of the trimeric RANKL adopts a typical b-sandwichwith jellyroll topology consisting of two five-stranded antiparallelb-sheets. The first b-sheet, containing strands A”, A, H, C, and F, is
involved in trimer formation by making intersubunit contacts. Thesecond b-sheet contains strands B’, B, G, D, and E and forms theouter surface. The two b-sheets form a core scaffold that is highlystructurally conserved, whereas the loops linking the b-strands areinvolved in receptor binding and highly variable within the TNFfamily. Superposition of the RANKL monomer from the complexwith the previously published structure of RANKL gives an rmsd of0.5 A for 145 Ca atoms of 156 residues. As expected, conforma-tional changes are only observed in the surface loops of the mole-cule, including the AA”, CD, DE, and EF loops, all of which areinvolved in interactions with the receptor (Fig. 1B). The largestconformational change observed is in the CD loop, where the Caatom and the ring center of Tyr234 are shifted by 3.3 A and 6.5 A,respectively.
The RANKL–RANK complex
RANKLandRANKformaheterohexameric complexwitha receptormolecule bound along each of the three clefts formed by neighboringmonomers of the ligand homotrimer. Of the four CRDs of RANK,only the middle two are involved in direct contacts with the ligand.Superposition of RANKL–RANK with TNF-b–TNFRSF1A andTRAIL–DR5 complexes reveal that, although the RANK CRD2 isbound in a similar orientation as its counterparts in TNFRSF1A andDR5, there is a large difference in the orientation of CRD3, witha tilt of some 45˚ and 11˚ away from the ligand, whereas the positionof RANK as a whole is ∼2 A lower (Fig. 3A).Each of the three interfaces buries 2660 A2 solvent accessible
surface area, 1290 A2 from the ligand, and 1370 A2 from the re-ceptor. Of the surface area buried on RANKL by each receptor, 540A2 is on subunit A and 780 A2 on subunit B, whereas, of the areaburied on the receptor, 840 A2 is from CRD2 and 530 A2 from
FIGURE 2. The structure of RANK. A, A diagram showing the structural differences and highlighting the flexibility of the three copies of RANK. The
superposition was performed using CRD1 and CRD2 only. The RANK molecule in the complex is shown in red and the two copies in the crystal of RANK
are shown in blue and yellow. B, The structure of TNFRSF1A CRD2 comprised of A1-B2 modules (pdb ID, 1ext), a typical CRD with five irregular
b-strands linked by three interstrand disulphides. The side chains of key residues for structure stability are drawn as sticks with the position of each cystine
marked and numbered. The backbone direction is indicated by arrows. C–E, The structures of RANK CRDs 2–4, which are all comprised of A1-B1
modules lacking a 3–5 disulphides. F, Close-up view of the b2b3 loop of CRD3, shows how the loop is stabilized by an intrastrand disulphide and
a network of hydrogen bonds (yellow broken lines). G, The coordination of the Na+ binding site. H and I, Comparison of the relative conformations
between CRD3 and CRD4 in TNFRSF1A (H) and RANK (I). In TNFRSF1A, the CXC linker between the two domains positions, the b2b3 loop of CRD4
on the left side of the b3b4 loop of CRD4, allows little interaction between the two domains, whereas the CXXC linker in RANK enables the b2b3 loop of
CRD4 to be positioned on the right side of the b3b4 loop of CRD3 and makes a number of stabilizing interactions.
CRD3 (Fig. 3B). The areas involved in the interactions in bothreceptor and ligand are significantly different from the two pub-lished structures of TNF family complexes. In the TRAIL–DR5complex, the buried area on TRAIL is contributed not only equallyby the two repeats of the receptor, but also shared equally betweenthe two ligand subunits. In the case of TNF-b–TNFRSF1Acomplex, the CRD2 and CRD3 of TNFRSF1A contribute ina similar ratio to their counterparts of RANK, but they are madeapproximately equally to the two ligand subunits. A unique featureof the RANKL–RANK complex is that the solvent accessible areasburied on the receptor and on subunit B of the ligand arecontinuous, contrasting with the TNF-b–TNFRSF1A andTRAIL–DR5 complexes, where contact regions are discontinuousand form two distinct patches on both ligands and receptors.All solvent accessible loops bridging the b-strands in RANKL,
apart from A”B’, B’B, and BC, are involved in receptor binding.These loops are structurally unique in terms of length and confor-mation and there is little sequence conservation compared withother members of the TNF family. Receptor residues from theb2b3 loop, the b3 strand of CRD2 and the b2b3 loop of CRD3contribute most of the interactions. Residues involved in the li-gand–receptor interactions are mainly hydrophilic in nature (34 of50), forming 9 hydrogen bonds and 12 salt bridges as well as themajority of hydrophobic contacts.The top patch of RANKL subunit A interacts with the receptor
through the AA” and CD loops to the b2b3 loop of CRD3 (Fig.3C). The interactions between the AA” loop and the b2b3 loopare predominantly hydrophilic; Asp124 in the b2b3 loop makesa salt bridge to His179 and a hydrogen bond to the main chain of
Lys180, whose side chain, in turn, forms a salt bridge with Glu126and a hydrogen bond to the carbonyl oxygen of Ser123. Theinteractions from the CD loop are mainly van der Waals contacts;Tyr240 is fully buried, largely by Glu126 of the receptor. Tyr234undergoes the largest conformational change on receptor binding,making side chain stacking interactions with Arg129, which, inturn, is stabilized by a salt bridge with Glu268 from the EF loop ofsubunit B. The unique intrastrand disulphide Cys125-Cys127 ofthe receptor is solvent inaccessible, contacting both the CD loopof subunit A and the EF loop of subunit B. The AA” loops ofTNFSF and the b2b3 loops of TNFRSF members are both struc-turally the most divergent, having the greatest number of aminoacid deletions and insertions. The AA” loop in RANKL foldstoward the top third of the molecule and is positioned above theb2b3 loop of the receptor. In contrast, the much longer AA” loopof TRAIL runs across the middle surface of the ligand and liesbelow the b2b3 loop to make a salt bridge from Arg149 to Glu147of DR5, whereas the same loop in TNF-b is very short and doesnot make any interaction with TNFRSF1A (38). Deletion mutationof the AA” loop in both RANKL and TRAIL completely abolishesbiological activity (33, 40). The structural diversity between themembers of the TNF family, charged interactions, andmutagenesis all suggest that the AA” loop confers specificity.The contact area on the lower part of RANKL subunit A is
mediated by residues Ile248 andHis252 of theDE loop toGlu84 andLeu88 of the b2b3 loop of CRD2. Ile248 corresponds to Tyr108 inTNF-b and Tyr214 in TRAIL and has been predicted to makestrong hydrophobic interactions with the receptor analogous tothose in the TNF-b–TNFRSF1A and TRAIL–DR5 complexes
FIGURE 3. The RANKL–RANK complex. A, A ribbon and coil diagram showing the positions and orientations of receptors relative to the ligands in
RANKL–RANK (green), TNF-b–TNFRSF1A (red), and TRAIL-DR5 (blue) complexes by superimposing the ligands. B, Open book view of RANKL–
RANK interface. Residues in the interface are colored by the percentage of surface area buried on complex formation (1–20%, blue; 20–40%, cyan; 40–
60%, green; 60–80%, orange; 80–100%, red). C–E, Close up of three key interface areas with residues in RANKL labeled in red and those in RANK
labeled in black; (C) residues Asp124 and Glu126 from the b2b3 loop of the receptor form salt bridges with residues His179 and Lys180 from the AA” loop
of the ligand. D, Hydrophobic interactions centered at Leu89 of the receptor. E, Interface area centered at His224 of the ligand, where the His224 is fully
buried, Glu225 and Glu268 make two salt bridges to Arg129 and Arg130 of the receptor. On receptor binding the ring center of Tyr234 from the ligand has
moved 6.5 A to make stacking interactions with Arg129 (from red to blue).
(38–40). Tyr108 in TNF-b is fully buried in a hydrophobicdepression formed by the b2b3 loop consisting of residues 60–71of TNFRSF1A, and Tyr214 in TRAIL makes similar interactionswith the b2b3 loop (residues 50–61, the 50S loop) of DR5; the onlyconserved ligand–receptor interactions in the two complexes. Theimportance of this tyrosine has been shown by mutagenesis inTNF-a, TNF-b, FasL, and TRAIL that abolishes receptor binding(39, 40, 56–58). An Ile248 Asp mutation in mouse RANKL,however, showed only an 8-fold decrease in activity (33). Ile248has direct contact with a charged residue (Glu84) of RANK, at anequivalent position to Leu67 of TNFRSF1A and Leu57 of DR5, andthe reduction in activity is likely due to the introduction of anelectrostatic repulsive force. The DE loop is one of the regions thathas the highest B factors and the side chains of Lys247 and Ile248do not have well-defined electron density. It is likely that inRANKL, unlike other members of the TNFSF, the DE loop is notcritical for receptor binding. Arg283 of the FG loop in RANKLforms a salt bridge with Asp85 that is at an equivalent position toArg68 of TNFRSF1A and Leu58 of DR5 (Fig. 4). The FG loop isnot involved in receptor binding in either TNF-b or TRAIL.
There are two key interface areas between subunit B and thereceptor. One, at the lower part of the interface, is mediated by theb1b2 loop and b3 of CRD2 of the receptor that interacts with theGH loop and the N and C termini of the AA” loop from the ligand.The interaction is centered on Leu89 that nests in a hydrophobicpocket formed by Tyr187, Arg190, and Gln302 of RANKL (Fig.3D). This interface area, together with the two salt bridges formedfrom Asp94 and Lys97 of b3b4 loop at the C-terminal part ofCRD2 to residues Arg222 and Asp229, appears to be the deter-minant for the position and orientation of CRD2. The second keyinterface area is centered on His224 of the CD loop that is locatedin a pocket formed between the two ligand-binding repeats of thereceptor (Fig. 3E). Residues lining the pocket include Ala98 fromthe b3b4 loop of CRD2, Tyr119 from the b1b2 loop, the firstdisulphide Cys115-Cys128, and Cys128 of CRD3. The hydropho-bic interactions centered on His224 are sandwiched by two clus-ters of charged interactions: the two salt bridges mentionedpreviously, and two additional salt bridges made from Glu225 ofthe CD loop and Glu268 of the EF loop to Arg129 and Arg130 ofthe b3 strand of CRD3. The CD and AA” loops, located oppositeeach other across the receptor binding cleft, act as two anchorpoints for the CRD3 of the receptor. The interactions centeredon His224 are not observed in the TNF-b–TNFRSF1A or in theTRAIL–DR5 complexes, because of the different orientations ofthe receptors.
High-affinity binding is critical for functional osteoclast for-mation
The affinity of RANKL and RANK was measured by Biacore withRANK immobilized in a channel of a chelating NTA sensor chipand RANKL as the mobile phase. Recombinant TNFRSF9, asa nonspecific protein control, was immobilized in a differentchannel in the same chip. As seen in Fig. 5A, the affinity betweenRANK and RANKL is very high with a Kd of 6.8 3 10211 M.To elucidate the critical residues responsible for this very tight
binding, and the contribution of the binding affinity to functional
osteoclast formation, the following RANKL mutants were madeaccording to the buried area on complex formation (Fig. 3B):Asn266Ala, 1–20%; Glu225Ala, 40–60%; Arg222Ala, 60–80%,and Asp299Ala 80–100%. All four of these residues interact withthe receptor via either hydrogen bonds or salt bridges. The bindingaffinities of RANKL mutants for immobilized RANK were mea-sured as before. The binding affinities of Glu225Ala, Arg222Ala,and Asp299Ala RANKL mutants for RANK are dramaticallydecreased by .100-fold (Fig. 5A) and these mutants have com-pletely lost their ability to promote functional osteoclast formation(Fig. 5B). In contrast, amino acid Asn266 (marked in blue on Fig.3B) contributes moderately to binding with ,20% buried area. Itsmutant Asn266Ala only marginally affects its binding to RANK(Kd 8.83 10211 M compared with a Kd of 6.83 10211 M for wildtype). Interestingly, this slightly reduced affinity between RANKLand RANK significantly affects the ability of RANKL to promoteosteoclast formation (Fig. 5B, 5C), demonstrating that a strongassociation between RANK and RANKL is prerequisite for properRANK signaling and subsequent osteoclast formation.
DiscussionMembers of the TNFSF adopt the same trimeric structural scaffoldwith each receptor-binding cleft formed between two neighboringligand subunits.Receptor binding ismediatedpredominantly by sur-face loops with little sequence homology andmuch structural diver-gence between family members. The multidomain TNFRs possesstwo ligand-binding CRDs with a similar overall fold (but differingdisulphide topology) and a flexible CXC domain junction. Thus,for a given TNF ligand-receptor pair, the structural diversity ofthe ligand surface loops is coupled to structural variations and do-mainflexibility of the receptor, leading to a distinct, specific, bindingmode. It is therefore unsurprising that ligand–receptor interactionsin the RANKL–RANK complex are significantly different to thestructurally known complexes of TNF-b–TNFRSF1A and TRAIL-DR5.The majority of residues involved in complex formation are hy-
drophilic in nature, achieving both surface and electrostatic com-plementary. Of the three key interface areas identified in theRANKL–RANK complex (the AA” loopmediated interactions withthe b2b3 loop of CRD3, the area centered on His224 of the ligand,and the hydrophobic contacts centered on Leu89 of the receptor)none is conserved in the other two complexes, giving an indicationof the complexity of ligand–receptor binding in the superfamily. Ourobservations are in agreement with the notion that the interactions ofthe b2b3 loop of CRD3 with the ligand may have an important rolein controlling the specificity and cross-reactivity among the super-family members, but do not support the proposal that the hydropho-bic interaction between the DE loop of the ligand and the b2b3 loopof CRD2 of the receptor, as observed in the complexes of TNF-b–TNFRSF1A and TRAIL-DR5, is a general feature important forbinding in the superfamily (39).The decoy receptor OPG is a soluble protein containing four N-
terminal CRDs, followed by two death domains and a C-terminalbasic domain. It has been shown that in vivo the protein exists intwo states: as a homodimer cross-linked via C-terminal cysteines oras a C-terminal truncated monomer, both of which appear to have
similar specific activities in the inhibition of osteoclastogenesis(59). However, in a more recent report it has been shown thatthe dimerization of OPG is a result of noncovalent interactionsmediated by the two death domains, and the dimer binds RANKLwith an affinity of three orders of magnitude tighter than themonomer lacking the death domains. One dimer interacts withone RANKL trimer by occupying two of the three binding siteson the ligand (60). Nevertheless, aligning the OPG sequence withour structure of RANKL–RANK complex suggests that OPGwould bind RANKL via its CRD2-3 in a similar mode to RANK(Fig. 4). The CRD2 of OPG has the same disulphide connectivityas the first three CRDs of TNFRSF1A, comprising A1-B2 struc-tural modules; whereas OPG CRD3 is made of A1-B1 modules.The Asn131 and Leu144 at the third and fifth cystine positions ofRANK CRD3 are however substituted by a histidine and a glycinein OPG; replacing the disulphide constraint by stacking and hy-drogen bond interactions. The CRD3 of OPG is thus expected tobe structurally similar to the CRD2 of RANK (Fig. 2C). Most ofthe key structural features observed in the RANKL–RANK com-plex are expected to be conserved in the RANKL–OPG complex,despite the two receptors having only 30% sequence identity. Theb2b3 loop of CRD3 in OPG is two residues shorter and does notcontain the intrastrand disulphide and so would adopt a moreextended conformation to make similar interactions to the AA”loop of the ligand as observed for RANK (Fig. 3C). A tyrosineresidue in OPG that substitutes Leu89 could fill up the extra spaceof the leucine binding pocket and make ring-stacking interactionswith Tyr187 of RANKL (Fig. 3D).
The last ordered residue at the C terminus of RANK in thecomplex is Met199, 15 residues away from the start of thetransmembrane helix. Thus, the apex of RANKL is probably morethan 50 A from the target cell surface, in contrast to the ∼6 A and25 A for TRAIL and TNF-b. The distance between the C terminiof two neighboring RANK molecules is 76 A, much larger thanthe ∼50 A spacing of the DR5 transmembrane helices and theintracellular receptor binding sites on the trimeric TRAFmolecules, which initiate downstream signaling (61, 62). Agreater separation between the RANK transmembrane helices isto be expected because the RANK intracellular domain is muchlarger than that of DR5 and the spacing of the receptor bindingsites on the trimeric TRAF molecules is similar. It has beensuggested that extracellular ligand-receptor binding may triggerthe signal cascade by dictating a precise positioning of the threetransmembrane helices (40). This is in agreement with the obser-vation of the rigid domain connection between CRD3 and CRD4of RANK. The sodium ion bound between these two CRDs mayplay an important role in the RANKL–RANK signaling by main-taining the structural integrity of these two domains.Bone remodeling is a dynamically equilibrated process regulated
by the RANKL/RANK/OPG system. Perturbation of the processby mutations in genes of the molecular system results in variousbone diseases. The structure of the RANKL–RANK complex isessential for our understanding of the structural mechanism ofthese disease related mutations. Four autosomal-recessive osteo-petrosis-related mutations in the extracellular region of RANKhave been reported recently: Gly53Arg, Arg129Cys, Arg170Gly,
FIGURE 5. The affinity of RANKL–RANK interaction and the key residues involved in binding and osteoclast formation. A, The affinities of RANKL
and its mutants for binding to RANK were measured by Biacore analyses using a chelating NTA sensor chip. Recombinant TNFRSF9 was used as
a negative control. Different concentrations of RANKL or its mutants (0, 1.88, 3.75, 7.5, 15, and 30 nM) were injected into channels and signals recorded as
sensorgrams. The equilibrium-dissociation constants were calculated using BIA evaluation software 4.1. B, TRAP staining of bone marrow-derived
monocytes cells. Bone marrow-derived cells were cultured with M-CSF and in the presence or absence of 25 ng/ml RANKL or its mutants for 8 d and
TRAP stained. Scale bar, 200 mm. C, RAW264.7 cells were cultured in the presence or absence of 50 ng/ml RANKL or its mutants for 4 d. The numbers of
matured osteoclastic cells (TRAP-positive, multinucleated) per well were scored and expressed as means 6 SD of four wells. The experiments were
and Cys175Arg (27), which correspond to residues of Gly54,Arg130, Lys171, and Cys176, respectively, in the mouse RANK.Lys171 forms a salt bridge with the conserved Asp162, which liesbetween the Na+ binding residues Ser161 and Val163 (Fig. 2G). Amutation to glycine is likely to disturb the Na+ binding. Cys176forms the 4–6 disulphide of CRD4 with Cys195; removing thedisulphide constraint by an arginine mutation will result in con-formational flexibility in the second structural module of CRD4,especially the b5 strand leading to the transmembrane helix.Gly54 is highly conserved in the TNFRSF, and its structural rolehas been discussed earlier. The Gly54Arg mutation is likely tocause conformational changes of the b1b2 loop of CRD2 anddisrupt the interactions between Leu89 and the ligand. Arg130is directly involved in ligand binding, forming a salt bridge withGlu225 and a hydrogen bond to Asn266. An Arg130Cys mutationwould abolish these interactions.Three RANKL mutations, an N-terminal deletion of residues
145–177, a C-terminal truncation starting at residue 277, anda Met199Lys substitution, have been identified in patients withautosomal-recessive osteopetrosis (28). The deletion of 145–177results in loss of bA and half of the AA” loop. The loop interactswith the CRD3 b2b3 loop and is crucial for receptor binding (33).The C-terminal deletion mutation causes loss of two central strandsof the jellyroll scaffold, likely resulting in an unfolded ligand. Theside chain of Met199 nests in a hydrophobic core between the twob-sheets and has direct contact with the backbone of Phe165 thatstacks against Phe213 and Phe280 from a neighboring subunit.Mutation of Met199 to a charged residue in a hydrophobic envi-ronment is expected to cause local conformational changes anddisturb the trimer interface.In summary, we have elucidated at atomic resolution the
structures of the extracellular domain of mouse RANK and of theRANK–RANKL complex. Our data show that although the com-plex between RANK and its cognate ligand RANKL is similar inoverall architecture to that observed for other members of theTNFSF, there are significant differences in the position and ori-entation of the receptor and, notably, in the conformation of thebound RANK. This leads to each interaction surface in RANKL–RANK being continuous, whereas for the other examples, theinteraction consists of two distinct patches. Mutations of individualresidues of RANKL involved in receptor binding demonstrate theirfunctional significance in terms of osteoclastogenesis. The struc-tural information obtained additionally helps to explain someforms of human osteopetrosis linked to mutations in the RANKand RANKL genes.
AcknowledgmentsWe thank the staff at the United Kingdom beamline, BM14, and at ID14-
EH4 and ID23-EH2 ESRF, Grenoble, France, for help with data collections
and Prof. Fremont for GST–RANKL construct.
DisclosuresThe authors have no financial conflicts of interest.
References1. Theill, L. E., W. J. Boyle, and J. M. Penninger. 2002. RANK-L and RANK:
T cells, bone loss, and mammalian evolution. Annu. Rev. Immunol. 20: 795–823.2. Walsh, M. C., N. Kim, Y. Kadono, J. Rho, S. Y. Lee, J. Lorenzo, and Y. Choi.
2006. Osteoimmunology: interplay between the immune system and bone me-tabolism. Annu. Rev. Immunol. 24: 33–63.
3. Binder, N. B., B. Niederreiter, O. Hoffmann, R. Stange, T. Pap, T. M. Stulnig,M. Mack, R. G. Erben, J. S. Smolen, and K. Redlich. 2009. Estrogen-dependentand C-C chemokine receptor-2-dependent pathways determine osteoclast behav-ior in osteoporosis. Nat. Med. 15: 417–424.
4. Fata, J. E., Y. Y. Kong, J. Li, T. Sasaki, J. Irie-Sasaki, R. A. Moorehead,R. Elliott, S. Scully, E. B. Voura, D. L. Lacey, et al. 2000. The osteoclast
differentiation factor osteoprotegerin-ligand is essential for mammary gland de-velopment. Cell 103: 41–50.
5. Loser, K., A. Mehling, S. Loeser, J. Apelt, A. Kuhn, S. Grabbe, T. Schwarz,J. M. Penninger, and S. Beissert. 2006. Epidermal RANKL controls regulatoryT-cell numbers via activation of dendritic cells. Nat. Med. 12: 1372–1379.
6. Hanada, R., A. Leibbrandt, T. Hanada, S. Kitaoka, T. Furuyashiki, H. Fujihara,J. Trichereau, M. Paolino, F. Qadri, R. Plehm, et al. 2009. Central control offever and female body temperature by RANKL/RANK. Nature 462: 505–509.
7. Anderson, D. M., E. Maraskovsky, W. L. Billingsley, W. C. Dougall,M. E. Tometsko, E. R. Roux, M. C. Teepe, R. F. DuBose, D. Cosman, andL. Galibert. 1997. A homologue of the TNF receptor and its ligand enhanceT-cell growth and dendritic-cell function. Nature 390: 175–179.
8. Marsters, S. A., A. D. Frutkin, N. J. Simpson, B. M. Fendly, and A. Ashkenazi.1992. Identification of cysteine-rich domains of the type 1 tumor necrosis factorreceptor involved in ligand binding. J. Biol. Chem. 267: 5747–5750.
9. Boyle, W. J., W. S. Simonet, and D. L. Lacey. 2003. Osteoclast differentiationand activation. Nature 423: 337–342.
10. Chung, J. Y., Y. C. Park, H. Ye, and H. Wu. 2002. All TRAFs are not createdequal: common and distinct molecular mechanisms of TRAF-mediated signaltransduction. J. Cell Sci. 115: 679–688.
11. Feng, X. 2005. Regulatory roles and molecular signaling of TNF family mem-bers in osteoclasts. Gene 350: 1–13.
12. Liu, W., D. Xu, H. Yang, H. Xu, Z. Shi, X. Cao, S. Takeshita, J. Liu, M. Teale,and X. Feng. 2004. Functional identification of three receptor activator of NF-kappa B cytoplasmic motifs mediating osteoclast differentiation and function.J. Biol. Chem. 279: 54759–54769.
13. Takayanagi, H., S. Kim, K. Matsuo, H. Suzuki, T. Suzuki, K. Sato, T. Yokochi,H. Oda, K. Nakamura, N. Ida, et al. 2002. RANKL maintains bone homeostasisthrough c-Fos-dependent induction of interferon-beta. Nature 416: 744–749.
14. Xing, L., E. M. Schwarz, and B. F. Boyce. 2005. Osteoclast precursors, RANKL/RANK, and immunology. Immunol. Rev. 208: 19–29.
15. Wong, B. R., J. Rho, J. Arron, E. Robinson, J. Orlinick, M. Chao, S. Kalachikov,E. Cayani, F. S. Bartlett, III, W. N. Frankel, et al. 1997. TRANCE is a novelligand of the tumor necrosis factor receptor family that activates c-Jun N-terminal kinase in T cells. J. Biol. Chem. 272: 25190–25194.
16. Lacey, D. L., E. Timms, H. L. Tan, M. J. Kelley, C. R. Dunstan, T. Burgess,R. Elliott, A. Colombero, G. Elliott, S. Scully, et al. 1998. Osteoprotegerin ligandis a cytokine that regulates osteoclast differentiation and activation. Cell 93:165–176.
17. Yasuda, H., N. Shima, N. Nakagawa, K. Yamaguchi, M. Kinosaki, S. Mochizuki,A. Tomoyasu,K.Yano,M.Goto,A.Murakami, et al. 1998. Osteoclast differentiationfactor is a ligand for osteoprotegerin/osteoclastogenesis-inhibitory factor and is iden-tical to TRANCE/RANKL. Proc. Natl. Acad. Sci. USA 95: 3597–3602.
18. Dougall, W. C., and M. Chaisson. 2006. The RANK/RANKL/OPG triad incancer-induced bone diseases. Cancer Metastasis Rev. 25: 541–549.
19. Burgess, T. L., Y. Qian, S. Kaufman, B. D. Ring, G. Van, C. Capparelli,M. Kelley, H. Hsu, W. J. Boyle, C. R. Dunstan, et al. 1999. The ligand forosteoprotegerin (OPGL) directly activates mature osteoclasts. J. Cell Biol. 145:527–538.
20. Hsu,H.,D.L.Lacey,C.R.Dunstan, I. Solovyev,A.Colombero,E.Timms,H.L.Tan,G. Elliott, M. J. Kelley, I. Sarosi, et al. 1999. Tumor necrosis factor receptor familymember RANK mediates osteoclast differentiation and activation induced byosteoprotegerin ligand. Proc. Natl. Acad. Sci. USA 96: 3540–3545.
21. Nakagawa, N., M. Kinosaki, K. Yamaguchi, N. Shima, H. Yasuda, K. Yano,T. Morinaga, and K. Higashio. 1998. RANK is the essential signaling receptorfor osteoclast differentiation factor in osteoclastogenesis. Biochem. Biophys. Res.Commun. 253: 395–400.
22. Simonet, W. S., D. L. Lacey, C. R. Dunstan, M. Kelley, M. S. Chang, R. Luthy,H. Q. Nguyen, S. Wooden, L. Bennett, T. Boone, et al. 1997. Osteoprotegerin: a novelsecreted protein involved in the regulation of bone density. Cell 89: 309–319.
23. Bucay, N., I. Sarosi, C. R. Dunstan, S. Morony, J. Tarpley, C. Capparelli,S. Scully, H. L. Tan, W. Xu, D. L. Lacey, et al. 1998. osteoprotegerin-deficientmice develop early onset osteoporosis and arterial calcification. Genes Dev. 12:1260–1268.
24. Kong, Y. Y., H. Yoshida, I. Sarosi, H. L. Tan, E. Timms, C. Capparelli,S. Morony, A. J. Oliveira-dos-Santos, G. Van, A. Itie, et al. 1999. OPGL is a keyregulator of osteoclastogenesis, lymphocyte development and lymph-node or-ganogenesis. Nature 397: 315–323.
25. Dougall, W. C., M. Glaccum, K. Charrier, K. Rohrbach, K. Brasel, T. De Smedt,E. Daro, J. Smith, M. E. Tometsko, C. R. Maliszewski, et al. 1999. RANK isessential for osteoclast and lymph node development. Genes Dev. 13: 2412–2424.
26. Li, J., I. Sarosi, X. Q. Yan, S. Morony, C. Capparelli, H. L. Tan, S. McCabe,R. Elliott, S. Scully, G. Van, et al. 2000. RANK is the intrinsic hematopoietic cellsurface receptor that controls osteoclastogenesis and regulation of bone mass andcalcium metabolism. Proc. Natl. Acad. Sci. USA 97: 1566–1571.
27. Guerrini, M. M., C. Sobacchi, B. Cassani, M. Abinun, S. S. Kilic, A. Pangrazio,D. Moratto, E. Mazzolari, J. Clayton-Smith, P. Orchard, et al. 2008. Humanosteoclast-poor osteopetrosis with hypogammaglobulinemia due to TNFRSF11A(RANK) mutations. Am. J. Hum. Genet. 83: 64–76.
28. Sobacchi, C., A. Frattini, M. M. Guerrini, M. Abinun, A. Pangrazio, L. Susani,R. Bredius, G. Mancini, A. Cant, N. Bishop, et al. 2007. Osteoclast-poor humanosteopetrosis due to mutations in the gene encoding RANKL. Nat. Genet. 39:960–962.
29. Villa, A., M. M. Guerrini, B. Cassani, A. Pangrazio, and C. Sobacchi. 2009.Infantile malignant, autosomal recessive osteopetrosis: the rich and the poor.Calcif. Tissue Int. 84: 1–12.
30. Cohen, S. B., R. K. Dore, N. E. Lane, P. A. Ory, C. G. Peterfy, J. T. Sharp, D. vander Heijde, L. Zhou, W. Tsuji, and R. Newmark Denosumab Rheumatoid Ar-thritis Study Group. 2008. Denosumab treatment effects on structural damage,bone mineral density, and bone turnover in rheumatoid arthritis: a twelve-month,multicenter, randomized, double-blind, placebo-controlled, phase II clinical trial.Arthritis Rheum. 58: 1299–1309.
31. Cummings, S. R., J. San Martin, M. R. McClung, E. S. Siris, R. Eastell,I. R. Reid, P. Delmas, H. B. Zoog, M. Austin, A. Wang, et al; FREEDOM Trial.2009. Denosumab for prevention of fractures in postmenopausal women withosteoporosis. N. Engl. J. Med. 361: 756–765.
32. Jones, E. Y., D. I. Stuart, and N. P. Walker. 1989. Structure of tumour necrosisfactor. Nature 338: 225–228.
33. Lam, J., C. A. Nelson, F. P. Ross, S. L. Teitelbaum, and D. H. Fremont. 2001.Crystal structure of the TRANCE/RANKL cytokine reveals determinants ofreceptor-ligand specificity. J. Clin. Invest. 108: 971–979.
34. Ito, S., K. Wakabayashi, O. Ubukata, S. Hayashi, F. Okada, and T. Hata. 2002.Crystal structure of the extracellular domain of mouse RANK ligand at2.2-A resolution. J. Biol. Chem. 277: 6631–6636.
35. Karpusas, M., Y. M. Hsu, J. H. Wang, J. Thompson, S. Lederman, L. Chess, andD. Thomas. 1995. 2 A crystal structure of an extracellular fragment of humanCD40 ligand. Structure 3: 1426.
36. Graham, S. C., M. W. Bahar, N. G. Abrescia, G. L. Smith, D. I. Stuart, andJ. M. Grimes. 2007. Structure of CrmE, a virus-encoded tumour necrosis factorreceptor. J. Mol. Biol. 372: 660–671.
37. Naismith, J. H., T. Q. Devine, T. Kohno, and S. R. Sprang. 1996. Structures ofthe extracellular domain of the type I tumor necrosis factor receptor. Structure 4:1251–1262.
38. Banner, D. W., A. D’Arcy, W. Janes, R. Gentz, H. J. Schoenfeld, C. Broger,H. Loetscher, and W. Lesslauer. 1993. Crystal structure of the soluble human 55kd TNF receptor-human TNF beta complex: implications for TNF receptoractivation. Cell 73: 431–445.
39. Hymowitz, S. G., H. W. Christinger, G. Fuh, M. Ultsch, M. O’Connell, R. F. Kelley,A. Ashkenazi, and A. M. de Vos. 1999. Triggering cell death: the crystal structure ofApo2L/TRAIL in a complex with death receptor 5.Mol. Cell 4: 563–571.
40. Mongkolsapaya, J., J. M. Grimes, N. Chen, X. N. Xu, D. I. Stuart, E. Y. Jones, andG. R. Screaton. 1999. Structure of the TRAIL-DR5 complex reveals mechanismsconferring specificity in apoptotic initiation. Nat. Struct. Biol. 6: 1048–1053.
41. Naismith, J. H., and S. R. Sprang. 1998. Modularity in the TNF-receptor family.Trends Biochem. Sci. 23: 74–79.
42. Walter, T. S., J. M. Diprose, C. J. Mayo, C. Siebold, M. G. Pickford, L. Carter,G. C. Sutton, N. S. Berrow, J. Brown, I. M. Berry, et al. 2005. A procedure forsetting up high-throughput nanolitre crystallization experiments. Crystallizationworkflow for initial screening, automated storage, imaging and optimization.Acta Crystallogr. D Biol. Crystallogr. 61: 651–657.
43. Walter, T. S., C. Liu, P. Huang, S. Zhang, L. R. Wedderburn, B. Gao,R. J. Owens, D. I. Stuart, P. Tang, and J. Ren. 2009. Crystallization andpreliminary X-ray analysis of mouse RANK and its complex with RANKL. ActaCrystallogr. Sect. F Struct. Biol. Cryst. Commun. 65: 597–600.
44. Otwinowski, Z, W. Minor. 1997. Processing of x-ray diffraction data collected inoscillation mode. Methods Enzymol. 276: 307–326.
45. Vagin, A., and A. Teplyakov. 1997. MOLREP: an Automated Program forMolecular Replacement. J. Appl. Cryst. 30: 1022–1025.
46. CCP4. 1994. The CCP4 suite: programs for protein crystallography. ActaCrystallogr. Sect. C, Crystal structure communications.
47. McCoy, A. J., R. W. Grosse-Kunstleve, P. D. Adams, M. D. Winn, L. C. Storoni,and R. J. Read. 2007. Phaser crystallographic software. J. Appl. Cryst. 40: 658–674.
48. Brunger, A. T. 2007. Version 1.2 of the Crystallography and NMR system. Nat.Protoc. 2: 2728–2733.
49. Emsley, P., and K. Cowtan. 2004. Coot: model-building tools for moleculargraphics. Acta Crystallogr. D Biol. Crystallogr. 60: 2126–2132.
50. Nieba, L., S. E. Nieba-Axmann, A. Persson, M. Hamalainen, F. Edebratt,A. Hansson, J. Lidholm, K. Magnusson, A. F. Karlsson, and A. Pluckthun.1997. BIACORE analysis of histidine-tagged proteins using a chelating NTAsensor chip. Anal. Biochem. 252: 217–228.
51. Polek, T. C., M. Talpaz, B. G. Darnay, and T. Spivak-Kroizman. 2003. TWEAKmediates signal transduction and differentiation of RAW264.7 cells in theabsence of Fn14/TweakR. Evidence for a second TWEAK receptor. J. Biol.Chem. 278: 32317–32323.
52. Lamothe, B., A. Besse, A. D. Campos, W. K. Webster, H. Wu, and B. G. Darnay.2007. Site-specific Lys-63-linked tumor necrosis factor receptor-associated fac-tor 6 auto-ubiquitination is a critical determinant of I kappa B kinase activation.J. Biol. Chem. 282: 4102–4112.
53. Compaan, D. M., and S. G. Hymowitz. 2006. The crystal structure of the cos-timulatory OX40-OX40L complex. Structure 14: 1321–1330.
54. Naismith, J. H., T. Q. Devine, B. J. Brandhuber, and S. R. Sprang. 1995.Crystallographic evidence for dimerization of unliganded tumor necrosis factorreceptor. J. Biol. Chem. 270: 13303–13307.
55. Muller, P., S. Kopke, and G. M. Sheldrick. 2003. Is the bond-valence methodable to identify metal atoms in protein structures? Acta Crystallogr. D Biol.Crystallogr. 59: 32–37.
56. Goh, C. R., C. S. Loh, and A. G. Porter. 1991. Aspartic acid 50 and tyrosine 108are essential for receptor binding and cytotoxic activity of tumour necrosis factorbeta (lymphotoxin). Protein Eng. 4: 785–791.
57. Schneider, P., J. L. Bodmer, N. Holler, C. Mattmann, P. Scuderi, A. Terskikh,M. C. Peitsch, and J. Tschopp. 1997. Characterization of Fas (Apo-1, CD95)-Fasligand interaction. J. Biol. Chem. 272: 18827–18833.
58. Yamagishi, J., H. Kawashima, N. Matsuo, M. Ohue, M. Yamayoshi, T. Fukui,H. Kotani, R. Furuta, K. Nakano, and M. Yamada. 1990. Mutational analysis ofstructure—activity relationships in human tumor necrosis factor-alpha. ProteinEng. 3: 713–719.
59. Tomoyasu, A., M. Goto, N. Fujise, S. Mochizuki, H. Yasuda, T. Morinaga,E. Tsuda, and K. Higashio. 1998. Characterization of monomeric and homodi-meric forms of osteoclastogenesis inhibitory factor. Biochem. Biophys. Res.Commun. 245: 382–387.
60. Schneeweis, L. A., D. Willard, and M. E. Milla. 2005. Functional dissection ofosteoprotegerin and its interaction with receptor activator of NF-kappaB ligand.J. Biol. Chem. 280: 41155–41164.
61. Park, Y. C., V. Burkitt, A. R. Villa, L. Tong, and H. Wu. 1999. Structural basis forself-association and receptor recognition of human TRAF2. Nature 398: 533–538.
62. Ye, H., J. R. Arron, B. Lamothe, M. Cirilli, T. Kobayashi, N. K. Shevde,D. Segal, O. K. Dzivenu, M. Vologodskaia, M. Yim, et al. 2002. Distinct mo-lecular mechanism for initiating TRAF6 signalling. Nature 418: 443–447.
![RANKL/RANK/OPG system beyond bone remodeling: …Intriguingly, RANKL/RANK axis is also required for hormone-driven mammary gland development during pregnancy [21]. Given the proliferative](https://static.documents.pub/doc/80x56/6096c353589f381291245d5f/ranklrankopg-system-beyond-bone-remodeling-intriguingly-ranklrank-axis-is-also.jpg)
