Crystal Structure of Human RANKL Complexed with Its Decoy Receptor

This information is current as Xudong Luan, Qingyu Lu, Yinan Jiang, Senyan Zhang, Qing of September 26, 2021. Wang, Huihui Yuan, Wenming Zhao, Jiawei Wang and Xinquan Wang J Immunol 2012; 189:245-252; Prepublished online 4 June 2012;

doi: 10.4049/jimmunol.1103387 Downloaded from http://www.jimmunol.org/content/189/1/245

Supplementary http://www.jimmunol.org/content/suppl/2012/06/04/jimmunol.110338 Material 7.DC1 http://www.jimmunol.org/ References This article cites 45 articles, 13 of which you can access for free at: http://www.jimmunol.org/content/189/1/245.full#ref-list-1

Why The JI? Submit online.

• Rapid Reviews! 30 days* from submission to initial decision by guest on September 26, 2021

• No Triage! Every submission reviewed by practicing scientists

• Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2012 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

Crystal Structure of Human RANKL Complexed with Its Decoy Receptor Osteoprotegerin

Xudong Luan,*,1 Qingyu Lu,*,1 Yinan Jiang,* Senyan Zhang,* Qing Wang,* Huihui Yuan,† Wenming Zhao,† Jiawei Wang,* and Xinquan Wang*

Receptor activator of NF-kB (RANKL), its signaling receptor RANK, and its decoy receptor osteoprotegerin (OPG) constitute a molecular triad that is critical in regulating remodeling, and also plays multiple roles in the . OPG binds RANKL directly to block its interaction with RANK. In this article, we report the 2.7-A˚ crystal structure of human RANKL trimer in complex with the N-terminal fragment of human OPG containing four cysteine-rich TNFR homologous domains (OPG-CRD). The structure shows that RANKL trimer uses three equivalent grooves between two neighboring monomers to interact with three OPG-CRD monomers symmetrically. A loop from the CRD3 domain of OPG-CRD inserts into the shallow groove of RANKL, providing the major binding determinant that is further confirmed by affinity measurement and Downloaded from differentiation assay. These results, together with a previously reported mouse RANKL/RANK complex structure, reveal that OPG exerts its decoy receptor function by directly blocking the accessibilities of important interacting residues of RANKL for RANK recognition. Structural comparison with / complex also reveals structural basis for the cross-reactivity of OPG to TRAIL. The Journal of Immunology, 2012, 189: 245–252. http://www.jimmunol.org/ one remodeling is a dynamic and balanced progress that creted by inhibits osteoclastogenesis by binding involves bone matrix formation mediated by osteoblasts RANKL to prevent RANKL/RANK interaction. In the bone mi- B and mediated by . Osteoblasts croenvironment, the local ratio of RANKL and OPG determines are cells that arise from stem cells, whereas osteo- the level of osteoclast activation. The normal osteoclast activation clasts are multinucleated cells derived from hematopoietic pre- results in physiological bone resorption in , cursors of the monocyte/ lineage. Receptor activator whereas overactivated osteoclasts result in various bone diseases of NF-kB ligand (RANKL), its signaling receptor RANK, and its such as and (8, 9). Besides the decoy receptor osteoprotegerin (OPG) play key roles in regulating best-known critical roles in bone remodeling, RANKL also plays bone remodeling (1–3). Experiments using knockout mice have multiple roles in immune system (10), develop- by guest on September 26, 2021 unanimously established their pivotal roles as central regulators of ment (11), bone metastasis (12), hormone-derived breast osteoclast function (4–7). Specifically, the interaction of RANKL cancer development (13), and thermal regulation (14). Therefore, expressed by osteoblasts with RANK expressed on osteoclast the RANKL-RANK-OPG molecular triad is an attractive target precursors is essential for the differentiation, activation, and sur- to develop rational therapy for many osteopenic conditions and vival of osteoclasts. The decoy receptor OPG expressed and se- prevent bone destruction in osteoporosis and arthritis (15). RANKL is a member in the TNF superfamily (TNFSF), con- sisting of a membrane-anchor domain, a connecting stalk, and *Center for Structural Biology, School of Life Sciences, Ministry of Education Key Laboratory of Science, Tsinghua University, Beijing 100084, People’s Re- a receptor-binding ectodomain. RANKL also exists in a soluble public of China; and †Department of Immunology, Capital Medical University, Bei- form from proteolytic cleavage of the transmembrane form or jing 100069, People’s Republic of China alternative mRNA splicing (16, 17). The ectodomain of RANKL 1 X.L. and Q.L. contributed equally to this work. adopts the characteristic jellyroll b-sandwichfoldinTNFSF Received for publication November 28, 2011. Accepted for publication March 30, and assembles into a homotrimer in either membrane- 2012. bound or soluble form (18, 19). The receptor RANK belongs to This work was supported by the Ministry of Science and Technology (2010CB912402, 2011CB910502, and 2010CB529106), the National Natural Science Foundation of the TNFR superfamily (TNFRSF), consisting of four extracellular China (30970573 and 31070667), and the Fok Ying Tung Education Foundation. cysteine-rich TNFR homologous domains (CRD1-CRD4), a trans- The coordinates and structure factors presented in this article for the human RANKL/ membrane helix, and a cytoplasmic region. RANKL/RANK in- OPG-CRD complex have been deposited in the (http://www.rcsb. teraction induces RANK receptor trimerization, triggering the org/pdb/) under access code 3URF for immediate release on publication. association of RANK cytoplasmic region with TNFR-associated Address correspondence and reprint request to Dr. Xinquan Wang, Medical Science Building C226, Tsinghua University, Beijing 100084, People’s Republic of China. factors and subsequent multiple signaling pathways (20, 21). The E-mail address: [email protected] decoy receptor OPG represents an atypical member in the TNFRSF The online version of this article contains supplemental material. because it lacks the transmembrane region and is secreted as a Abbreviations used in this article: CRD, cysteine-rich TNFR homologous domain; soluble protein. The OPG has four CRD domains in the N-ter- DR, death receptor; OPG, osteoprotegerin; OPG-CRD, N-terminal four CRDs of minal region that are necessary and sufficient for the inhibition of human OPG; RANK, receptor activator of NF-kB; RANKL, receptor activator of NF-kB ligand; SPR, surface plasmon resonance; TNFSF, TNF superfamily; TNFRSF, osteoclastogenesis (7, 22), and its C-terminal region contains two TNFR superfamily; TRAP, tartrate-resistant acid phosphatase; vWF, von Willebrand death domain homologous regions, a heparin-binding domain, and factor. a cysteine residue at the C-terminal end necessary for dimer for- Copyright Ó 2012 by The American Association of Immunologists, Inc. 0022-1767/12/$16.00 mation (7, 22, 23). www.jimmunol.org/cgi/doi/10.4049/jimmunol.1103387 246 COMPLEX STRUCTURE OF RANKL WITH OPG N-TERMINAL CRDs

The crystal structure of mouse RANKL was determined 10 years SCALA from the CCP4 suite (29). The final statistics of data processing ago (18, 19), and residues in RANKL that are important for are listed in Table I. RANK recognition have also been proposed based on structural Structural determination and refinement modeling and mutational studies (18, 19, 24, 25). The crystal structure of the mouse RANKL/RANK complex was recently The position of human RANKL monomer in the asymmetric unit was solved independently by two groups, providing precise molecular determined by using molecular replacement program PHASER (30) with mouse RANKL monomer (Protein Data Bank code: 1IQA) as search details of RANKL/RANK interaction (26, 27). In this article, we model. Starting with initial phases provided by RANKL, the OPG-CRD report the complex structure of human RANKL ectodomain with model was built gradually by cycles of iterative manual model building N-terminal four CRD domains of human OPG (OPG-CRD). Key with program COOT (31) and structural refinement with program PHENIX residues in OPG-CRD for RANKL binding were identified by (32) refinement module. In the final model, 80.4, 16.4, 2.1, and 1.1% of the amino acids reside in the most favorable, additional allowed, generously structure-based mutagenesis of OPG-CRD and surface plasmon allowed, and disallowed regions of the Ramachandran plots, respectively. resonance (SPR) analyses, which was further collaborated by The final refinement statistics are listed in Table I. osteoclast differentiation assay. These results, together with pre- vious reported mouse RANKL/RANK complex structure, provide SPR experiment a complete picture of ligand/receptor interactions in the RANKL- The binding affinity between RANKL and OPG-CRD and its mutants RANK-OPG molecular triad. was determined by SPR using BIAcore 3000 at 25˚C. The RANKL was immobilized to ∼50 response units on a research-grade CM5 sensor chip in 10 mM sodium acetate, pH 5.5 by standard amine coupling method. The Materials and Methods flow cell 1 was left blank as a reference. OPG-CRD in 10 mM HEPES, pH Constructs 7.2, 150 mM NaCl, and 0.005% Tween 20 was injected over the flow cells Downloaded from at different concentrations at a flow rate of 30 ml min21 to measure binding The coding sequence of human RANKL ectodomain (residues 162–317) affinity of OPG-CRD or its mutants by RANKL. The RANKL/OPG-CRD was cloned into a modified pET30a vector (Novagen) with NusA/His tag complexes were allowed to associate for 90 s and dissociate for 150 s. The followed by a tobacco etch virus protease cleavage site. Constructs surfaces were regenerated with 5 mM NaOH between each cycle if needed. expressing GST-tagged wild type human RANKL and its mutants for GST Data were analyzed with BIAcore 3000 evaluation software BIAevaluation pull-down experiments were made by using vector pGEX-6P-1 vector (GE 4.1 by globally fitting to a 1:1 Langmuir binding model. Healthcare). The coding sequences of human OPG-CRD containing four http://www.jimmunol.org/ CRD domains (residues 1–165) and human RANK (residues 1–166) were Osteoclast differentiation assay cloned into pFastBac (Invitrogen) vector with C-terminal 63 His tag. The residues in OPG-CRD and RANK ectodomain were numbered without The biological activities of RANKL and OPG were estimated using counting the cleaved signal peptide in the mature receptors. Mutations on an osteoclast differentiation assay. Mouse osteoclast precursors were ex- OPG-CRD were made by overlap extension PCR using synthetic oligo- tracted from the tibia and femur of 7-wk-old male mice, cultured in RPMI nucleotide primers. 1640 containing 10% FBS and 30 ng ml21 M-CSF and 100 ng ml21 RANKL. Then osteoclast precursors were seeded in 96-well plates (2 3 Protein expression and purification 104 cells/well) and incubated at 37˚C in a humidified atmosphere con- taining 5% CO2. Three days later, the medium was changed with fresh The OPG-CRD and RANK ectodomain were expressed and purified from 2 2 RPMI 1640 containing 10% FBS and 30 ng ml 1 M-CSF, and 100 ng ml 1 baculovirus-infected insect cells. In brief, the pFastBac vector containing

RANKL. The inhibitory effects of OPG-CRD and its mutants were tested by guest on September 26, 2021 of interest was transformed into DH10bac competent cells, and high 2 by adding them into wells with a final concentration of 500 ng ml 1. m.w. mini-prep DNA was prepared from selected Escherichia coli clones Differentiation of osteoclast precursors was measured in vitro by containing the recombinant bacmid. This DNA was then used to transfect quantitation of tartrate-resistant acid phosphatase (TRAP) activity. In the Sf9 insect cells. After two cycles of amplification in Sf9 cells, the re- seventh day, cells were fixed and stained using the TRAP staining (Sigma combinant virus was used to infect Hi5 cells at a density of 2 3 106 cells/ Aldrich no. 387), and the reaction product was quantified by spectroscopic ml in HyQ SFX medium (HyClone). The cell culture medium was col- measurement at 405 nm. TRAP-positive multinucleated cells were evalu- lected after 48 h, and the secreted protein was purified using Ni-NTA resin ated under the microscope, and each experiment was repeated three times (Qiagen) followed by size exclusion chromatography (Superdex 200 10/ for statistical analysis. 300; GE Healthcare). The OPG-CRD mutants were expressed and purified in a similar way. The RANKL was overexpressed in E. coli strain BL21 (DE3) at 16˚C for 20 h with 0.5 mM isopropyl b-D-1-thiogalactopyr- Results anoside induction. The cells were collected and lysed in a buffer con- taining 50 mM Tris-HCl, 150 mM NaCl, pH 8.0. The fusion protein NusA- Overall structure His-RANKL in the soluble fraction was purified using Ni-NTA resin We have determined the structure of the human RANKL in com- (Qiagen), followed by the cleavage with tobacco etch virus protease to plex with human OPG-CRD at a resolution of 2.7 A˚ using a com- remove NusA and His tag. The released RANKL was further purified by size exclusion chromatography (Superdex 200 10/300; GE Healthcare). bination of molecular replacement method with mouse RANKL GST-tagged wild type RANKL and its mutants were also expressed in E. as search model with subsequent manual model building of the coli strain BL21(DE3) and purified by glutathione-Sepharose beads (GE OPG-CRD region (Table I, Supplemental Fig. 1). We also tried the Healthcare) and size exclusion chromatography (Superdex 200 10/300; molecular replacement method to search the position of OPG- GE Healthcare). The RANKL/OPG-CRD complex was reconstituted by mixing OPG- CRD with the available TNF family receptor structures includ- CRD and RANKL in ∼1:1 molar ratio and incubated at 4˚C for 10 h. ing TNFR1, TNFR2, and others. However, the extensive search The complex was further purified by size exclusion chromatography did not result in a clear solution, reflecting structural flexibility (Superdex 200 10/300; GE Healthcare) and ion exchange chromatography within each CRD domain and domain flexibility between the CRD (Resource S; GE Healthcare). domains in the TNFRSF. The final model in the asymmetric unit Crystallization and x-ray data collection consists of one RANKL monomer (Ala162-Asp317) and one human OPG-CRD monomer (Pro5-Ser165), with additional three histidine Crystals were grown using the hanging-drop vapor-diffusion method by 166 168 mixing RANKL/OPG-CRD complex (10 mg ml–1 in 10 mM HEPES so- residues (His -His ) at the C-terminal end of OPG-CRD (Fig. dium, pH 7.2 and 150 mM NaCl) with an equal volume of reservoir so- 1A). We were also able to build a glycan attached to residue lution containing 100 mM sodium phosphate monobasic monohydrate, 100 Asn157 of OPG-CRD (Fig. 1A). In the RANKL/OPG-CRD com- mM potassium phosphate monobasic, 100 mM MES monohydrate pH 6.5, plex structure, functional RANKL trimer bound with three OPG- and 2.0 M sodium chloride. Crystals appeared in 1 wk at 18˚C. X-ray diffraction data were collected at Shanghai Synchrotron Research Facil- CRD monomers can be generated by rotating the monomeric ity beamline BL17U at a wavelength of 0.97 A˚ . Data images were indexed RANKL/OPG-CRD along the 3-fold rotational axis of the sym- and integrated using program MOSFLM (28) and scaled using program metry (space group P4132) (Fig. 1B, 1C). The Journal of Immunology 247

Table I. Data collection and refinement statistics

Data Collection RANKL/OPG-CRD

Space group P4132 Cell dimensions a, b, c (A˚ ) 144.80, 144.80, 144.80 a, b, g (˚) 90.0, 90.0, 90.0 Resolution (A˚ )a 40.0 – 2.70 (2.84 – 2.70) Rmerge 0.111 (0.690) I / sI 11.8 (3.2) Completeness (%) 99.9 (100.0) Redundancy 9.2 (9.5) Refinement Resolution (A˚ )b 40.0 – 2.70 (2.91 – 2.70) Reflections 14,783 (2738) Rwork/Rfree 0.202/0.251 No. of atoms Protein 2555 Glycan 14 Water 36 B-factors Protein 79.4 Downloaded from Glycan 124.3 Water 51.6 Root mean square deviations Bond lengths (A˚ ) 0.009 Bond angles (˚) 1.23 Values in parentheses are for highest resolution shell. a Statistics from data processing. http://www.jimmunol.org/ bStatistics from structure refinement.

FIGURE 2. Structure of OPG-CRD and its comparison with RANK. (A) The OPG-CRD fragment, corresponding to the extracellular four OPG-CRD is composed of four CRDs (CRD1-CRD4), similar to the ar- CRD domains of RANK (Fig. 2A), has been shown to be necessary chitecture of RANK. (B) The rigid-body movements of CRD1 and CRD4 and sufficient for the inhibition of osteoclastogenesis (7, 22). The between RANK (blue) and OPG-CRD (magenta) after structural super- imposition based on CRD2 and CRD3. CRD1-CRD4 domains of OPG-CRD adopt the so-called A1-B1, A1-B2, A1-B1, and A1-B1 module, respectively (Fig. 2A), where A and B denote the type of topology and 1 and 2 the number of interstrand disulfide bond within the module (33). The CRD1- by guest on September 26, 2021 CRD4 domains of RANK adopt the A1-B2, A1-B1, A1-B1, and A1-B1 module, respectively (Fig. 2A). Besides the difference of interstrand disulfide bond distribution in CRD1 and CRD2 domains, an additional intrastrand disulfide bond (Cys125-Cys127) in the CRD3 domain of RANK is also absent in that of OPG-CRD. Structural alignment of each CRD domain between OPG-CRD and RANK results in a root mean square deviation of 1.16, 1.36, 1.49, and 1.69 A˚ for CRD1 (33 Ca atoms), CRD2 (41 Ca atoms), CRD3 (37 Ca atoms), and CRD4 (45 Ca atoms), re- spectively. There are apparent rigid-body movements in CRD1 and CRD4 after overall alignment based on CRD2 and CRD3 re- sponsible for ligand binding (Fig. 2B). Similar rigid-body move- ments were also observed among different states of RANK in- cluding unbound and bound with RANKL (27). The RANKL monomer in the complex is composed of 10 b-strands, forming the outer b-sheet (B’, B, G, D, and E) and inner b-sheet (A’, A, H, C, and F) (Fig. 1A). The most significant conformation changes of RANKL on OPG-CRD binding are at AA’ and CD loops that are involved in the direct interaction, revealed by structural compar- ison of RANKL in unbound and OPG-CRD bound states (Sup- plemental Fig. 2). In our protein purification experiment, the OPG-CRD contain- FIGURE 1. Overall structure of RANKL/OPG-CRD complex. (A) ing the N-terminal four CRD domains was purified as a monomer. Ribbon diagram of one RANKL (green) with one OPG-CRD in the The complex structure also shows that trimeric RANKL interacts asymmetric unit (magenta). The disulfide bonds (magenta) and N-linked with monomeric OPG-CRD to form a stable 1:3 trimer/monomer glycan (cyan) in OPG-CRD are shown in sticks. The b-strands in RANKL are labeled according to standard TNF-b–sandwich nomenclature. (B) Side complex. In contrast, the full-length OPG is a disulfide-bond linker view of the RANKL trimer (green) bound with three OPG-CRD (magenta) dimer, and it was reported that trimeric RANKL interacts with monomers. (C) View down the 3-fold axis of the RANKL/OPG-CRD dimeric OPG to mainly form a stable 1:1 trimer/dimer complex complex, perpendicular to (B). All figures were made with the PyMol in solution (34). We will further discuss the binding stoichiometry program (http://www.PyMol.org). between RANKL and OPG later. 248 COMPLEX STRUCTURE OF RANKL WITH OPG N-TERMINAL CRDs

Interactions between RANKL and OPG-CRD (Fig. 4C, 4D). These data indicate that site II plays a more critical In the RANKL/OPG-CRD complex, the trimeric RANKL uses three role in the binding than site I. spatially distinct but equivalent shallow grooves to bind three OPG- Comparison with RANKL/RANK complex CRD molecules (Fig. 1B, 1C). One groove, formed by two neigh- boring RANKL monomers, accommodates one OPG-CRD mono- The overall architectures of the human RANKL/OPG-CRD and mer and the buried solvent-accessible surface area is 2200 A˚ 2 at the mouse RANKL/RANK complexes are similar with a trimeric interface. The binding interface is composed of two binding sites RANKL bound with three monomeric OPG-CRD or RANK ˚ 2 (Fig. 3A, 3B). At site I, the OPG “50s loop” (His47-Leu65)from receptors. The buried surface-accessible area is 2660 A at the ˚ 2 CRD2 domain is placed nearly parallel to the RANKL along the RANKL/RANK interface, similar to the value (2200 A ) at the groove (Fig. 3A), whereas at site II, the OPG “90s loop” (Arg90 RANKL/OPG interface. Both ligand/receptor interfaces are also -Leu98) from CRD3 domain inserts deeply into the groove and is composed of two binding sites (I and II; Fig. 5A). The DE, FG, clamped by loops AA’ and CD from neighboring RANKL mono- and GH loops of RANKL involved in site I interaction are well mers (Fig. 3A). Site I consists of relative small and separate contact conserved in RANKL/RANK and RANKL/OPG-CRD complexes, patches in both RANKL and OPG-CRD (Fig. 3B). Residues in- and the CRD2 domain is also similar in its orientation at the volved in the site I interaction are from loops AA’ (Arg191,Gly192), binding interface (Fig. 5A). We have shown that the OPG-CRD 57 58 DE (Lys248,His253), FG (Lys282,Arg284), and GH (Pro301,Asp302, Asp Ala/Glu Ala mutant retained the nearly same RANKL Gln303) of RANKL and the 50s loop (His47,Tyr48,Tyr49,Ser56, binding affinity and inhibition of osteoclast differentiation as wild Asp57,Glu58,Tyr61,Pro64,Val65,Leu69) of OPG CRD2 domain type OPG-CRD (Fig. 4C, 4D). Similarly, it has also been shown

85 Downloaded from (Supplemental Table I). Site II consists of continuous patches in both that mutation of RANK residue Asp at site I did not significantly RANKL and OPG-CRD (Fig. 3B), including residues from loops affect the RANKL binding and inhibition of osteoclast differen- AA’ (Ser179,His180,Lys181), strands C (Arg223,His224,His225), D tiation by soluble RANK (26). At site II, the general binding mode (Gln237,Met239,Tyr241), E (Lys257), and F (Phe270,His271,Phe272)of of an inserted loop from receptor into RANKL trimer is also RANKL and the 90s loop (Arg90,Leu92,Glu93,Ile94,Glu95,Phe96, conserved in the RANKL/OPG-CRD and RANKL/RANK com- Cys97,Leu8,Glu116) of OPG CRD3 domain (Supplemental Table I). plexes. There are significant conformation changes at the central Detailed inspection revealed obvious hydrophilic interactions part of the long AA’ and CD loops in RANKL (Fig. 5A), and the http://www.jimmunol.org/ including ionic and hydrogen bonds in both sites (Fig. 4A, 4B). 90s loop of CRD3 also has positional shift resulting from the Compared with separate interactions at site I (Fig. 4A), the more change of relative orientation between CRD3 and CRD2 in two interactions at site II are focused around the tip the OPG 90s loop different complexes (Fig. 5B). Therefore, the detailed interactions and form a network (Fig. 4B). Residues involved in the interaction at site II are different in these two complexes, especially at the network are Arg223, Gln237,Tyr241, and Lys257 of RANKL and central part of the inserted 90s loop of OPG-CRD and loop3 of 93 94 95 96 Ile94, Glu95, and Phe96 of the OPG-CRD 90s loop. RANKL resi- RANK, whose sequences are –Glu 2Ile 2Glu 2Phe 2 and – 124 125 126 127 dues Arg223,Tyr241, and Lys257 all interact with central Glu95 Glu 2Cys 2Glu 2Cys 2, respectively. In the human residue in the 90s loop of OPG-CRD (Fig. 4B). RANKL/OPG-CRD complex, the Glu95 from the OPG-CRD has by guest on September 26, 2021 To confirm the roles of sites I and II in the binding of RANKL interactions with RANKL residues Arg223,Tyr241, and Lys257, by OPG-CRD, residues in OPG-CRD involved in the binding forming a hydrophilic interaction network (Fig. 5B). In the mouse through ionic and hydrogen bonds (Asp57 and Glu58 in site I and RANKL/RANK complex, Glu126 from loop3 of RANK keeps the Glu95 in site II) were mutated to Ala. The SPR analyses showed interaction with RANKL Tyr240 with additional interaction with that wild type OPG-CRD bound RANKL with affinity of 0.65 nM RANKL Lys180 and has no direct interactions with RANKL (Fig. 4C). The binding affinity of RANKL by OPG-CRD Glu95Ala Arg222 and Lys256 (Fig. 5B). There is a close to 180˚ side-chain mutant was significantly reduced to 2.4 mM. Consistently, this rotation of mouse RANKL Arg222 compared with human RANKL OPG-CRD mutant lost most of the capacity to inhibit RANKL- Arg223 (Fig. 5C), which enables the interaction with RANK Asp94. induced osteoclast differentiation compared with wild type OPG- Mouse RANKL Lys256 has interaction with RANK Lys97 at the CRD (Fig. 4D). In contrast, the OPG-CRD Asp57Ala/Glu58Ala interface. Mutation of either Glu126 or Lys97 at RANK completely double-mutant bound RANKL with an affinity of 0.69 nM and abrogated the RANKL binding and inhibition of osteoclast dif- still retained the same level of inhibition as wild type OPG-CRD ferentiation by soluble RANK (26). We also performed a GST

FIGURE 3. Overall view of RANKL/OPG-CRD binding interface. (A) Sites I and II in the binding interface. Two neighboring RANKL monomers are shown in surface, colored with green and cyan, re- spectively. The OPG-CRD is shown in ribbon col- ored with magenta. The “50s loop” in site I and “90s loop” of OPG-CRD in site II are colored with or- ange. (B) An open-book view of the contact residues in binding sites I and II. The two neighboring RANKL monomers and OPG-CRD are presented in surface with gray color. The contact residues in binding sites I and II are colored with red and blue, respectively. The Journal of Immunology 249 Downloaded from http://www.jimmunol.org/ by guest on September 26, 2021

FIGURE 4. Specific interactions at sites I and II, RANKL binding affinity measurement, and inhibition of osteoclast differentiation by OPG-CRD and its mutants. (A) Interactions at site I. RANKL residues of two monomers are colored with green and cyan, respectively, and OPG-CRD residues are colored with magenta. The salt-bridge and hydrogen-bonding interactions are represented by black dashed lines. (B) Interactions around the tip of the inserted OPG- CRD loop in site II. (C) Binding affinity measurement of RANKL by OPG-CRD and its mutants with SPR method. (D) Inhibitory activities of OPG-CRD and its mutants measured with TRAP assay. pull-down experiment using human GST-RANKL and its mutants Structural basis for the cross-reactivity of OPG with human OPG-CRD or human RANK. The mutation of residue Besides binding RANK, OPG is able to interact with TRAIL (35– 223 241 257 Arg ,Tyr , and Lys reduced the binding of human RANKL 37), which is also a member in the TNFSF and induces by both human OPG-CRD and human RANK (Fig. 5C). All these of tumor cells through the cell-surface receptors structural and biochemical data suggest that site II is the major (DR4) and DR5 in the TNFRSF family (38). It has been shown binding determinant of RANKL by both OPG-CRD and RANK, that OPG inhibits the TRAIL-induced cell apoptosis and TRAIL although there are some subtle differences in the specific inter- inhibits OPG-mediated inhibition of osteoclastogenesis (35, 37). actions. The RANKL/OPG-CRD complex is similar to TRAIL/DR5 250 COMPLEX STRUCTURE OF RANKL WITH OPG N-TERMINAL CRDs

FIGURE 5. Structural comparison of RANKL/ OPG-CRD with RANKL/RANK. (A) Sites I and II in both complexes. (B) Side-by-side comparison of interactions around the tip of inserted loop at site II. (C) GST pull-down experiments showing the effects of mutagenesis of RANKL residues at site II on the binding of OPG-CRD (left Coomassie-stained SDS- PAGE gel) and RANK (right gel). The upper bands

in both gels are GST-tagged RANKL wild type and Downloaded from mutants. The middle bands in both gels are GST that is still coeluted with target protein after gluta- thione beads and size exclusion column purifica- tions. The control experiment using GST instead of GST-tagged RANKL in the binding shows that GST does not bind to OPG-CRD and RANK. The bottom bands are OPG-CRD (left gel) and RANK (right http://www.jimmunol.org/ gel) that are retained by RANKL. by guest on September 26, 2021

complex (39–41) in both ligand/receptor docking mode and the in plasma (43, 44). The complex of vWF with factor VIII has organization of ligand/receptor binding sites (Supplemental Fig. also been shown to interact with OPG through the vWF and 3A). The TRAIL/DR5 complex also has two major binding sites abolish the inhibitory effect of OPG on TRAIL-induced apoptosis in the interface, corresponding to the sites I and II in the RANKL/ (45). The modeling indicated that the contact surface of OPG for OPG-CRD complex (Supplemental Fig. 3A). The site I in TRAIL/ vWF is composed of residues Cys41 to Tyr48 and Tyr61 to Glu68 (45), DR5 complex is clustered around residue Tyr216 that interacts with which has partial overlap with binding site I at the RANKL/OPG residues in a hydrophobic groove in the DR5 CRD2 domain (40). interface. This tyrosine residue is critical for the bioactivity of TRAIL and its interaction with DR5 (42), and is also strictly conserved in Discussion some TNFSF members including TNF, LT, FasL, and Apo2L (40). Full-length OPG is a disulfide-bond–linked homodimer, and each In RANKL, this corresponding position is replaced by residue monomer has N-terminal four CRD domains followed by two Ile249 (Supplemental Fig. 3B). However, in the RANKL/OPG- death domain homologous regions, one heparin-binding domain, CRD site I, residue Ile249 does not have close interactions with and a cysteine residue at the C-terminal end forming interchain OPG-CRD, neither with RANK in the site I of RANKL/RANK disulfide bond. The RANKL/OPG-CRD structure we report in this complex. The four major RANKL residues Arg223, Gln237,Tyr241, article shows that the N-terminal four CRD domains behave as an and Lys257 for OPG-CRD interaction at site II are strictly con- independent structural module, and CRD2 and CRD3 contribute served in both primary and three-dimensional structure in TRAIL to the interaction with RANKL. Although the C-terminal region (Supplemental Fig. 3B, 3C). TRAIL residues in Arg223,Tyr241, of OPG does not participate in the direction interaction with and Lys257 positions also have interactions with Glu96 in the loop RANKL, it has been reported that the C-terminal death domain of DR5 CRD3 domain, corresponding to Glu95 in the OPG 90s homologous regions are essential for OPG dimer formation (34). loop (Supplemental Fig. 3C). The highly conserved contact resi- The native dimeric OPG has stronger avidity for RANKL than dues and specific interactions observed in site II provide structural monomer form, providing the structural basis for its ability to basis for the cross-reactivity of OPG to TRAIL. effectively compete with signaling receptor RANK for ligand It has been reported that OPG is physically associated with von binding (34). This also raises the question about the binding Willebrand factor (vWF), a glycoprotein involved in primary he- stoichiometry of trimeric RANKL with full-length dimeric OPG. mostasis, within the Weibel–Palade bodies of endothelial cells and Previous ultracentrifuge analyses showed that trimeric RANKL The Journal of Immunology 251 interacts with dimeric OPG to mainly form a stable 1:1 trimer/ a novel secreted protein involved in the regulation of bone density. Cell 89: 309– 319. dimer complex in solution (34). The two OPG monomers per 8. Anandarajah, A. P. 2009. Role of RANKL in bone diseases. Trends Endocrinol. dimer both bind to the same RANKL trimer, and each monomer of Metab. 20: 88–94. OPG cannot bind a separate RANKL trimer (34). Based on these 9. Kearns, A. E., S. Khosla, and P. J. Kostenuik. 2008. Receptor activator of nuclear factor kappaB ligand and osteoprotegerin regulation of bone remodeling in biochemical data and the RANKL/OPG-CRD complex structure, health and disease. Endocr. Rev. 29: 155–192. we suggest that the shape of OPG dimer would be like a “Y” 10. Takayanagi, H. 2007. : shared mechanisms and crosstalk be- shape. The bottom part is the dimeric C-terminal region of OPG tween the immune and bone systems. Nat. Rev. Immunol. 7: 292–304. 11. Fata, J. E., Y. Y. Kong, J. Li, T. Sasaki, J. Irie-Sasaki, R. A. Moorehead, including death domain homologous regions, heparin-binding R. Elliott, S. Scully, E. B. Voura, D. L. Lacey, et al. 2000. The osteoclast dif- domain, and the C-terminal disulfide bond, and the top part is ferentiation factor osteoprotegerin-ligand is essential for mammary gland de- two separate N-terminal regions including four CRD domains. In velopment. Cell 103: 41–50. 12. Jones, D. H., T. Nakashima, O. H. Sanchez, I. Kozieradzki, S. V. Komarova, the 1:1 trimer/dimer RANKL/OPG complex, the two separate N- I. Sarosi, S. Morony, E. Rubin, R. Sarao, C. V. Hojilla, et al. 2006. Regulation of terminal arms of OPG occupy two of three grooves in RANKL cancer cell migration and bone metastasis by RANKL. Nature 440: 692–696. 13. Schramek, D., A. Leibbrandt, V. Sigl, L. Kenner, J. A. Pospisilik, H. J. Lee, trimer. This would allow efficient inhibition of RANKL-induced R. Hanada, P. A. Joshi, A. Aliprantis, L. Glimcher, et al. 2010. Osteoclast dif- RANK trimerization, which is essential for subsequent signaling ferentiation factor RANKL controls development of progestin-driven mammary and osteoclast maturation. cancer. Nature 468: 98–102. 14. Hanada, R., A. Leibbrandt, T. Hanada, S. Kitaoka, T. Furuyashiki, H. Fujihara, The human RANKL/OPG-CRD and mouse RANKL/RANK J. Trichereau, M. Paolino, F. Qadri, R. Plehm, et al. 2009. Central control of structural comparison showed that OPG exerts its decoy recep- fever and female body temperature by RANKL/RANK. Nature 462: 505–509. tor function by directly occupying the major binding determinant 15. Tanaka, S. 2007. Signaling axis in osteoclast biology and therapeutic targeting in the RANKL/RANK/OPG system. Am. J. Nephrol. 27: 466–478. site II and blocking the accessibilities of key interacting residues 16. Ikeda, T., M. Kasai, M. Utsuyama, and K. Hirokawa. 2001. Determination of Downloaded from such as Arg223,Tyr241, and Lys257 in human RANKL for RANK three isoforms of the receptor activator of nuclear factor-kappaB ligand and their recognition. This information could help the development of differential expression in bone and . Endocrinology 142: 1419–1426. 17. Lum, L., B. R. Wong, R. Josien, J. D. Becherer, H. Erdjument-Bromage, therapeutic agents targeting the RANKL/RANKL/OPG molecular J. Schlo¨ndorff, P. Tempst, Y. Choi, and C. P. Blobel. 1999. Evidence for a role of triad for the treatment of bone-related diseases (46, 47). Current a -alpha (TNF-alpha)-converting enzyme-like protease in strategies under investigation include the usage of OPG-Fc, shedding of TRANCE, a TNF family member involved in osteoclastogenesis and survival. J. Biol. Chem. 274: 13613–13618.

RANK-Fc, anti-RANKL Ab, OPG peptidomimetics, RANKL 18. Ito, S., K. Wakabayashi, O. Ubukata, S. Hayashi, F. Okada, and T. Hata. 2002. http://www.jimmunol.org/ vaccination, and modulators of intracellular signaling pathways Crystal structure of the extracellular domain of mouse RANK ligand at 2.2-A resolution. J. Biol. Chem. 277: 6631–6636. (47). Designing small molecules targeting protein–protein inter- 19. Lam, J., C. A. Nelson, F. P. Ross, S. L. Teitelbaum, and D. H. Fremont. 2001. face is an enormous challenge in drug development, but the ex- Crystal structure of the TRANCE/RANKL reveals determinants of istence of a small subset of residues that contribute to most of the receptor-ligand specificity. J. Clin. Invest. 108: 971–979. 20. Leibbrandt, A., and J. M. Penninger. 2008. RANK/RANKL: regulators of im- free energy of binding (“hot spots”) in the interface still makes it mune responses and bone physiology. Ann. N. Y. Acad. Sci. 1143: 123–150. possible (48). The existence of critical RANKL residues such as 21. Tanaka, S., I. Nakamura, J. Inoue, H. Oda, and K. Nakamura. 2003. Signal Arg223,Tyr241, and Lys257 at site II for RANK interaction suggests transduction pathways regulating osteoclast differentiation and function. J. Bone Miner. Metab. 21: 123–133. that designing small molecules that could target these “hot-spot” 22. Yamaguchi, K., M. Kinosaki, M. Goto, F. Kobayashi, E. Tsuda, T. Morinaga, and residues in RANKL is still a potential way in targeting the K. Higashio. 1998. Characterization of structural domains of human osteoclas- by guest on September 26, 2021 RANKL-RANK-OPG molecular triad. togenesis inhibitory factor. J. Biol. Chem. 273: 5117–5123. 23. Yasuda, H., N. Shima, N. Nakagawa, S. I. Mochizuki, K. Yano, N. Fujise, Y. Sato, M. Goto, K. Yamaguchi, M. Kuriyama, et al. 1998. Identity of osteo- Acknowledgments clastogenesis inhibitory factor (OCIF) and osteoprotegerin (OPG): a mechanism by which OPG/OCIF inhibits osteoclastogenesis in vitro. Endocrinology 139: The diffraction data were collected at Shanghai Synchrotron Research 1329–1337. Facility Beamline BL-17U. We thank J.H. He, S. Huang, and L. Tang 24. Cheng, T., N. J. Pavlos, C. Wang, J. W. Tan, J. M. Lin, J. Cornish, M. H. Zheng, at Shanghai Synchrotron Research Facility for onsite assistance. We also and J. Xu. 2009. Mutations within the TNF-like core domain of RANKL impair thank D.W. Chen, M.H. Zhang, X. Chen, and H. Qi for providing mouse osteoclast differentiation and activation. Mol. Endocrinol. 23: 35–46. 25. Ito, S., and T. Hata. 2004. Crystal structure of RANK ligand involved in bone osteoclast precursor cells. metabolism. Vitam. Horm. 67: 19–33. 26. Ta, H. M., G. T. Nguyen, H. M. Jin, J. Choi, H. , N. Kim, H. Y. Hwang, and Disclosures K. K. Kim. 2010. Structure-based development of a receptor activator of nuclear factor-kappaB ligand (RANKL) inhibitor peptide and molecular basis for The authors have no financial conflicts of interest. . Proc. Natl. Acad. Sci. USA 107: 20281–20286. 27. Liu, C., T. S. Walter, P. Huang, S. Zhang, X. Zhu, Y. Wu, L. R. Wedderburn, P. Tang, R. J. Owens, D. I. Stuart, et al. 2010. Structural and functional insights References of RANKL-RANK interaction and signaling. J. Immunol. 184: 6910–6919. 1. Boyce, B. F., and L. Xing. 2008. Functions of RANKL/RANK/OPG in bone 28. Leslie, A. 1992. Recent changes to the MOSFLM package for processing film and image plate data. Joint CCP4 + ESF-EAMCB Newsletter on Protein Crys- modeling and remodeling. Arch. Biochem. Biophys. 473: 139–146. 2. Theill, L. E., W. J. Boyle, and J. M. Penninger. 2002. RANK-L and RANK: tallography 26. 29. Collaborative Computational Project, Number 4. 1994. The CCP4 suite: pro- T cells, bone loss, and mammalian evolution. Annu. Rev. Immunol. 20: 795–823. grams for protein crystallography. Acta. Crystallogr. D Biol. Crystallogr. 50: 3. Theoleyre, S., Y. Wittrant, S. K. Tat, Y. Fortun, F. Redini, and D. Heymann. 760–763. 2004. The molecular triad OPG/RANK/RANKL: involvement in the orchestra- 30. McCoy, A. J., R. W. Grosse-Kunstleve, P. D. Adams, M. D. Winn, L. C. Storoni, tion of pathophysiological bone remodeling. Cytokine Rev. 15: and R. J. Read. 2007. Phaser crystallographic software. J. Appl. Cryst. 40: 658– 457–475. 674. 4. Dougall, W. C., M. Glaccum, K. Charrier, K. Rohrbach, K. Brasel, T. De Smedt, 31. Emsley, P., and K. Cowtan. 2004. Coot: model-building tools for molecular E. Daro, J. Smith, M. E. Tometsko, C. R. Maliszewski, et al. 1999. RANK is graphics. Acta Crystallogr. D Biol. Crystallogr. 60: 2126–2132. essential for osteoclast and development. Dev. 13: 2412– 32. Adams, P. D., R. W. Grosse-Kunstleve, L. W. Hung, T. R. Ioerger, A. J. McCoy, 2424. N. W. Moriarty, R. J. Read, J. C. Sacchettini, N. K. Sauter, and T. C. Terwilliger. 5. Kong, Y. Y., H. Yoshida, I. Sarosi, H. L. Tan, E. Timms, C. Capparelli, 2002. PHENIX: building new software for automated crystallographic structure S. Morony, A. J. Oliveira-dos-Santos, G. Van, A. Itie, et al. 1999. OPGL is a key determination. Acta Crystallogr. D Biol. Crystallogr. 58: 1948–1954. regulator of osteoclastogenesis, lymphocyte development and lymph-node or- 33. Naismith, J. H., and S. R. Sprang. 1998. Modularity in the TNF-receptor family. ganogenesis. Nature 397: 315–323. Trends Biochem. Sci. 23: 74–79. 6. Li, J., I. Sarosi, X. Q. Yan, S. Morony, C. Capparelli, H. L. Tan, S. McCabe, 34. Schneeweis, L. A., D. Willard, and M. E. Milla. 2005. Functional dissection of R. Elliott, S. Scully, G. Van, et al. 2000. RANK is the intrinsic hematopoietic cell osteoprotegerin and its interaction with receptor activator of NF-kappaB ligand. surface receptor that controls osteoclastogenesis and regulation of bone mass and J. Biol. Chem. 280: 41155–41164. . Proc. Natl. Acad. Sci. USA 97: 1566–1571. 35. Emery, J. G., P. McDonnell, M. B. Burke, K. C. Deen, S. Lyn, C. Silverman, 7. Simonet, W. S., D. L. Lacey, C. R. Dunstan, M. Kelley, M. S. Chang, R. Lu¨thy, E. Dul, E. R. Appelbaum, C. Eichman, R. DiPrinzio, et al. 1998. Osteoprotegerin H. Q. Nguyen, S. Wooden, L. Bennett, T. Boone, et al. 1997. Osteoprotegerin: is a receptor for the cytotoxic ligand TRAIL. J. Biol. Chem. 273: 14363–14367. 252 COMPLEX STRUCTURE OF RANKL WITH OPG N-TERMINAL CRDs

36. Truneh, A., S. Sharma, C. Silverman, S. Khandekar, M. P. Reddy, K. C. Deen, 42. Hymowitz, S. G., M. P. O’Connell, M. H. Ultsch, A. Hurst, K. Totpal, M. M. McLaughlin, S. M. Srinivasula, G. P. Livi, L. A. Marshall, et al. 2000. A. Ashkenazi, A. M. de Vos, and R. F. Kelley. 2000. A unique zinc-binding site Temperature-sensitive differential affinity of TRAIL for its receptors. DR5 is the revealed by a high-resolution X-ray structure of homotrimeric Apo2L/TRAIL. highest affinity receptor. J. Biol. Chem. 275: 23319–23325. Biochemistry 39: 633–640. 37. Vitovski, S., J. S. Phillips, J. Sayers, and P. I. Croucher. 2007. Investigating the 43. Shahbazi, S., P. J. Lenting, C. Fribourg, V. Terraube, C. V. Denis, and interaction between osteoprotegerin and receptor activator of NF-kappaB or O. D. Christophe. 2007. Characterization of the interaction between von Wil- tumor necrosis factor-related apoptosis-inducing ligand: evidence for a pivotal lebrand factor and osteoprotegerin. J. Thromb. Haemost. 5: 1956–1962. role for osteoprotegerin in regulating two distinct pathways. J. Biol. Chem. 282: 44. Zannettino, A. C., C. A. Holding, P. Diamond, G. J. Atkins, P. Kostakis, 31601–31609. A. Farrugia, J. Gamble, L. B. To, D. M. Findlay, and D. R. Haynes. 2005. 38. Johnstone, R. W., A. J. Frew, and M. J. Smyth. 2008. The TRAIL apoptotic Osteoprotegerin (OPG) is localized to the Weibel-Palade bodies of human vas- pathway in cancer onset, progression and therapy. Nat. Rev. Cancer 8: 782–798. cular endothelial cells and is physically associated with von Willebrand factor. J. 39. Cha, S. S., B. J. Sung, Y. A. Kim, Y. L. Song, H. J. Kim, S. Kim, M. S. Lee, and Cell. Physiol. 204: 714–723. B. H. Oh. 2000. Crystal structure of TRAIL-DR5 complex identifies a critical 45. Baud’huin, M., L. Duplomb, S. Te´letche´a, C. Charrier, M. Maillasson, role of the unique frame insertion in conferring recognition specificity. J. Biol. M. Fouassier, and D. Heymann. 2009. Factor VIII-von Willebrand factor com- Chem. 275: 31171–31177. plex inhibits osteoclastogenesis and controls cell survival. J. Biol. Chem. 284: 40. Hymowitz, S. G., H. W. Christinger, G. Fuh, M. Ultsch, M. O’Connell, 31704–31713. R. F. Kelley, A. Ashkenazi, and A. M. de Vos. 1999. Triggering cell death: the 46. Ando, K., K. Mori, F. Re´dini, and D. Heymann. 2008. RANKL/RANK/OPG: key crystal structure of Apo2L/TRAIL in a complex with death receptor 5. Mol. Cell therapeutic target in bone oncology. Curr. Drug Discov. Technol. 5: 263–268. 4: 563–571. 47. Tanaka, S., K. Nakamura, N. Takahasi, and T. Suda. 2005. Role of RANKL in 41. Mongkolsapaya, J., J. M. Grimes, N. Chen, X. N. Xu, D. I. Stuart, E. Y. Jones, physiological and pathological bone resorption and therapeutics targeting the and G. R. Screaton. 1999. Structure of the TRAIL-DR5 complex reveals RANKL-RANK signaling system. Immunol. Rev. 208: 30–49. mechanisms conferring specificity in apoptotic initiation. Nat. Struct. Biol. 6: 48. Wells, J. A., and C. L. McClendon. 2007. Reaching for high-hanging fruit in 1048–1053. drug discovery at protein-protein interfaces. Nature 450: 1001–1009. Downloaded from http://www.jimmunol.org/ by guest on September 26, 2021