Crystal Structure of LGR4-Rspo1 Complex

Background: LGR receptors play important roles in many developmental processes. Results: The structure of human LGR4-Rspo1 complex was solved. Conclusion: Diverse mechanisms are utilized by LGRs in ligand recognition. Significance: Our structures are important for potential drug design. Leucine-rich repeat G-protein-coupled receptors (LGRs) are a unique class of G-protein-coupled receptors characterized by a large extracellular domain to recognize ligands and regulate many important developmental processes. Among the three groups of LGRs, group B members (LGR4–6) recognize R-spondin family proteins (Rspo1–4) to stimulate Wnt signaling. In this study, we successfully utilized the “hybrid leucine-rich repeat technique,” which fused LGR4 with the hagfish VLR protein, to obtain two recombinant human LGR4 proteins, LGR415 and LGR49. We determined the crystal structures of ligand-free LGR415 and the LGR49-Rspo1 complex. LGR4 exhibits a twisted horseshoe-like structure. Rspo1 adopts a flat and β-fold architecture and is bound in the concave surface of LGR4 in the complex through electrostatic and hydrophobic interactions. All the Rspo1-binding residues are conserved in LGR4–6, suggesting that LGR4–6 bind R-spondins through an identical surface. Structural analysis of our LGR4-Rspo1 complex with the previously determined LGR4 and LGR5 structures revealed that the concave surface of LGR4 is the sole binding site for R-spondins, suggesting a one-site binding model of LGR4–6 in ligand recognition. The molecular mechanism of LGR4–6 is distinct from the two-step mechanism of group A receptors LGR1–3 and the multiple-interface binding model of group C receptors LGR7–8, suggesting LGRs utilize the divergent mechanisms for ligand recognition. Our structures, together with previous reports, provide a comprehensive understanding of the ligand recognition by LGRs.

to stimulate the Wnt signaling pathways, regulating cell proliferation, differentiation, and adult stem cell maintenance (5,6).
Although LGR4 -6 contain a 7TM domain as the group A and group C receptors, the stimulation of Wnt signaling by LGR4 -6 was found to be independent of downstream G-protein activation, suggesting the functional difference of LGR4 -6 from LGR1-3 and LGR7-8. Here, we successfully utilized the "hybrid LRR technique" to obtain human LGR4 proteins from insect cells. We determined the crystal structure of human LGR4 alone at a resolution of 2.2 Å, as well as its complex structure with Rspo1 at a resolution of 2.25 Å. Our studies, together with the previously determined structures (19 -22), suggest that LGR4 -6 recognize R-spondins via a distinct mechanism from those of LGR1-3 and LGR7-8.

Plasmids, Reagents, and Mutagenesis-To express human
LGR4 and Rspo1 in insect cells, the LGR4 15 and LGR4 9 hybrids ( Fig. 1) were cloned into the pFastBac1 vector (Invitrogen) with a gp67 secretory signal peptide at the N terminus and a Fc-9ϫHis tag with a tobacco etch virus protease cleavage site at the C terminus. The two furin-like cysteine-rich regions of Rspo1 (residues 34 -135) were cloned into the pFastBac1 vector with an N-terminal gp67 signal peptide. For TOP/FOPFlash assays, full-length wild-type and mutant LGR4 were cloned into the pCDNA4 vector with a CD8 secretory signal peptide and a FLAG tag at the N terminus. Full-length Rspo1 and its mutants were cloned into the pCDNA4 vector with a C-terminal HA tag. All site-directed mutageneses were performed with the standard overlap-extension PCR method, and all plasmids were verified by DNA sequencing.
Protein Expression and Purification-Recombinant proteins were expressed in Sf9 insect cells using the Bac-to-Bac baculovirus expression system (Invitrogen). LGR4 15 and LGR4 9 recombinant proteins were purified from cell medium using nickel-nitrilotriacetic acid resins. After tobacco etch virus protease digestion to remove the C-terminal Fc-9ϫHis tag, the proteins were further purified by size exclusion chromatography using Superdex 200 10/300 GL column (GE Healthcare). Both purified proteins were concentrated to 10 mg/ml in the Tris-HCl buffer containing 50 mM Tris-HCl, pH 8.0, and 150 mM NaCl for crystallization. To obtain the recombinant LGR4-Rspo1 complex, Sf9 cells were co-infected with baculoviruses encoding LGR4 9 and human Rspo1 (residues 34 -135). After 60 h post-infection, the LGR4 9 -Rspo1 complex was purified from cell medium using nickel-nitrilotriacetic acid column and further purified using a Mono Q column (GE Healthcare). The purified LGR4 9 -Rspo1 complex was concentrated to 40 mg/ml in the buffer containing 50 mM Tris-HCl, pH 8.0, and 150 mM NaCl for crystallization.
Crystallization and Data Collection-All crystallization experiments were carried out with the hanging-drop vapor diffusion method at 16°C. In initial crystal screening, crystals of LGR4 15 and LGR4 9 appeared after 3 days. Crystals of LGR4 15 were selected for structural determination. After extensive optimization, diffraction-qualified crystals of LGR4 15 were obtained in the well condition containing 0.1 M HEPES, pH 7.5, and 2.0 M ammonium sulfate. Crystals of LGR4 15 were transferred into the well solution supplemented with 15% glycerol as the cryoprotectant and then flash-cooled into liquid nitrogen for data collection. The LGR4 9 -Rspo1 complex was crystallized after 7 days in the well solution containing 0.1 M MES, pH 6.5, 0.2 M sodium thiocyanate, and 22% PEG 3350. The crystals were cryoprotected with the well solution supplemented with 20% glycerol in liquid nitrogen. All diffraction data were collected on the BL17U1 beamline of Shanghai Synchrotron Radiation Facility and processed with the HKL-2000 package (23).
Structural Determination-The LGR4 15 structure was solved with the molecular replacement method using the partial model of LRR repeats from TLR5 (PDB code 3V44) (24) as the search mode in the Phaser program (25) in CCP4 (26). The final structure was refined to 2.2 Å with R work /R free of 20.7%/23.4% in Refmac5 (27).There is one LGR4 15 molecule in an asymmetric unit. The LGR4 9 -Rspo1 complex structure was solved with the molecular replacement method using LGR4 15 and VLR (residues 129 -200) as the search modes. The final structure of LGR4 9 -Rspo1 complex was refined to 2.25 Å with R work /R free of 21.3%/24.2% in Refmac5. There are two LGR4 9 -Rspo1 complex molecules in an asymmetric unit. Because of the lack of electron density, the N-terminal 18 residues and C-terminal 5 residues of Rspo1 were missed in the final complex model. All model building was performed in Coot (28). All structures were checked with the program Procheck (29). All structural pictures were prepared in PyMOL. Statistics of data collection and refinement are listed in Table 1.
TOP/FOPFlash Assays-TOP/FOPFlash assays were carried out similarly as described (30). To test the effect of Rspo1 mutations, HEK293T cells cultured in 12-well plates were transfected with 500 ng of Super 8ϫ TOPFlash or FOPFlash firefly luciferase reporter, 10 ng of Renilla luciferase reporter, and 750 ng of wild-type LGR4 plasmids. Wild type and mutant Rspo1 conditional medium (CM) were prepared by culturing HEK293T cells in 6-well plates, which were transiently transfected with 2 g of Rspo1 variant plasmids in each well. Wnt3a CM was obtained from mouse L cells stably expressing Wnt3a. After a 24-h co-stimulation by Wnt3a and wild-type or mutant Rspo1 CM, the transfected HEK293T cells were harvested for Dual-Luciferase reporter assays by using the Dual-Glo luciferase assay kit (Promega). To test the effect of LGR4 mutations on the Rspo1 stimulation of Wnt signaling, HEK293T cells cul-tured in 12-well plates were transfected with 500 ng of Super 8ϫ TOPFlash or FOPFlash firefly luciferase reporter, 10 ng of Renilla luciferase reporter, and 1 g of wild-type or mutant LGR4 plasmids. After stimulation for 24 h with Wnt3a and wild-type Rspo1 CM, the cells were harvested for Dual-Luciferase reporter assays. The relative TOP/FOPFlash luciferase activities presented were normalized against the levels of vector control.

Production of Recombinant Human LGR4 and LGR4-Rspo1
Complex Proteins-The ectodomain of human LGR4 includes residues 28 -528. The two central tandem Fu-CRD domains of human Rspo1, which are required for binding the LGRs, include residues 34 -135. When expressed in insect cells, the human LGR4 ectodomain severely aggregated during purification. To enhance the solubility and stability of the LGR4 ectodomain, we designed LGR4 chimeras that contain the LRR repeats of the LGR4 ectodomain and the hagfish VLR protein using the hybrid LRR technique, which was previously used to facilitate the soluble expression and crystallization of TLR4 ( Fig. 1A) (31). Previous studies have suggested that the N-terminal residues of LGRs are important for interaction with R-spondins (16). Therefore, we attempted to fuse the N-terminal LRR repeats of LGR4 with the C-terminal regions of the hagfish VLR protein according to the "LXXLXLXXNXLXXL" consensus sequence of LRR modules.
After many attempts, we successfully expressed and crystallized two LGR4 ectodomain fusion proteins, LGR4 15 and LGR4 9 (Fig. 1, A and B). In LGR4 15 , the N-terminal 15 LRRs of LGR4 (residues 28 -396) were fused with the VLR residues 131-200 (VLR A ). In LGR4 9 , the N-terminal nine LRRs of LGR4 (residues 28 -253) were fused with the C-terminal 72 residues of VLR (residues 129 -200, VLR B ). Both chimeric proteins formed stable protein complexes with Rspo1 during purification, suggesting that our designed LGR4 proteins contain the entire Rspo1-binding region. LGR4 15 crystals were selected for structural determination because the construct contains more residues of LGR4. The LGR4 15 structure was finally refined to 2.2 Å with R work /R free values of 20.7%/23.4% (Table 1). Because LGR4 15 precipitated in polyethylene glycols, which are extensively used in the crystallization of protein complexes, we co-expressed LGR4 9 with the two central tandem Fu-CRD domains of Rspo1 (residues 34 -135) in insect cells to assemble the receptor-ligand complex for crystallization. We successfully crystallized the LGR4 9 -Rspo1 complex and determined its structure to 2.25 Å with high geometry quality ( Table 1). Details of the data collection and structural refinement statistics are summarized in Table 1.
Human LGR4 Ectodomain Structure-The final model of the ligand-free LGR4 15 structure includes all residues of the designed hybrid construct. The overall architecture of LGR4 15 adopts a typical twisted solenoid-like structure (Fig. 1B). As expected, the central 15 LRR repeats of LGR4 15 are flanked by the N-terminal LRRNT (residues 28 -57) and the C-terminal VLR module (Fig. 1B). In total, 17 ␤-strands are present in the LGR4 15 structure, of which 15 originate from the LRR repeats, one from LRRNT, and one from VLR. The 17 ␤-strands form the inner concave surface of LGR4 15 , which interacts with Rspo1 in the LGR4-Rspo1 complex structure (details below). The outer convex surface of LGR4 15 is composed of 10 linking loops and four short linking ␣-helices from the LRR repeats ( Fig. 1B).
A structural homology search using the LGR4 15 structure (LRRNT and LRR(1-15)) as the bait in the Dali server (32) revealed that LGR4 15 shares significant structural similarities with several LRR receptors, including LINGO-1, TLR2, TLR4, and TLR5 (24,31,33,34). LGR4 15 exhibits the highest structural similarity to the extracellular LRR subdomain of the human neuron co-receptor LINGO-1 (PDB code 2ID5) with a Z-score value of 32.4 and a root mean square deviation value of 4.4 Å for 347 aligned residues (Fig. 1D) (34). LINGO-1 forms a ternary complex with NgR1 and p75 to constitute the Nogo receptor for myelin and mediates the inhibition of axonal growth (34). Although the overall structure of LGR4 15 is highly similar to the extracellular LRR subdomain of LINGO-1, the C terminus of LGR4 15 (LRR11-15) is largely different from that of LINGO-1 (Fig. 1D).
Structure of the Human LGR4-Rspo1 Complex-The recombinant LGR4-Rspo1 complex for crystallization was prepared from the co-expression of LGR4 9 with the two central Fu-CRD domains of Rspo1 in insect cells ( Fig. 2A). LGR4 9 formed a stable 1:1 complex with Rspo1 during size exclusion chroma-tography. We successfully crystallized the LGR4 9 -Rspo1 complex and determined its structure by the molecular replacement method using LGR4 15 and VLR as the search models. In the final structural model of the LGR4 9 -Rspo1 complex, there are two LGR4 9 -Rspo1 complex copies related by the noncrystallography symmetry in an asymmetric unit. Because the two complex molecules adopt identical conformation, hereafter, we describe one of them in our structural analysis.
Like the ligand-free LGR4 15 structure, Rspo1-bound LGR4 9 adopts a short twisted solenoid-like structure and contains the N-terminal LRRNT, the central nine leucine-rich repeats (LRR1-9), and the C-terminal hagfish VLR region (Fig. 2B). All of the residues in the designed LGR4 9 chimera are included in the final LGR4 9 model. In the complex, Rspo1 horizontally crosses the inner concave surface of LGR4 9 from one side to the other side and binds to the concave surface through extensive interactions with the central six LRR repeats (LRR4 -9) of LGR4 9 (Fig. 2B).
Rspo1 Structure in the LGR4-Rspo1 Complex-Rspo1 in complex with LGR4 is composed of eight ␤-strands that form an elongated and flat overall architecture (Fig. 3A). The final Rspo1 model only includes the central 78 residues, which constitute eight ␤-strands (␤1-␤8) in the structure (Fig. 3B). The N-terminal 18 residues (Ala-34 -Asn-51) and the C-terminal five residues (Glu-131-Ala-135) of Rspo1 are missing in the structure due to lack of electron density. These residues are not conserved in the four R-spondin proteins (Fig. 3B) and do not participate in any crystal contact.
Rspo2-4 share high sequence homology with Rspo1 in their furin-like domains (Fig. 3B) (7,18), suggesting that the furinlike domains of Rspo2-4 also adopt the elongated flat ␤-fold structures similar to Rspo1. The 10 cysteine residues of Rspo1, which form five disulfide bonds within the hairpins, are highly conserved in all R-spondin proteins (Fig. 3B). Previous clinical studies have revealed that the mutations of the cysteine residues completely disrupt Rspo4 function and cause human autosomal-recessive anonychia (14,15), suggesting that the five disulfide bonds are critical for maintaining the flat structures of R-spondins. Furin-like cysteine-rich regions also exist in a variety of proteins that are involved in signal transduction via receptor tyrosine kinases (35). Our structural homology search revealed that Rspo1 just exhibits low structural similarity to subdomain IV, the furin-like region of the extracellular region of the EGF receptors HER2, -3, and -4 (36 -38). Subdomain IV of EGF receptors has been reported to mediate the inter-receptor dimerization of HER2, -3, and -4 to regulate cellular growth and differentiation (36 -38). Rspo1 was superposed with subdomain IV only at its middle hairpin ␤3-␤4 in structural superimposition.
In Mutagenesis Analysis-To examine the interactions observed in the complex structure, we mutated the interacting residues of Rspo1 and LGR4 and carried out the ␤-catenin-driven luciferase reporter assays (Fig. 4, D and E). A single mutation of the nonconserved Asp-85 of Rpso1 to alanine just had a minor effect on the stimulation of Wnt signaling. But the R87A mutant and the D85A/R87A mutant of Rspo1 could not sufficiently stimulate Wnt signaling (Fig. 4D), which indicates that the conserved Arg-87 plays a key role in the LGR4-Rspo1 interaction, and the electrostatic interactions in region 1 are required for Rspo1 recognition by LGR4. The Rspo1 mutations (F106A and F110A) of Phe-106 and Phe-110 at the hydrophobic binding region completely abolished the Rspo1-mediated stimulation of Wnt signaling (Fig. 4D), which suggests that the hydrophobic interactions between Rspo1 and LGR4 are also essential for the LGR4-Rspo1 interaction, and the conserved residues Phe-106 and Phe-110 of Rspo1 are critical for the interactions. Different from the F106A and F110A mutations, the N119A and K122A mutants of Rspo1 slightly affected the stimulation of Wnt signaling, suggesting that the nonconserved Asn-119 and Lys-122 of Rspo1 just play supplementary roles in the interactions between Rspo1 and LGR4. The critical residues Arg-87, Phe-106, and Phe-110 of Rspo1 are located in the first furin-like domain (␤1-␤4) and the second furin-like domain (␤5-␤8), respectively, which well explains why both furin-like domains of R-spondins are required for the recognition by LGRs.  JANUARY 23, 2015 • VOLUME 290 • NUMBER 4

JOURNAL OF BIOLOGICAL CHEMISTRY 2459
We also mutated the interacting residues of LGR4 to evaluate their effects on the Rspo1 stimulation of Wnt signaling (Fig.  4E). Consistent with the mutational analysis of Rspo1, the mutations of Arg-135, Asn-226, and Thr-229/Glu-252 in LGR4, which interact with the nonconserved residues of Rspo1, had a minor effect on the Rspo1 stimulation of Wnt signaling. The double mutation (N114A/D137A) of Asn-114 and Asp-137 of LGR4, which interact with Arg-87 of Rspo1 in the hydrophilic binding region, severely affected the Rspo1 stimulation of Wnt signaling. The A181D/T183D and A181D/V204D mutations, which disrupt the hydrophobic binding region of LGR4, also completely abolished the Rspo1 stimulation of Wnt signaling. Therefore, both the electrostatic and hydrophobic interactions between Rspo1 and LGR4 are required for Rspo1 recognition by LGR4.
Comparison of LGR4 with LGR5 and LGR6 -We further examined the conformational changes of LGR4 upon Rspo1 binding. The structure of the LGR4 9 -Rspo1 complex was superimposed with ligand-free LGR4 15 . As shown in Fig. 5, LRRNT and the central LRR1-9 repeats of LGR4 9 adopt identical conformations as those of LGR4 15 (Fig. 5A). The main chain and secondary structures of LGR4 do not undergo conformational changes upon Rspo1 binding, suggesting that LGR4 utilizes a rigid structural fold to recognize R-spondins to stimulate Wnt signaling. Our LGR4 9 and Rspo1 structures were also well superimposed with the previously determined LGR4-Rspo1 complex structure (PDB code 4KT1) (Fig. 5B) (21). All results of our luciferase assays for the LGR4 and Rspo1 mutations are consistent with the previously published data (20,21), which further confirm our structures and mutagenesis analysis.
LGR5 and LGR6 serve as markers of specific stem cells and play important roles in stem cell maintenance. The ectodomains of LGR5 and LGR6 have significant sequence similarities to that of LGR4. Previous studies had reported the LGR5-Rpso1 complex structures (20). Structural comparison between LGR4 and LGR5 revealed that the ectodomain of LGR5 adopts a similar horseshoe-like structure as that of LGR4 (Fig. 5B). But the C-terminal LRRCT region of LGR5 contains an additional LGR4 and Rspo1 are colored in green and purple, respectively. The LGR4-interacting residues in Rspo1 are labeled with the R subscripts. D, effects of Rspo1 mutations on the stimulation of Wnt signaling. The potentiation of Wnt signaling by Rspo1 mutants was examined using ␤-catenin-driven Super 8ϫ TOP/FOPFlash luciferase reporter assays. HEK293T cells were transiently transfected with full-length LGR4, TOPFlash, or FOPFlash firefly luciferase, and Renilla luciferase plasmids and were then stimulated with the CM of Rspo1 mutants in the presence of Wnt3a CM. E, effects of LGR4 mutations on the stimulation of Wnt signaling. HEK293T cells were transfected with the LGR4 mutants, TOPFlash or FOPFlash firefly luciferase, and Renilla luciferase plasmids and were then stimulated with the CM of wild-type Rspo1 and Wnt3a. LGR4 9 (LRRNT and LRR(1-9)) was superimposed with LGR4 15 (LRRNT and LRR(1-15)). The VLR modules in LGR4 15 and LGR4 9 were omitted in structural superimposition. LGR4 9 and Rspo1 are colored as in Fig. 2B. LGR4 15 is colored in blue. B, structural comparison of the LGR4 9 -Rspo1 complex with the previously determined LGR4-Rspo1 and LGR5-Rspo1 structures. The LGR4 9 -Rspo1 complex is colored in green (LGR4 9 ) and purple (Rspo1). The previously determined LGR4-Rspo1 and LGR5-Rspo1 structures are colored in blue and orange, respectively. C, sequence alignment of the Rspo1-binding residues. The Rspo1-binding residues are colored in blue. All the Rspo1-binding residues of LGR4 are conserved in LGR5 and LGR6. D, sequence alignment of the residues around ␣ C in the LRRCT regions. The cysteine residues of disulfide bonds in the LRRCT regions are colored in red. The sequence of the ␣ C helix of LGR5 is colored in orange.
␣-helix (␣ C ), which mediated the dimerization of the LGR5-Rspo1 complex in crystals (details below) (20). Sequence alignment showed that the residues near the ␣ C region are not conserved in LGR4 -6 ( Fig. 5D).
Because all the Rspo1-interacting residues of LGR4 and LGR5 are conserved in LGR6 (Fig. 5C), LGR6 recognizes R-spondins through an identical surface as LGR4 and LGR5, which includes the central LRR4 -9 repeats. All R-spondin proteins bind to the LGRs, but the binding affinity varies slightly among the members (30). Rspo1 exhibits ϳ40% sequence similarity with the other three members (7,18). The critical LGR4binding residues Arg-87, Phe-106, and Phe-110 of Rspo1 are conserved in the four R-spondin proteins (Fig. 3B). But the interacting residues Asp-85, Asn-109, and Lys-122 of Rspo1 are highly divergent in Rspo2-4 (Fig. 3B). Therefore, the limited difference in the binding affinities between Rspo1-4 and LGRs is likely determined by these variable interacting residues in the four R-spondins.

Divergence of Ligand Recognition by
LGRs-To uncover the differences in ligand recognition, we structurally analyzed the molecular mechanisms of all LGRs. The molecular mechanisms of group A (LGR1-3) and group C receptors (LGR7-8) have been extensively studied. The complex structures of LGR1 (FSH receptor, FSHR) with FSH revealed that the heterodimeric FSH contains similarly ␤-folded ␣and ␤-chains and is bound into the concave surface of LGR1 in a hand-clasp fashion (Fig. 6A) (39,40). In addition, LGR1 provides a sulfotyrosine (sulfated Tyr-335) in the LRRCT/hinge region as a second interaction site with FSH (Fig. 6A). The concave surface in the LRR region of LGR1 represents a high affinity FSH-binding site. Interestingly, the monoclonal TSHR-stimulating antibody M22 also binds to the corresponding concave surface of LGR3 (TSHR) in the LGR3-M22 complex (Fig. 6B) (41), suggesting that the concave surface of LGR1 is the major binding site for FSH. The high affinity binding site of LGR1 was suggested to recruit FSH and mediate its conformational changes to form a sulfotyrosine-binding pocket for the sulfated Tyr-335 insertion. Tyr-335 of LGR1 is conserved in all group A receptors, and the tyrosine sulfation is also essential for the activation of LGR2 and LGR3 by luteinizing hormone and thyroid-stimulating hormone, respectively (42), suggesting that group A LGR universally utilizes the two-step mechanism of LGR1 for ligand recognition.
Relaxin and INSL3 are two-chain peptide hormones and play important roles in the control of adult fertility (43). Relaxin and INSL3 are specifically recognized by LGR7 and LGR8, respec- tively (44). Distinct from R-spondins and FSH, Relaxin and INSL3 are ␣-folded heterodimers of A-and B-chain linked by disulfide bonds, which are structurally similar to insulin. Previous biochemical studies had revealed that the recognition of relaxin and INSL3 by LGR7 and LGR8 requires multiple interactions between the ligands and various receptor domains. The primary binding occurs via the interactions between B-chain of relaxin/INSL3 with the inner ␤-sheets of the LRRs in the ectodomains of LGR7 and LGR8 (45,46). The primary binding site provides a high affinity interface for the ligands. A secondary low affinity binding site is located at the extracellular loops in the 7TM domains of LGR7 and LGR8, which binds the A chains of relaxin and INSL3 (47,48). Therefore, both the LRR regions and the 7TM extracellular loops are involved in the relaxin/ INSL3 recognition by LGR7 and LGR8, suggesting a unique multiple-interface mechanism of group C receptors.
Our LGR4-Rspo1 complex structure and previously determined LGR4 -5 structures revealed that the LRR region of LGR4 is the sole binding site for Rspo1, suggesting that LGR4 -6 recognize R-spondins via a one-site binding mechanism (Fig. 2) (20,21). Although the transmembrane E3 ubiquitin ligase ZNRF3 and its homologue RNF43 were reported to interact with R-spondins and form complexes with LGR4 -6 (49,50), the LGR5-Rspo1-RNF43 ternary complex structure demonstrated that LGR5 does not interact with RNF43 to affect Rspo1 binding (Fig. 6C) (19). In addition, LGR5 contains more redundancy of LRR repeats than LGR1 and LGR3 (Fig. 6, D and  E). The distinct one-site binding mechanism of R-spondin recognition by LGR4 -6, together with the structural differences between LGR5 and LGR1, suggests the mechanism divergence in ligand recognition of LGRs.
Potential Oligomerization of LGRs in Ligand Recognition-It was suggested that glycoprotein hormone receptors undergo dimerization in living cells (51,52). The complex of FSH and LGR1 without the LRRCT/hinge region was demonstrated to form a dimer at high concentration in solution (39). The crystal structure of FSH with the hinge-truncated LGR1 revealed that the dimer was formed by two FSH-LGR1 complex molecules via three ␤-strands on the convex surface of LGR1 in an asymmetric unit (Fig. 7A). The small dimerization interface indicated the weak interactions between the two protomers ( Fig.  7A), consistent with that the dimerization was just detected by the chemical cross-linking assays (39). In addition, recent studies suggested that LGR1 underwent trimerization upon FSH binding when the conserved Tyr-335 residue in the LRRCT/ hinge domain was sulfated (Fig. 7B) (40, 53). The trimerization LGR5 and Rspo1 are colored in blue and red, respectively. The disordered loops in the LRRCT region of LGR5 are shown as dashed lines. D, dimerization of the LGR5-Rspo1-RNF43 ternary complex in an asymmetric unit (PDB code 4KNG). LGR5, Rspo1, and RNF43 are colored in orange, blue, and cyan, respectively. E, two LGR4 9 -Rspo1 copies in an asymmetric unit in our complex structure. The two LGR4-Rspo1 copies interact with each other via the engineered VLR module and the LRRNT region.
of LGR1 provided an additional binding site for FSH, which involves the interaction of the convex surface of LGR1 with a neighboring FSH (40). The trimerization was also observed in the crystals of LGR1 with FSH that was deglycosylated at Asn-52 in the ␣-chain (53). As the trimerization buried a large surface area of each protomer, the FSH-binding capacity of LGR1 was increased to 3-fold for the deglycosylated FSH (53). Although the trimer model of LGR1 in FSH recognition could well explain some observation in biochemical and functional studies (53), the in vivo relevance of the LGR1-FSH trimerization and the actual oligomerization form in living cells still need to be determined.
Dimerization of the LGR5-Rspo1 complex was also observed in crystals, although LGR5 formed a 1:1 complex with Rspo1 in solution (Fig. 7C) (20). In the LGR5-Rspo1 binary complex structure (PDB code 4BSR and space group P22 1 2 1 ) (20), the dimerization of the LGR5-Rspo1 complex was mediated by the LGR5 ␣ C -helices and the Rspo1-LGR5 LRRCT interactions, forming a cross-dimer of LGR5-Rspo1 complex in an asymmetric unit (Fig. 7C). But in the LGR5-Rspo1-RNF43 ternary complex structure (19) (PDB code 4KNG and space group P2 1 2 1 2 1 ), a back-to-back dimer of the LGR5-Rspo1-RNF43 complex was formed by the LGR5-LGR5 and LGR5-RNF43 trans-interactions (Fig. 7D). It is unknown which form of dimerization functions in cells and whether the two dimerization forms represent different states of LGR5 during ligand recognition. Different from the LGR5-Rspo1 complex, the LGR4-Rspo1 complex did not dimerize in the previously determined structure (21) and our structures. The previously determined LGR4-Rspo1 structure (PDB code 4KT1 and space group P3 2 ) only contained one LGR4-Rspo1 molecule in an asymmetric unit. Although there are two LGR4-Rspo1 molecules in our structure, the two copies do not form a real dimer because the two copies interact with each other via the engineered VLR module and the LRRNT region (Fig. 7E).
Interestingly, Xenopus LGR4 was found to dimerize both in solution and in crystals (22). The dimerization of Xenopus LGR4 was mediated by the side-by-side interactions between LRRNT and the first four LRRs of the two LGR4 copies in an asymmetric unit, forming a tail-to-tail dimer in an asymmetric unit (22). But our human LGR4 15 , which contains the intact LRRNT and the first four LRR repeats, did not dimerize either in solution or in crystals. Given that human LGR4 has significant sequence similarities to that of Xenopus, it is possible that the C-terminal engineered VLR module of LGR4 15 has negative effects on the human LGR4 dimerization.
LGR7 and LGR8 were also reported to homodimerize and heterodimerize in cells, which regulated the negative cooperativity in relaxin/INSL3 binding (54). Dimerization and negative cooperativity in ligand binding have been generally demonstrated for the insulin receptor (55), ␤2-adrenergic receptor (56), and chemokine receptors (57). It will be interesting to investigate the negative cooperativity in R-spondin binding and its correlation with the dimerization of LGR4 -6. Taken together, LGRs likely undergo different types of oligomerization during ligand recognition.
Closing Remarks-LGR4 -6 receptors are important for stem cell maintenance by stimulating Wnt signaling. It is important to uncover the molecular mechanism of R-spondin recognition by LGR4 -6. In this study, we utilized the hybrid LRR technique to obtain functional human LGR4 proteins, and we determined the crystal structure of the LGR4-Rspo1 complex. Our studies, together with previous reports, suggest that LGR4 -6 recognize R-spondins via a one-site binding mechanism, which involves the electrostatic and hydrophobic interactions between R-spondins and the concave surface in the LRR region of LGR4 -6. Our analysis reveals the structural differences of the three groups of LGRs in ligand recognition, highlighting the divergence in the mechanisms of ligand recognition. Because of the important roles of LGR4 -6 receptors and R-spondin proteins in embryonic development and in many human diseases, our structures are important for potential drug design.