doi:10.1016/j.jmb.2009.01.006 J. Mol. Biol. (2009) 386, 841–853 Available online at www.sciencedirect.com Crystal Structure of Myeloid Cell Activating Receptor Leukocyte Ig-like Receptor A2 (LILRA2/ILT1/LIR-7) Domain Swapped Dimer: Molecular Basis for Its Non-binding to MHC Complexes Yong Chen1,2,3, Feng Gao1,4, Fuliang Chu1, Hao Peng1, Lili Zong1,5, Yiwei Liu6, Po Tien1 and George F. Gao1,2,3⁎ 1Key Laboratory of Pathogenic The leukocyte Ig-like receptor (LILR/ILT/LIR) family comprises 13 Microbiology and Immunology, members that are either activating or inhibitory receptors, regulating a Institute of Microbiology, broad range of cells in the immune responses. LILRB1 (ILT2), LILRB2 (ILT4) Chinese Academy of Sciences, and LILRA1 (LIR6) can recognize MHC (major histocompatibility complex) Beijing 100101, China class I or class I-like molecules, and LILRB1/HLA-A2, LILRB1/UL18 and LILRB2/HLA-G complex (extracellular domains D1D2) structures have 2China-Japan Joint Laboratory of been solved recently. The details of binding to MHC have been described. Molecular Immunology and Despite high levels of sequence similarity among LILRA1, LILRA2 (ILT1), Molecular Microbiology, LILRA3 (ILT6) and LILRB1/B2, all earlier experiments showed that LILRA2 Institute of Microbiology, does not bind to MHC, but the reason is unknown. Here, we report the Chinese Academy of Sciences LILRA2 extracellular D1D2 domain crystal structure at 2.6 Å resolution, (CAS), Beijing 100101, China which reveals structural shifts of the corresponding MHC-binding amino 3College of Life Sciences, acid residues in comparison with LILR B1/B2, explaining its non-binding to Graduate University, Chinese MHC molecules. We identify some key residues with great influence on the Academy of Sciences, Beijing local structure, which exist only in the MHC-binding receptors. Moreover, 100049, China we show that LILRA2 forms a domain-swapped dimer. Further work with 4 these key swapping residues yields a monomeric form, confirming that the Institute of Biophysics (IBP), domain-swapping is primarily amino acid sequence-specific. The structure Chinese Academy of Sciences, described here supports the dimer conformation in solution observed Beijing 100101, China earlier, and implies a stress-induced regulation by dimerization, consistent 5Department of Obstetrics and with its function as a heat shock promoter. Gynaecology, Zhujiang © 2009 Elsevier Ltd. All rights reserved. Hospital, Nanfang Medical University, Guangzhou 510280, China 6Laboratory of Structural Microbiology, Institute of Microbiology, Chinese Academy of Sciences, Beijing 100101, China *Corresponding author. Key Laboratory of Pathogenic Microbiology and Immunology, Institute of Microbiology, Chinese Academy of Sciences, Beijing 100101, China. E-mail address: [email protected]. Present address: F. Chu, Department of Lymphoma and Myeloma, Center for Cancer Immunology Research, the University of Texas M. D. Anderson Cancer Center, Houston 77054, USA. Abbreviations used: LILR/LIR, leukocyte immunoglobulin-like receptor; ILT, immunoglobulin-like transcript; RA, rheumatoid arthritis; MHCI, MHC class I molecule; MIC, MHC class I-related chain. 0022-2836/$ - see front matter © 2009 Elsevier Ltd. All rights reserved. 842 Structure of Leukocyte Ig-like Receptor A2 Received 28 September 2008; received in revised form 7 January 2009; accepted 9 January 2009 Available online 15 January 2009 Keywords: LILRA2; ILT1; crystal structure; MHC non-binding; domain Edited by I. Wilson swapping dimer Introduction results) and suggest its functional role in virus control mechanisms. Moreover, even with a highly conserved Leukocyte immunoglobulin-like receptors amino acid sequence, LILRA2 may be the only group 1 member that is a non-MHC class I (MHCI) binding (LILRs/ LIRs), also called immunoglobulin-like 15,20 transcripts (ILTs)1,2 or CD85,3 including 13 members receptor. For the LILR family, the key factor/s of (two are pseudogenes) that are either activating the MHC binding specificity remains elusive. receptors (LILRA) or inhibitory receptors (LILRB), Here, we present the complete crystal structure of can be divided into two groups according to the LILRA2 D1D2 at 2.6 Å resolution. To reveal that the molecular basis of LILRA2 does not bind to amino acid similarity responsible for the interaction 15,20 between LILRB1 (ILT2/LIR-1/CD85j) and HLA-A2/ MHCIs or MHC class I-related chain (MIC) β2m.4 Group 1 members include the inhibitory molecules as we have verified (our unpublished receptors LILRB1 and LILRB2 (ILT4/LIR-2/ results), we analyzed the structural differences of CD85d), the activating receptors LILRA1 (LIR6/ LILRA2 from other LILR family members. We found CD85i) and LILRA2 (ILT1/LIR-7/CD85h), and the the crucial residues that exist only in the MHC- soluble receptor LILRA3 (ILT6/LIR-4/CD85e) (Fig. binding members and have a substantial influence 1a). Group 2 includes LILR A4/A5/A6 and B3/B4/ on the local structure. Moreover, we describe 3D domain swapping of LILRA2 dimer formation as B5 (ILT 7/11/8 and ILT 5/3/LIR8), showing less 14 than 60% sequence identity with group 1 members our previous experiments observed in solution. (Fig. 1b). We identified two Trp residues in the hinge loop Most of the LILR family members contain four with a key role in the domain swapping procedure. immunoglobulin (Ig)-like domains in their extra- cellular domains (D1, D2, D3 and D4). The available Results crystal structures of LILR B1/B2/A5 include only – the D1 and D2 domains (D1D2).4 12 Though D1D2 of LILR B1/B2/B4/A5 can be purified in vitro using Overall structure classical refolding protocols,5,12,13 only the instabil- ity and precipitation of wild type LILRA2 (ILT1/ Unlike other group 1 members with known struc- LIR-7/CD85h) D1D2 has been documented.13 We tures,5,6,12 the structure of LILRA2 shows a swapped recently developed a method to solve this problem dimer, consistent with our previous work on this on the basis of the close sequence and structural protein in solution. As shown in Fig. 2,thetwoLILRA2 similarity between LILRA2 and LILR B1/B2/A5. A molecules form an intertwined dimer through cysteine-introduced mutation (R142C) was designed exchange with identical β-strands of domain 2, follow- to form a disulfide bond with the spare Cys132, ing the canonical 3D domain-swapping mechanism.21 which stabilizes the protein and enhances the The refined crystal structure of LILRA2 R142C recombinant production of the LILRA2 D1D2 pro- contains two molecules (I and II) in one crystal- tein for the subsequent structural and functional lographic asymmetric unit. Each molecule includes studies.14 two domains, D1 (residues 3–96) and D2 (residues LILRA2, a member of the group 1 activating 97–196), with Ser43 of molecule I disordered. receptors, is expressed mainly on myeloid cells, There are disulfide bonds between cysteine resi- and shows N80% sequence identity to LILR B1/B2, dues 26 and 74, 120 and 172, and 132 and 142, which but follows a different ligand binding and signal is consistent with the concept of stability engineer- pathway (Fig. 1a).15 The ligand to which LILRA2 ing by mutating Arg142 to Cys142 (R142C), which binds is unknown. We know only that LILRA2 re- forms a disulfide bond with Cys132 (equivalent to cruits the γ-chain of FcɛRI with a charged Arg resi- the relevant bonds of LILR B1/B2).14 due in the transmembrane region. Via ITAM motifs Two Ig domains of LILRA2 are arranged in a of FcɛRIγ, LILRA2 activates human basophils and V-shaped conformation (Fig. 2a). Each domain 1 is eosinophils16,17 as potential therapeutic targets in composed primarily of β strands arranged into two rheumatoid arthritis (RA) and other inflammatory anti-parallel β sheets, with a β sheet containing three diseases.18,19 anti-parallel β strands (A, B, and E), and another β The interactions of LILRA2 with virus-infected cells sheet containing five anti-parallel β strands (C′,C,F, were observed recently (Y.C. et al., unpublished G, and A′). The main structural folding involves two Structure of Leukocyte Ig-like Receptor A2 843 Fig. 1. Amino acid sequence alignments of LILR family receptors. The residues conserved in all receptors are shaded in red. The location of pairs of cysteine residues and disulfide bonds had been marked with the corresponding numbers. Secondary structure elements of LILRA2 are shown above the sequences. The asterisks indicate the residues involved in the MHCI binding of LILRB1 (blue for residues involved in an α3 domain contact; orange for residues involved in a β2m contact). The site-directed mutagenesis position for stability engineering of LILRA2 is marked with green squares. Accession codes for the sequences are: ILT1/LILRA2, AAC51176; ILT2/LILRB1, AAC51179; ILT3/LILRB4, AAG02024; ILT4/LILRB2, AAC51882; ILT5/LIR3/LILRB3, AAB88120 ILT6/LILRA3, AAC51885; LIR6/LILRA1, AAB87664; ILT7/ LILRA4, NM_012276; ILT8/LILRA6, AAD02204; ILT11/LILRA5, NP_871715. disulfide bonds between two pairs of highly con- D1,5,6 while a β strand termed C′ in LILRA2 D1 served Cys residues. Additional short regions of 310 pairs with strand C. helix are found between the E and F strands in LILRA2 D1, just like LILR B1/B2/A5 (Fig. 2b).5,6,12 Molecular basis for the non-binding of LILRA2 The polyproline II helix, observed in typical C2 Ig- to MHCI complexes like domain, is located in the F-G loop of each domain of LILRA2 D1D2. Nevertheless, the classical The crystal structures of the LILRB1-HLA-A24/ WSXWS motif22 is different in D1 and D2, with LILRB2-HLA-G complex10 showed that residues SSEYS in D1 but WSLPS in D2. The side chains of the involved in MHCI binding were highly conserved serine residues within these motifs are hydrogen among group 1 receptors (Fig. 1a).4 However, recent bonded to main chain atoms in strand F and con- studies indicated that LILRA2 is the only group 1 tribute to the stability of the core Ig fold.23 receptor that does not bind to any of the MHCIs The topology of domain 1 is similar to that of tested.15,20 LILR B1/B2/A5, but with some noteworthy diffe- Figure 3a shows an overview of the LILRB1- rences (Fig.