LILRA2/ILT1/LIR-7) Domain Swapped Dimer: Molecular Basis for Its Non-Binding to MHC Complexes

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.

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 extracellular 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. 2b). There is no D β-strand in all of HLA-A2 complex and the superposition of LILRA2 LILR B1/B2/A2/A5 D1. Corresponding to the D and LILRB1. Two distinct contact areas in the strand between strands C–E of a typical Ig-like LILRB1-HLA-A2 interface have been circled: the A′ α domain, two 310 helices replace it in LILR B1/B2 CFG of LILRB1 D1 contacting with the HLA-A2 3 844 tutr fLuoyeI-ieRcpo A2 Receptor Ig-like Leukocyte of Structure

Fig. 2 (legend on next page) Structure of Leukocyte Ig-like Receptor A2 845 domain; the LILRB1 D1-D2 inter-domain hinge LILRB1-HLA-A2. More hydrophobic interactions region binding to β2m domain.4 between LILRB2/α3 domain of HLA-G but less Superposition of the structures of LILRA2 D1D2 LILRB2/β2m binding is seen, providing a β2m- and LILRB1 in complex with HLA-A2 (PDB code independent MHCI recognition of LILRB2.10 Super- 1P7Q),4 as well as LILRB2 in complex with HLA-G position of the structures of LILRA2 and LILRB2 (PDB code 2DYP)10 results in a root-mean-square bound to the HLA-G-heavy chain reveals the non- deviation (R.M.S.D) of 1.41 Å and 2.10 Å for all Cα binding property of LILRA2 to HLA-G (Fig. 3d). atoms, respectively. A magnified view of the HLA- First, Arg36→His36 and the directional shift of A2 α3 domain-contacting region is shown in Fig. 3b. Tyr38 makes it impossible for LILRA2 to bind with There are many differences in the corresponding the AB loop in the HLA-G α3 domain. Second, the residues between LILRA2 and LILRB1 in this region, different conformations of residues 41 and 42 despite the high levels of sequence identity. First, the between LILRA2/LILRB2, because of a unique 310 shift of LILRB1 Arg36 to LILRA2 His36 results in a helix structure in LILRB2 (discussed in detail later), shorter side chain and can no longer interact via a make the hydrogen-bonding with α3domain salt bridge with the HLA-A2 Asp196. In addition, impossible for LILRA2. the directional change of LILRA2 Tyr38 is stabilized It is worth mentioning that all the corresponding by Trp46 and the hydrophobic environment. This β2m-binding residues in LILRA2, in comparison change shortens the C strand and reduces the with LILRB2, are highly conserved (Fig. 1a). Some hydrophobic interaction with HLA-A2 main chain local conformational difference is seen (Fig. 3e) but atoms. Second, Asn41 (Lys41 in LILRB1) loses the we cannot rule out the possibility that there might be salt bridge connecting Glu198 of HLA-A2 and some interaction between β2m and LILRA2. How- results in the positional and conformational changes ever, the fact that LILRA2 does not bind to MHC of Lys42. This eliminates the hydrogen bond within and LILRB2 binding to MHC is β2m-independent, Lys42 and Asp96 of β2m. Third, the close interaction suggests that these interactions, if any, would not between the C-C′ β strands in LILRA2 generates contribute a great deal to a measurable binding. steric hindrance for Arg39 not to re-orient away, unlike Arg39 in LILRB1/B2, from the region that Strands–helices differentiation in the LILR HLA α3 domain to bind. family members Figure 3c illustrates the interaction of the LILRB1 D1-D2 inter-domain hinge region with the β2m The LILRA2 structure has remarkable similarities domain. First, the shorter side chain of Val182 in to LILRA5, with the R.M.S.D. of LILR A2/A5 D1D2 LILRA2, compared to LILRB1 Glu184, prevents the as 0.93 [LILRA5 (2D3V)] for all Cα atoms, in contrast formation of a salt bridge with β2m Lys91. Second, to 1.32 [LILRA2-LILRB1 (1G0X)] and 1.51 [LILRA2- hydrophobic interaction between Ile100 of LILRB1 LILRB2 (2GW5)] with LILR B1/B2, respectively. and β2m Val85 contributes to the LILRB1-HLA-A2 In Fig. 3a, one of the differences between LILR A2/ binding, but the corresponding residue of LILRA2 A5 and LILR B1/B2 is presented in a red circle. The is Ser98, which is a hydrophilic residue. Third, in two 310 helices in LILR B1/B2 D1 between the C and LILRB1, the Gln18 residue interacts with Gln89 of E strands4,10 are replaced by the additional β strand the β2m domain. However, Gln18 in LILRA2 forms C′ in LILR A2/A5 D1 (Fig. 4a). a hydrogen bond with the carbonyl oxygen of Ile64 As discussed above, conformational differences of in the molecule, making Gln18–Gln89 interaction identical residues in the HLA contacting region impossible. Lastly, the unique local conformation of mainly explain the non-binding property of Trp66 in LILRA2, in comparison with LILRB1, pre- LILRA2/MHCIs. The strands–helices transition vents hydrogen bonding with β2m Ile92 (carbonyl induces the neighboring conformational and posi- oxygen) and the hydrophobic interaction with β2m tional changes of the MHCI-contacting residues. Val93. This uniqueness results from the hydrogen Many hydrophobic residues are located within this bonding between Gln18 and Ile64 mentioned above, region, and some are highly conserved. While making Gln18 almost parallel with Trp66. forming a deep hydrophobic pocket surrounded HLA-G is a non-classical MHC molecule. The by the C′ and E strands of LILRA2 D1 (Fig. 4b), LILRB2-HLA-G complex exhibits an overlapping these residues in LILR B1/B2 provide a hydropho- but distinct recognition mode compared with bic core10 (Fig. 4c) for the formation of each helix,

Fig. 2. An overview of the crystal structure of the domain-swapped LILRA2 (ILT1/LIR-7/CD85h) D1D2 dimer. The crystal structure of LILRA2 R142C contains two molecules (I and II) per crystallographic asymmetric unit. The contiguous β-strands (E of one molecule, B of another molecule) of the two molecules per crystallographic asymmetric unit, form inter-molecular hydrogen bonds and interact with each other. Strands E, F and G of LILRA2 D2 exchanged between the dimeric molecules as the “Swapped Domain”. (a) A ribbon drawing of the structure of LILRA2 colored by chain: molecule I, cyan; molecule II, green. (b) A topological diagram of the LILRA2 structure of the two molecules per crystallographic asymmetric unit. The arrows show the direction of the β-strands. Cylinders labeled PP indicate the polyproline II-type ′– – helices. Cylinders labeled 310indicate the 310 helices. The C E (149 153) loops of D2 were marked out in red as the hinge loop. The topology structures of the D2 of the two molecules were modified to present the interaction between two molecules according to the details observed. (c) Model diagram of two LILRA2 molecules form dimer through domain swapping as they exchange their identical structural element (“domain”). The hinge loop is presented as red lines. O interface is the additional interaction interface between domain-swapped dimers that does not exist in the normal dimer. 846 Structure of Leukocyte Ig-like Receptor A2

Fig. 3. Structural comparison of the binding region of LILR B1/B2 and the corresponding sites of LILRA2. (a) An overview of the crystal structure of the LILRB1-HLA-A2 complex. The structure of LILRA2 (green) was superposed with LILRB1 (magenta). The MHC class I heavy chain is shown in blue while the light chain (β2m) is shown in brown. The residues with changes are shown in sticks. The regions involved in the LILRB1-MHCIs contact are enclosed in a black β α circle. The region of the switch from 310 helix to strand is marked with red circle. (b) Details of the LILRB1-HLA-A2 3 ′ β domain (blue) contact interface. The direction of both the 310 helix in LILRB1 and the C strand in LILRA2 are indicated. (c) Details of the LILRB1-HLA-A2 light chain (β2m) (brown) contact interface. (d) Details of the LILRB2 (yellow) -HLA-G α3 domain (blue) contact interface. (e) Details of the LILRB2-HLA-G light chain (β2m) (brown) contact interface. resulting in conformational changes of the residues LILR family demonstrate the key binding residues interacting with MHCIs such as Tyr38, Arg39 and Lys41 and Lys42 (KK), and crucial residues influen- Lys42 around this region (Fig. 3b). Studies of the cing the helices formation; Ile47 and Thr48 (IT) for crystal structures (LILR A2/A5/B1/B2), ligand helix1 and Leu54 and Val55 (LV) for helix2 (Fig. 4d). binding properties and multiple alignments of the These highly conserved residues are present in all Structure of Leukocyte Ig-like Receptor A2 847

Fig. 4. A structural comparison of the deep hydrophobic pocket of activating receptors LILR A2/A5 and the corresponding region of the inhibitory receptors LILR B1/B2. (a) Superposition of the local structures around the 310 helix of LILR B1/B2 and the C′ strand of LILR A2/A5: LILRA2, green; LILRB1, magenta; LILRB2, orange; LILRA5, blue. (b) Details of the structure of the C′ strand of LILRA2. The hydrophobic residues forming the deep hydrophobic pocket of LILRA2 are marked in magenta. Residues in the 310 helix that are changed between the E and F strands are shown in orange. The location of the unique potential N-linked glycosylation sites of LILRA2 is high lighted in blue. (c) Details of the structure of the 310 helix of LILRB1. The hydrophobic residues located around this area are marked in magenta. (d) Multiple alignments of the residues around strands/helices region among LILR family members. The secondary structures of this region of known structure receptors is shown as: strands, red lines; helices, blue lines. The crucial residues involved in the formation of each helix are in the orange/green box. The conserved residues of contact with MHCIs are in a magenta box. Whether each member is a receptor of MHCI molecules is marked at the right-hand side: MB, MHCI binding receptors; NMB, non-MHCI binding receptors.

MHCIs binding receptors. Clearly, this strands– LILRA2 D2 are exchanged between molecules I and helices transition in the LILR receptor family can II. As a result, the strands of the C-C′-E pair with differentiate the members into two sub-groups, NB each other, in contrast to the typical conformation and NMB as shown in Fig. 4d. (Fig. 5a). Moreover, corresponding to the D strand between Domain swapping between molecules I and II strands C′ and E in LILR B1/B2/A5 D2, residues 147–153 of LILRA2 D2 still behave as a loop region, The structure of the domain-swapped LILRA2 and introduce additional hydrogen bonds (Fig. 5d). dimer is illustrated in Fig. 2a and b. Structures of the These hydrogen bonds have an important role as the typical Ig-like domains, such as LILR B1/B2/A5 D2 hinge loop in domain swapping (Fig. 2c); they and LILRA2 D1, show that the strands C′ and E extend the C′ strands and form the O interface of the belong to the two different anti-parallel β sheets, domain-swapped dimer, which is the extra interac- pairing with the β strands C and B, respectively. tion interface between the domain-swapped dimer However, as swapped domain, strands E, F and G of (Fig. 2c). Consequently, the domain-swapped dimer 848 Structure of Leukocyte Ig-like Receptor A2

Fig. 5. Domain swapping between the dimeric molecules of LILRA2 and the unique MW interaction pattern. Molecule I is cyan, molecule II is green. C, (a–c) C′ and E strands of molecule I, as well as A, B and F strands of molecule II, these six anti-parallel β-strands form a unique MW pattern. The inter-molecular interface, as well as the region of the hinge loop, is marked with a black circle. As the site of stability engineering, the FILC and EHPQR/C region is surrounded by a red circle, where disulfide bonds (orange line) form in LILR B1/B2 and a unique separate Cys132 residue is located in wild type LILRA2. (b) Details of the inter-molecular interface, as well as the domain-swapping region and the hinge loop. Molecule II of LILRA2 was superposed with LILRB1 (magenta). The distance between the residues within and between the molecules of LILRA2 are shown as orange/black lines. The distances for corresponding residues of LILRB1 and LILRA2 are shown as blue lines. (c) Details of the FILC and EHPQR/C region, which forms the foundation of stability engineering. LILRA2 was superposed with LILRB1 (magenta), LILRB2 (orange) and LILRA5 (blue). The cysteines located in this region are shown in sticks, and the disulfide bonds are shown as orange lines. (d) Cross-eye stereo view of the electron density map for the two anti-parallel C′-E β-strands containing residues Asn144–Phe157 (NSHSHARWSWAIF) located in the hinge loop region at the LILRA2 dimer interface. Hydrogen bonds between the two main chains are shown as broken red lines. The map is contoured at 1.0 σ. with the extended C′ strand and O interface is reveals that the C′ strand of LILRA2 D2 was drawn intrinsically lower in energy than the monomer away from the corresponding position of the C′ conformation. This change should result in the strand of LILRB1 D2 as a result of the interaction absence of a small separate D strand from LILRA2 with the contiguous swapped E strand. His148 of D2 (Fig. 2b), in contrast to LILR B1/B2/A5, and the LILRA2, is about 11 Å away from the corresponding stability of the domain-swapped dimer, which does residue His150 of LILRB1 (Fig. 5b). It performs just not easily exchange with the monomer form.14 like the arm of the letter W. In D2, as Fig. 2b shows, the swapped E strand of Intermolecular interface of LILRA2 and the molecule I, pairs with both the C′ strand of molecule unique MW interaction pattern I and the contiguous B strand of molecule II. Thus, these C′ and E strands, together with the β-strand C As illustrated in Fig. 5a, superposition of the of molecule I, as well as β-strands A, B and F of structure of LILRA2 and LILRB1 (PDB code 1G0X) molecule II, form a unique interaction pattern. This Structure of Leukocyte Ig-like Receptor A2 849

Fig. 6. Structure-based engineering and gel-filtration analysis to determine key residues for domain swapping. (a) Local structure of the LILRA2 dimer hinge loop. The two different Trp residues in the hinge loop are shown in ball-and-stick. The hydrogen bonds are shown as broken red lines. (b) Multiple alignments of successfully purified group 1 members. The two Trp residues of LILRA2 that are different from other members are marked by a green box. (c) The size-exclusion chromatography profiles of LILRA2 WSWR and LILRA2 R142C. The LILRB1 and LILRA5 results are overlapped as the monomer control. The profiles are marked, along with approx- imate positions of molecular mass standards of 43.0 kDa, 29.0 kDa and 13.7 kDa). As clearly shown, the monomeric form of LILRA2 runs in a similar position to LILRB1 and LILRA5, but between the dimer and monomer positions of the LILRA2 R142C. unique β sheet is called the MW pattern (Fig. 5a) those in LILR B1/B2/A3. These two aromatic resi- because it looks like the letters M and W. The E strand dues are located in the top of the hinge loop, and of molecule II acts in the same way at the opposite Trp154 provides an additional hydrogen bond with side, with which 12 anti-parallel β strands form two Ser147 (Fig. 6a). The corresponding residues of MW patterns, constituting a firm interface basket. LILR B1/B2/A3 are Ser152 and Arg154, respectively. So, we generated the LILRA2 mutant (named Identification of key residues for domain LILRA2WSWR) with these two Trp residues re- swapping in LILRA2 placed by Ser/Arg. The profiles of different gel-filtration chroma- Previous intensive studies indicated that domain tography runs with the same column are over- swapping is sequence-dependent.24 The native pro- lapped in Fig. 6c for comparison. As we observed tein topology alone is sufficient to determine whe- earlier,14 LILRA2 elutes as either the dimer or the ther domain swapping can occur and to define the monomer at the position of the lower molecular structure of the swapped complex.25 To investigate mass protein marker: the dimer behaves as if it is whether LILRA2 domain swapping obeys the same smaller than the 43 kDa marker protein, and the rule, and to identify which residue(s) has a crucial monomer behaves as if it is similar in size to the role in this protein dimerization, we performed the 13.7 kDa marker protein, consistent with the re- site-directed mutations based on the structure we sults of the dynamic light-scattering (DLS) studies. solved here. However, the gel-filtration chromatography, DLS Above, we described the conformational changes and analytical sedimentation results demonstrate of the hinge loop and the O interface involved in the mutant LILRA2 WSWR behaves as a typical the association of two molecules that makes the monomer, like LILRB1/A5, the two monomeric swapped state energy favorable (Fig. 5). Thus, we members belonging to group 1 and group 2, res- compared the amino acid sequences of the hinge pectively. This mutational analysis shows clearly loop and corresponding regions of all other success- that these two tryptophan residues are responsible fully obtained group 1 members, LILR B1/B2/A3, for domain swapping, in addition to their role in which exist as monomers. As shown in Fig. 6b, the formation of the hinge loop. Therefore, the do- multiple alignment of these proteins shows most main swapping of LILRA2 is also sequence-specific residues are conserved, except two residues in as observed in other classical domain-swapping LILRA2, Trp152 and Trp154, are different from molecules. 850 Structure of Leukocyte Ig-like Receptor A2

Discussion that whether the disulfide bonds exist (LILR B1/B2/ A2(R142C)) or not (LILR A2(wild type)/A5), the Here, we report the first example of the structure directions and positions of the C and C′ β strands are of a group 1 activating receptor, LILRA2 (D1D2), a almost identical. LILRA2 forms a domain-swapped non-MHC binder of the LILR family. As group 1 dimer through interchange of the hinge loop (C′-E members, LILRB1 in complex with either HLA-A2 loop) and the downstream β strands (E\F\G) or the CMV-derived UL18, and LILRB2 in complex between the dimeric molecules. This dimeric form with HLA-G have been described,4,10,11 Based on the remains stable due to the interaction of C′-E strands, comparison of these 3D structures, we demon- and the position of 142C-132C is far from the hinge strated the shifts of specific MHC class I-binding loop. Nevertheless, this disulfide bond connecting residues within LILRA2, which were caused by the the C-C′ strands would be an obstacle if C′-E strands neighboring strands–helices transition. On the basis tend to be at a much closer position during domain of the observation of this transition in the LILR swapping. Thus, R142C mutation would not have family, we propose the following: first, even though any effect on the domain-swapped dimer. Further- the structures and ligands of many LILR receptors more, our mutational work identified the primary have not been identified, we still have reason to amino acid sequence (Trp152 and Trp154) deter- believe that some receptors of the LILR family, if mining the swapping. they bind to MHCIs, either activating or inhibitory The LILRA2 dimerization might reflect its distinct receptors, may have the same 310 helix region loca- distribution and expression profiles. Transcriptional ted between the C and E strands of D1, and use a regulation analysis of the LILR family indicated mechanism in contacting MHCIs similar to that used that LILRA2 carries a core promoter whose activity by the LILRB1-HLA-A2 and LILRB2-HLA-G com- was augmented by heat shock elements,35 which is plexes. Second, the four crucial hydrophobic resi- a transcriptional mechanism distinct from that of dues (IT and LV) (Fig. 4d) affecting the 310 helix LILR B1/B2. Under stress conditions, viruses inva- formation (replacing the β strands), together with ding or malignant transition of the cells gives the the KK residues, may be one of the most important transient denaturing conditions and/or a heat characteristics with which to categorize the MHCIs shock signal, triggers the high expression of binding receptors of the LILR family.12 At least, the LILRA2, and finally induces the formation of known structures follow this rule. Third, LILRA3, domain-swapped dimers of LILRA2.28,36 As a result the only LILR member without a transmembrane of domain swapping, the two molecules are region but with four typical crucial hydrophobic brought into close proximity, providing a larger residues (IL and LV) and KK residues (Fig. 4d), ligand-binding platform with high-yield activating would have the potential to bind specific MHCIs. signals and mobilize the immune system to raise This would be consistent with its function as a the anti-viral (stress) pathways. soluble molecule that blocks the contact of other receptors with the same site of the MHCI,26 and needs to be addressed in the future. Materials and Methods We describe the particular pattern of domain swapping in LILRA2 molecules. The 3D domain Crystallization of LILRA2 D1D2, data collection and swapping provided a mechanism for protein to form processing the dimer or even oligomers as they exchange identical structural elements (domain).27 Protein dimers The gene of the first two Ig-like domains (D1D2) of have evolved because of their advantages over the LILRA2 was cloned into Escherichia coli expression monomers, which include a higher local concentra- vector and expressed as described.14 After a long period tion of active sites, larger binding surfaces, and of extensive crystal trials with limited amounts of the economic ways to produce large protein interaction wild-type LILRA2 D1D2 available, the LILRA2 R142C 21 D1D2 protein was rationally designed, prepared and networks and molecular machines. More than 100 14 cases of domain-swapped proteins have been struc- crystallized as described. Briefly, the protein was over- expressed in E. coli as inclusion bodies, renatured by turally characterized. 3D domain swapping holds dilution refolding, and purified by gel-filtration chro- great interest for the evolution of protein pathologi- 28 matography using a Superdex 75 column (GE Health- cal dimerization, and has been proven as a good care). After concentration, the concentration of LILRA2 pathway of dimer/monomer exchanges for many D1D2 R142C was adjusted to 10 mg/mL. All crystal- immune system-related proteins, such as CD4,29 lization attempts were done at 18 °C by the hanging- CD2,30 BCL-xL,31,32 Nod1,33 and GITRL.34 The study drop, vapor-diffusion method. Initial conditions were described here has identified a new member of this established using Crystal Screens I and II (Hampton kind of swapping, which might have functional Research). implications for LILRA2 ligand recognition. For data collection, crystals were transferred to 15% (v/v) The stability engineering site (R142C) introduces a glycerol in reservoir buffer for about 1 min before they were flash-frozen to 100 K in a stream of nitrogen gas. Cys residue to pair with the single Cys in wild type Data were collected with a Rigaku MicroMax007 rotating- LILRA2.14 As shown in Fig. 5c, the R142C mutation ′ anode X-ray generator operated at 40 kV and 20 mA is located in strand C , and forms a disulfide bond (CuKα; λ=1.5418 Å) equipped with an R-AXIS VII++ with Cys132 in strand C. Superposition of the struc- image-plate detector. Data were processed and scaled tures of LILRA2 R142C and LILR B1/B2/A5 reveals using DENZO and SCALEPACK.37 Structure of Leukocyte Ig-like Receptor A2 851

Model determination, structure solution, refinement Table 1. X-ray diffraction data processing and refinement and analysis statistics

The collected data were analyzed by molecular replace- A. Data collection ment using Molrep38 in the CCP4 package as described.6 Space group C2 Unit cell parameters The crystal packing contains two molecules (22 kDa each) a (Å) 41.84 per crystallographic asymmetric unit. b (Å) 72.97 To eliminate interference by the variety of inter- c (Å) 131.85 domain angles, we tried to locate domains 1 and 2 sepa- α (°) 90.00 rately. LILR B1/B2 D2 can be located individually; how- β (°) 90.32 ever, all of the D1 of LILR B1/B2 cannot provide an γ (°) 90.00 unambiguous solution (PDB: 1G0X, 1UGN, 1UFU, 1VDG, Resolution range (Å) 44.21–2.50 – a and 2GW5). (2.74 2.50) Finally, D1 of LILRA5 (PDB code 2D3V)12 was located Total reflections 42,760 Unique reflections 11,930 first, then D2 of LILRB1 (PDB code 1G0X) was found in a Molecules/asymmetric unit 2 rotation and translation function while D1 of LILRA5 was Average redundancy 6.68 (7.08) fixed. Manual rebuilding was accomplished with the Completeness (%) 97.0 (96.9) 39 – program Coot using 2Fo Fc and annealed omit maps, Rmerge (%) 7.4 (32.0) alternating with reciprocal space refinement in the crystal- I/σ 6.1 (1.8) lography and nuclear magnetic resonance system (CNS).40 Final rounds of refinement resulted in a final R-factor of B. Refinement Resolution (Å) 30–2.6 17.6% (Rfree 22.7%) for all data between 30 Å and 2.60 Å. For analysis of inter-domain contacts and buried surface R-factor (%) 17.6 R b (%) 22.7 areas, D1 is defined as residues 1–96 and D2 is defined as free – 6 R.M.S.D.from ideal residues 97 196, according to the structure of LILRB1. Bond lengths (Å) 0.011 Buried surface areas were calculated using SURFACE with Bond angles (°) 1.03 41 a 1.4 Å probe radius. The PyMOL molecular graphics Ramachandran plot quality system† was used to prepare the figures. The geometry of Residues in most favored regions (%) 72.3 the refined structure was validated according to the Residues in additionally allowed (%) 24.9 Ramachandran plot criteria.42 Average B-factor versus Residues in generously allowed (%) 2.8 residues was calculated using BAVERAGE.41 The refine- Residues in disallowed regions (%) 0 ment statistics of structure are given in Table 1. Average B value Main chain (Å2) 47.002 Side chains and water (Å2) 46.440 2 Structure-based engineering All atoms (Å ) 46.718 a Values in parentheses are given for the highest resolution On the basis of the structure we solved, we hypothesized shell. b that Trp152 and Trp154 are responsible for the swapped Rfree is calculated over reflections in a test set (5%) not dimer. Therefore, site-directed mutation of LILRA2 WSWR included in atomic refinement. (W152S and W154R) was performed using LILRA2 R142C plasmid as template and the QuickChange® Site-directed Mutagenesis Kit according to the manufacturer's protocol (Strategene Cloning Systems). proteins were detected by monitoring the absorbance at The primers are shown below with the mutated nucleic 280 nm. acids in bold and underlined: Dynamic light-scattering (DLS) SRmu1: 5′-GCCCGTGGGTCGTCCCGCGCCATCTTCTCC- ′ GTG-3 The DLS experiments were done with the LILRA2 SRmu2: 5′-CACGGAGAAGATGGCGCGGGACGACCCA- ′ dimer fraction, the LILRA2 monomer fraction, LILRA2 CGGGC-3 WSWR, and LILRB1 using Dynapro Titan™ (Wyatt Technology Corporation). All proteins were dissolved at Analytical size-exclusion chromatography a concentration of 1.0 mg/mL in 30 μL of 20 mM Tris–HCl (pH 8.0), 50 mM NaCl. The sample was centrifuged at μ The oligomeric status of both purified LILRA2 R142C 21,600g for 10 min, and a 20 l sample of the supernatant and LILRA2 WSWR was determined by gel-filtration was used to collect scattering data at 4 °C. At least ten chromatography at 20 °C in a HiLoad 16/60 Superdex scattering measurements were taken and averaged for one 75 pg column with AKTA FPLC (GE Healthcare). LILRB1 experiment. A regularization histogram was analyzed (ILT2) and LILRA5 (ILT11) expression plasmids were kind using DYNAMICS software version 6 (Wyatt Technology gifts from Dr. Bent K. Jakobsen (Avidex Limited, UK) and Corporation). Dr. Katsumi Maenaka (Kyushu University, Japan), respectively. These proteins were used as the monomer control. Protein Data Bank accession number The mobile phase was 20 mM Tris–HCl, 50 mM NaCl, pH 8.0, and the flow rate was 1.0 mL/min. The molecular The atomic coordinates have been deposited in the mass standards were: ovalbumin, 43 kDa; carbonic anhy- RCSB Protein Data Bank‡ with accession code 2OTP. drase, 29 kDa; and ribonuclease-A, 13.7 kDa. The eluted

† http://www.pymol.org ‡ http://www.rcsb.org/pdb 852 Structure of Leukocyte Ig-like Receptor A2

Acknowledgements receptor LILRB1 modulates the differentiation and regulatory potential of human dendritic cells. Blood, 111, 3090–3096. We thank Dr. Bent K. Jakobsen (Avidex Limited, 9. Lichterfeld, M., Kavanagh, D. G., Williams, K. L., UK) and Dr. Katsumi Maenaka (Kyushu University, Moza, B., Mui, S. K., Miura, T. et al. (2007). A viral CTL Japan) for their expression plasmids of LILRB1 and escape mutation leading to immunoglobulin-like LILRA5, Dr. Wenhui Li (Harvard Medical School), transcript 4-mediated functional inhibition of myelo- Prof. Xiaojiang Chen (University of Southern Cali- monocytic cells. J. Exp. Med. 204, 2813–2824. fornia) for discussing the experiments and advice, 10. Shiroishi, M., Kuroki, K., Rasubala, L., Tsumoto, K., Mr. Christopher Pannell for his critical reading of the Kumagai, I., Kurimoto, E. et al. (2006). Structural manuscript, and Dr. Zheng Fan for assistance in DLS basis for recognition of the nonclassical MHC experiments operation. This work was supported by molecule HLA-G by the leukocyte Ig-like receptor a grant from the Ministry of Science and Technology B2 (LILRB2/LIR2/ILT4/CD85d). Proc. Natl Acad. Sci. USA, 103, 16412–16417. (MOST) of China for the basic research program 973, 11. Yang, Z. & Bjorkman, P. J. (2008). Structure of UL18, grant no. 2006CB504204; a grant from the National a peptide-binding viral MHC mimic, bound to a Natural Science Foundation (NSFC) of China, grant host inhibitory receptor. Proc. Natl Acad. Sci. USA, no. 30671903 and a grant from the Chinese Academy 105, 10095–10100. of Sciences (CAS) Knowledge Innovation Project, 12. Shiroishi, M., Kajikawa, M., Kuroki, K., Ose, T., grant no. KSCX2-SW-227. G.F.G. is a distinguished Kohda, D. & Maenaka, K. (2006). Crystal structure of young investigator of the NSFC (grant no. 30525010). the human monocyte-activating receptor, “Group 2” leukocyte Ig-like receptor A5 (LILRA5/LIR9/ILT11). Y.C. is a PhD student supported by Science Innova- 281 – tion Project 2008, Graduate University of Chinese J. Biol. Chem. , 19536 19544. Academy of Sciences (GUCAS), no. 0729031EE1. The 13. Garner, L. I., Salim, M., Mohammed, F. & Willcox, – B. E. (2006). Expression, purification, and refolding China Japan Joint Laboratory of Molecular Immu- of the myeloid inhibitory receptor leukocyte immunology and Molecular Microbiology is supported, in noglobulin-like receptor-5 for structural and ligand part, by the Japan Ministry of Education, Culture, identification studies. Protein Expr. Purif. 47, Sports, Science and Technology (MEXT). 490–497. 14. Chen, Y., Chu, F., Gao, F., Zhou, B. & Gao, G. F. (2007). Stability engineering, biophysical, and biological cha- References racterization of the myeloid activating receptor immunoglobulin-like transcript 1 (ILT1/LIR-7/LILRA2). 1. Colonna, M., Navarro, F., Bellon, T., Llano, M., Garcia, Protein Expr. Purif. 56, 253–260. P., Samaridis, J. et al. (1997). A common inhibitory 15. Nakajima, H., Samaridis, J., Angman, L. & Colonna, receptor for major histocompatibility complex class I M. (1999). Human myeloid cells express an activating molecules on human lymphoid and myelomonocytic ILT receptor (ILT1) that associates with Fc receptor cells. J. Exp. Med. 186, 1809–1818. gamma-chain. J. Immunol. 162,5–8. 2. Cosman, D., Fanger, N., Borges, L., Kubin, M., Chin, 16. Tedla, N., Bandeira-Melo, C., Tassinari, P., Sloane, W., Peterson, L. & Hsu, M. L. (1997). A novel immu- D. E., Samplaski, M., Cosman, D. et al. (2003). noglobulin superfamily receptor for cellular and viral Activation of human eosinophils through leukocyte MHC class I molecules. Immunity, 7, 273–282. immunoglobulin-like receptor 7. Proc. Natl Acad. Sci. 3. Saverino, D., Fabbi, M., Ghiotto, F., Merlo, A., USA, 100, 1174–1179. Bruno, S., Zarcone, D. et al. (2000). The CD85/ 17. Sloane, D. E., Tedla, N., Awoniyi, M., Macglashan, LIR-1/ILT2 inhibitory receptor is expressed by all D. W., Jr, Borges, L., Austen, K. F. & Arm, J. P. human T lymphocytes and down-regulates their (2004). Leukocyte immunoglobulin-like receptors: functions. J. Immunol. 165, 3742–3755. novel innate receptors for human basophil activation 4. Willcox, B. E., Thomas, L. M. & Bjorkman, P. J. (2003). and inhibition. Blood, 104, 2832–2839. Crystal structure of HLA-A2 bound to LIR-1, a host 18. Huynh, O. A., Hampartzoumian, T., Arm, J. P., Hunt, and viral major histocompatibility complex receptor. J., Borges, L., Ahern, M. et al. (2007). Down-regulation Nature Immunol. 4, 913–919. of leucocyte immunoglobulin-like receptor expression 5. Willcox, B. E., Thomas, L. M., Chapman, T. L., in the synovium of rheumatoid arthritis patients after Heikema, A. P., West, A. P., Jr & Bjorkman, P. J. treatment with disease-modifying anti-rheumatic (2002). Crystal structure of LIR-2 (ILT4) at 1.8 Å: drugs. Rheumatology (Oxford), 46, 742–751. differences from LIR-1 (ILT2) in regions implicated in 19. Tedla, N., Gibson, K., McNeil, H. P., Cosman, D., the binding of the human cytomegalovirus class I Borges, L. & Arm, J. P. (2002). The co-expression of MHC homolog UL18. BMC Struct. Biol. 2,6–15. activating and inhibitory leukocyte immunoglobulin- 6. Chapman, T. L., Heikema, A. P., West, A. P., Jr & like receptors in rheumatoid synovium. Am. J. Pathol. Bjorkman, P. J. (2000). Crystal structure and ligand 160, 425–431. binding properties of the D1D2 region of the inhibitory 20. Allen, R. L., Raine, T., Haude, A., Trowsdale, J. & receptor LIR-1 (ILT2). Immunity, 13, 727–736. Wilson, M. J. (2001). Leukocyte receptor complex- 7. Shiroishi, M., Tsumoto, K., Amano, K., Shirakihara, Y., encoded immunomodulatory receptors show diffe- Colonna, M., Braud, V. M. et al. (2003). Human ring specificity for alternative HLA-B27 structures. inhibitory receptors Ig-like transcript 2 (ILT2) and J. Immunol. 167, 5543–5547. ILT4 compete with CD8 for MHC class I binding and 21. Liu, Y.& Eisenberg, D. (2002). 3D domain swapping: as bind preferentially to HLA-G. Proc. Natl Acad. Sci. domains continue to swap. Protein Sci. 11, 1285–1299. USA, 100, 8856–8861. 22. Bazan, J. F. (1990). Structural design and molecular 8. Young, N. T., Waller, E. C., Patel, R., Roghanian, A., evolution of a cytokine receptor superfamily. Proc. Austyn, J. M. & Trowsdale, J. (2008). The inhibitory Natl Acad. Sci. USA, 87, 6934–6938. Structure of Leukocyte Ig-like Receptor A2 853

23. Foster, C. E., Colonna, M. & Sun, P. D. (2003). Crystal 33. Srimathi, T., Robbins, S. L., Dubas, R. L., Hasegawa, structure of the human natural killer (NK) cell acti- M., Inohara, N. & Park, Y. C. (2008). Monomer/dimer vating receptor NKp46 reveals structural relationship transition of the caspase-recruitment domain of to other leukocyte receptor complex immunoreceptors. human Nod1. Biochemistry, 47, 1319–1325. J. Biol. Chem. 278, 46081–46086. 34. Chattopadhyay, K., Ramagopal, U. A., Brenowitz, 24. Rousseau, F., Schymkowitz, J. W., Wilkinson, H. R. M.,Nathenson,S.G.&Almo,S.C.(2008). & Itzhaki, L. S. (2001). Three-dimensional domain Evolution of GITRL immune function: murine swapping in p13suc1 occurs in the unfolded state GITRL exhibits unique structural and biochemical and is controlled by conserved proline residues. properties within the TNF superfamily. Proc. Natl Proc. Natl Acad. Sci. USA, 98, 5596–5601. Acad. Sci. USA, 105, 635–640. 25. Ding, K., Louis, J. M. & Gronenborn, A. M. (2004). 35. Nakajima, H., Asai, A., Okada, A., Ping, L., Hamajima, Insights into conformation and dynamics of protein F., Sata, T. & Isobe, K. (2003). Transcriptional regulation GB1 during folding and unfolding by NMR. J. Mol. of ILT family receptors. J. Immunol. 171, 6611–6620. Biol. 335, 1299–1307. 36. Janket, M. L., DeRicco, J. S., Borowski, L. & Ayyavoo, 26. Borges, L., Hsu, M. L., Fanger, N., Kubin, M. & V. (2007). Human immunodeficiency virus (HIV-1) Cosman, D. (1997). A family of human lymphoid and Vpr induced downregulation of NHE1 induces myeloid Ig-like receptors, some of which bind to MHC alteration in intracellular pH and loss of ERM complex class I molecules. J. Immunol. 159, 5192–5196. in target cells. Virus Res. 126,76–85. 27. Bennett, M. J., Choe, S. & Eisenberg, D. (1994). 37. Otwinowski, Z. & Minor, W. (1997). Processing of X- Domain swapping: entangling alliances between ray diffraction data collected in oscillation mode. proteins. Proc. Natl Acad. Sci. USA, 91, 3127–3131. Methods Enzymol. 276, 307–326. 28. Bennett, M. J., Schlunegger, M. P. & Eisenberg, D. 38. Navaza, J. (1994). AMORE – an automated package for (1995). 3D domain swapping: a mechanism for molecular replacement. Acta Crystallogr. A, 50,157–163. oligomer assembly. Protein Sci. 4, 2455–2468. 39. Emsley, P. & Cowtan, K. (2004). Coot: model-building 29. Sanejouand, Y. H. (2004). Domain swapping of CD4 tools for molecular graphics. Acta Crystallogr. D, 60, upon dimerization. Proteins: Struct. Funct. Genet. 57, 2126–2132. 205–212. 40. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, 30. Murray, A. J., Lewis, S. J., Barclay, A. N. & Brady, R. L. W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). (1995). One sequence, two folds: a metastable struc- Crystallography & NMR system: A new software ture of CD2. Proc. Natl Acad. Sci. USA, 92, 7337–7341. suite for macromolecular structure determination. 31. Denisov, A. Y., Sprules, T., Fraser, J., Kozlov, G. & Acta Crystallogr. D, 54, 905–921. Gehring, K. (2007). Heat-induced dimerization of 41. Collaborative Computing Project Number 4. (1994). BCL-xL through alpha-helix swapping. Biochemistry, The CCP4 suite: programs for protein crystallography. 46, 734–740. Acta Crystallogr. D, 50, 760–763. 32. O'Neill, J. W., Manion, M. K., Maguire, B. & 42. Lovell, S. C., Davis, I. W., Arendall, W. B., III, de Hockenbery, D. M. (2006). BCL-XL dimerization by Bakker, P. I., Word, J. M., Prisant, M. G. et al. (2003). three-dimensional domain swapping. J. Mol. Biol. 356, Structure validation by Calpha geometry: phi,psi and 367–381. Cbeta deviation. Proteins, 50, 437–450.