- Email: [email protected]
SIRPα and Its Ligand CD47
doi:10.1016/j.jmb.2007.10.085
J. Mol. Biol. (2008) 375, 650–660
Available online at www.sciencedirect.com
Structural Insight into the Specific Interaction between Murine SHPS-1/SIRPα and Its Ligand CD47 Aki Nakaishi 1 , Mayumi Hirose 1 , Masato Yoshimura 1 , Chitose Oneyama 2 , Kazunobu Saito 2 , Nobuharu Kuki 1 , Makoto Matsuda 1 , Nakayuki Honma 3 , Hiroshi Ohnishi 4 , Takashi Matozaki 4 , Masato Okada 2 ⁎ and Atsushi Nakagawa 1 1
Laboratory of Supramolecular Crystallography, Research Center for Structural and Functional Proteomics, Institute for Protein Research, Osaka University, Suita, Osaka 565-0871, Japan 2
Department of Oncogene Research, Research Institute for Microbial Diseases, Osaka University, 3-1 Yamadaoka, Suita, Osaka 565-0871, Japan 3 Pharmaceutical Research Laboratories, KIRIN Brewery Co., Ltd., Takasaki, Japan 4
Laboratory of Biosignal Sciences, Institute for Molecular and Cellular Regulation, Gunma University, Maebashi, Gunma 371-8512, Japan Received 12 June 2007; received in revised form 22 October 2007; accepted 29 October 2007 Available online 7 November 2007 Edited by J. Karn
SRC homology 2 domain-containing protein tyrosine phosphatase substrate 1 (SHPS-1 or SIRPα/BIT) is an immunoglobulin (Ig) superfamily transmembrane receptor and a member of the signal regulatory protein (SIRP) family involved in cell–cell interaction. SHPS-1 binds to its ligand CD47 to relay an inhibitory signal for cellular responses, whereas SIRPβ, an activating member of the same family, does not bind to CD47 despite sharing a highly homologous ligand-binding domain with SHPS-1. To address the molecular basis for specific CD47 recognition by SHPS-1, we present the crystal structure of the ligand-binding domain of murine SHPS1 (mSHPS-1). Folding topology revealed that mSHPS-1 adopts an I2-set Ig fold, but its overall structure resembles IgV domains of antigen receptors, although it has an extended loop structure (C′E loop), which forms a dimer interface in the crystal. Site-directed mutagenesis studies of mSHPS-1 identified critical residues for CD47 binding including sites in the C′E loop and regions corresponding to complementarity-determining regions of antigen receptors. The structural and functional features of mSHPS-1 are consistent with the human SHPS-1 structure except that human SHPS-1 has an additional β-strand D. These results suggest that the variable complementarity-determining region-like loop structures in the binding surface of SHPS-1 are generally required for ligand recognition in a manner similar to that of antigen receptors, which may explain the diverse ligand-binding specificities of SIRP family receptors. © 2007 Elsevier Ltd. All rights reserved.
Keywords: SHPS-1; SIRP family; immunogloblin superfamily; CD47; complementarity-determining regions
*Corresponding author. E-mail address: [email protected]. Abbreviations used: SHPS-1, SRC homology 2 domain-containing protein tyrosine phosphatase substrate 1; Ig, immunoglobulin; SIRP, signal regulatory protein; mSHPS-1, murine SHPS-1; CDR, complementarity-determining region; ITIM, immunoreceptor tyrosine-based inhibition motif; ITAM, immunoreceptor tyrosine-based activation motif; LBD, ligand-binding domain; rSHPS-1, rat SHPS-1; mSIRPβ, mouse SIRPβ; PBS, phosphate-buffered saline; DMEM, Dulbecco's modified Eagle's medium.
Introduction Cell–cell interaction, mediated by specific cell surface receptors, is crucial to support tissue integrity and intercellular communication in multicellular organisms. SRC homology 2 domain-containing protein tyrosine phosphatase substrate 1 (SHPS-1),1 also known as SIRPα2 or BIT,3 is a member of the immunoglobulin (Ig) superfamily involved in cell–cell communications. SHPS-1 contains an extracellular region composed of three Ig domains with multiple
0022-2836/$ - see front matter © 2007 Elsevier Ltd. All rights reserved.
651
Structure of SHPS-1 Ligand-binding Domain
N-linked glycosylation sites and a cytoplasmic region with two immunoreceptor tyrosine-based inhibition motifs (ITIMs). It is abundantly expressed in leukocytes and neural cells and plays a crucial role in various immune functions, including phagocytosis,4–6 cell motility7–9 and T cell function10–12 as well as additional functions in neural cells.13,14 SHPS-1 has also been implicated in the regulation of cancer progression.15 SHPS-1 belongs to the signal regulatory protein (SIRP) family, whose members all contain an N-terminal extracellular region, a single transmembrane domain and a C-terminal intracellular region.16 Based on the primary structure of its cytoplasmic region and its potential role in signal transduction, SIRP proteins are classified into three subfamilies, SIRPα (SHPS-1), SIRPβ and SIRPγ. SIRPα has an ITIM in its cytoplasmic region and, upon phosphorylation, relays inhibitory signals that regulate cell responses by recruiting tyrosine phosphatases such as SHP-1/2.1,17,18 In contrast, both SIRPβ and γ have minimal intracellular tails. SIRPβ1, an isoform of SIRPβ identified in humans and mice, binds to the immunoreceptor tyrosine-based activation motif (ITAM)-containing adaptor protein DAP12 and induces a positive regulatory signal by recruiting the cytoplasmic tyrosine kinases ZAP-70 and Syk.19,20 The signaling mechanism mediated by SIRPγ remains elusive. Based on these differential signaling pathways downstream of SHPS-1 and SIRPβ, the SIRP family proteins have been classified as inhibitory/activating paired receptors.11,16 CD47 (or IAP) is another cell surface Ig superfamily protein that serves as a ligand for SHPS-1.21–24 The SHPS-1–CD47 interaction has been confirmed by direct protein binding and cell adhesion assays,7 and the membrane distal Ig domain of SHPS-1 has been identified as the CD47 binding domain.21,25 In contrast, SIRPβ does not bind to CD47, despite sharing a highly homologous Ig domain with SHPS-1 (∼90% identity). It has been shown that SHPS-1 and SIRPβ1 are coexpressed on the same cell surfaces of granulocyte and mononuclear cells but mediate opposite cell signaling responses26, suggesting that these paired receptors must precisely recognize their specific binding partners. The specificity of the SHPS-1– CD47 interaction has been addressed by examining the interactions of various combinations of SHPS-1 and CD47 from distinct species,27 which demonstrated that SHPS-1–CD47 interaction is highly species-specific. To elucidate the molecular mechanism for the specific SHPS-1–CD47 interaction, candidate amino acid residues required for the human SHPS-1–CD47 interaction have been determined by mutagenesis analyses.28 Very recently, the potential CD47 binding surface on the human SHPS-1 Ig domain has been identified on the basis of its crystal structure.29 However, to generalize this observation and to address the molecular basis for the species-specific SHPS-1–CD47 interaction, comparative analysis of SHPS-1 structure from various species would be required. Here we present the crystal structure of the ligand-binding domain (LBD) of murine SHPS-1
(mSHPS-1). The crystal structure reveals that the Ig domain of mSHPS-1 has a striking similarity to the IgV domains of T-cell receptor and Igs, although it is topologically classified as an I2 set of Ig fold. Furthermore, site-directed mutagenesis studies identified two essential residues involved in the specific binding of mSHPS-1 to mCD47. These residues reside in the unique loop structure and the regions corresponding to the complementaritydetermining regions (CDRs) of antigen receptor. Together with the comparative analysis to the tentative rat SHPS-1 structure and the recently solved structure of human SHPS-129 and SIRPβ, our structure-based studies suggest that the binding surface of SHPS-1 is composed of variable CDR-like loop structures in a manner similarly observed in antigen receptors.
Results Molecular structure of mouse SHPS-1 The ligand-binding domain of mSHPS-1 (mSHPS1 LBD) was expressed and purified from bacteria according to the methods used for the preparation of rat SHPS-1 (rSHPS-1).31 The crystal structure of mSHPS-1 was determined at 1.4-Å resolution (Table 1) based on the tentative structure of rSHPS-131, which was solved by a molecular replacement search using the V-set Ig domain structure of the anti-testosterone Fab fragment [Protein Data Bank (PDB) code 1L7T] as an initial model. mSHPS-1 LBD contains an Ig fold characteristic of the Ig superfamily, identified by a two-layer β-sandwich structure (Fig. 1a and c). In addition to the β strands, mSHPS-1 contains a 310 helix at the periphery of the E–F loop. The core packing residues of Ig fold are also functional in mSHPS-1: a canonical disulfide bond (Cys24 and Cys90) packs against the consensus Table 1. Data collection and refinement statistics Data collection Space group Unit cell parameters (Å) Resolution (Å)a Observed reflectionsa Unique reflectionsa Rmergea,b 〈I/σ(I)〉a Completeness (%)a Multiplicitya Refinement Rcrystc Rfreec rmsd Bonds (Å) Angles (°) a
P212121 a = 48.6, b = 57.7, c = 69.2 39.75–1.40 (1.40–1.48) 281,790 (40,529) 39,022 (5624) 0.055 (0.340) 7.3 (2.1) 100 (100) 7.2 (7.2) 0.174 0.199 0.022 2.04
The highest resolution shell statistics are shown in parentheses. Rmerge = ∑hk∑| i Ii(hkl) − 〈I(hkl)〉|/∑hkl∑iIi(hkl). R cryst and R free = ∑| |F obs|−|F calc||/∑|Fobs|; R free was calculated for 5% of reflections randomly chosen for cross validation. b c
652 Trp residue (Trp37) (Fig. 1a). The mSHPS-1 LBD framework exhibits a folding topology that resembles both V- and I2-set folds.32,33 Similar to these Ig folds, mSHPS-1 has a typical kink between the A and A′ strands mediated by a cis-proline (Pro8). This allows the A and A′ strands to hydrogen-bond to the B and G strands, respectively. The strand composition of mSHPS-1 (ABCEGF) is categorized as an I2-set Ig domain, whereas mSHPS-1 has bulges of C′ and G strands characteristic of V-set Ig domains. Furthermore, the B strand breaks into shorter strands mediated by Asn23, a potential glycosylation site, and the C′E loop is atypically arranged. In the V-set Ig domains, the regions corresponding to the C′E loop are integrated into the β-sandwich structure via the C″ and D strands, while the C′E loop of mSHPS-1 is unpacked and protrudes from the core β-sandwich structure (Fig. 1c). rSHPS-1 LBD has a structure quite similar to that of mSHPS-1 (rmsd 0.90 Å), except for substantial deviations in the C′E loop structure (Fig. 1b). Considering that the amino acid sequences in the C′E loop are almost identical between the two species (Fig. 5), it seems that the C′E loop structure is structurally flexible or variable. In the crystal of mSHPS-1 LBD, there are two monomers in the asymmetric unit, with a low
Structure of SHPS-1 Ligand-binding Domain
rmsd of 0.75 Å between equivalent Cα atoms (Fig. 2a and b). The C′E loop contributes to the formation of binding surfaces that are stabilized by hydrogen bonds between the two molecules (Fig. 2a, blue box). Similarly, rSHPS-1 LBD can also form a dimer via the C′E loop interface, although its mode of interaction is apparently different, reflecting the unique conformations of the C′E loop in the two species (Fig. 2c). Notably, the rmsd value of the C′E loop between rSHPS-1 LBD dimers (2.7 Å) is significantly higher than the resting β-sandwich structure (0.35 Å) (Fig. 2d), supporting the flexible nature of the C′E loop in rSHPS-1. However, gel filtration chromatography analysis revealed that the major population of either mSHPS-1 or rSHPS-1 LBD eluted as a single peak of ∼ 13 kDa, suggesting that these molecules exist as a monomer in solution (data not shown). The monomeric feature of SHPS-1 LBD was further confirmed by sedimentation equilibrium experiments (Fig. 2e). Fitting data to a single-component model provided a theoretically monomeric molecular weight for mSHPS-1 LBD and rSHPS-1 LBD. In addition, the dimer is antiparallel with the C terminus of each monomer pointing in opposite directions. This configuration is unlikely to be formed by a transmembrane protein on a single cell surface. Taken together, it is likely
Fig. 1. Structure of mSHPS-1 LBD. (a) Ribbon diagram of the mSHPS-1 LBD monomer. The front (cyan) and back (light blue) sheets are composed of A′GGFCC′ and ABBE strands, respectively. The disulfide bond is shown in yellow. Locations of Pro8, Cys24, Trp37 and Cys90 are indicated by red dots. (b) Superimposed mSHPS-1 (cyan) on rSHPS-1 (pink) structures. Regions corresponding to the C′E loop are indicated. (c) Folding topologies of mSHPS-1 LBD (I2 set) and T-cell receptor Vα (V set).
Structure of SHPS-1 Ligand-binding Domain
653
Fig. 2. (a) The configuration of the two monomers in the asymmetric unit of mSHPS-1 LBD. The diagram to the right is a higher magnification of the region boxed in blue. Hydrogen bonds from Ala65 are indicated by broken red lines. (b) The two monomers of mSHPS-1 LBD are superimposed. (c) The configuration of the two monomers in the asymmetric unit of rSHPS-1 LBD. (d) The monomers of rSHPS-1 LBD are superimposed. (e) Sedimentation equilibrium analysis of mSHPS-1 LBD. Open circles represent data collected at 25,000 rpm (lower diagram). The curve represents the best fit to a singlecomponent sedimentation model, yielding an estimated molecular mass of 13.7 kDa. The residuals of the corresponding points are randomly distributed, indicating that the sample consists of only monomers (upper diagram).
that SHPS-1 LBD is present as a monomer in solution, although it is possible that the full-length SHPS-1 could form a dimer under certain physiological conditions. Structural comparison of SHPS-1 with T-cell receptor A structure-based homology search of the PDB by the DALI program revealed that mSHPS-1 LBD is highly homologous to V-set Ig proteins rather than I2-set proteins. T-cell receptor Vα (PDB code 1TCR) shows the highest similarity to mSHPS-1 LBD (rmsd 2.6 Å for 96 equivalent Cα atoms), despite only 28% amino acid sequence identity. Several other V-set Ig proteins, including Igs (PDB code 1MFA; rmsd 2.1 Å), sialoadhesion (PDB code 1QFO;
rmsd 2.2 Å) and NKp44 (PDB code 1HKF; rmsd 2.4 Å), are also significantly similar to mSHPS-1. The core secondary structure elements of mSHPS-1 LBD are tightly superimposed with those of T-cell receptor Vα (Fig. 3a). The major structural differences between mSHPS-1 and T-cell receptor are present in the loops of B′C, C′E and FG, which correspond to the complementarity-determining regions CDR1, CDR2 and CDR3 of T-cell receptor, respectively (Fig. 1c and Fig. 3a). Furthermore, electrostatic potential analysis revealed that the mSHPS-1 LBD includes a number of solventaccessible hydrophobic residues surrounding the regions corresponding to the CDR loops (Fig. 3b). These structural features suggest that SHPS-1 contains the binding surface similar to those of antigen receptor Ig superfamily proteins.34,35
654
Structure of SHPS-1 Ligand-binding Domain
Fig. 3. Comparison of mSHPS-1 and T-cell receptor. (a) mSHPS-1 LBD (cyan) and T-cell receptor Vα (yellow) structure (PDB code 1TCR) are superimposed. Regions corresponding to CDR1, CDR2 and CDR3 in T-cell receptor are marked in blue, green and red, respectively. (b) Electrostatic potential surfaces of mSHPS-1 LBD and T-cell receptor Vα. Potential binding surfaces are shown in rotated views (45°). Negative and positive potentials are colored in red and blue, respectively.
Surface plasmon resonance assay for the mSHPS-1–mCD47 interaction In order to locate the ligand-binding surface on mSHPS-1, we first confirmed the direct interaction between the recombinant mSHPS-1 LBD and the mCD47-Fc fusion protein by surface plasmon resonance (SPR) assay (Fig. 4). Interaction was detected at 25 °C as response units reached equilibrium in a dose-dependent manner (Fig. 4a). These analytical data reveal that recombinant mSHPS-1 LBD can specifically bind to the mCD47-Fc fusion protein. The equilibrium dissociation constant (Kd) of the mSHPS-1–mCD47 interaction was calculated from a dose–response curve (Fig. 4b and c). Nonlinear Langmuir fitting of the binding data gave a Kd of ∼ 1.3 μM, which is almost equivalent to the Kd values reported for the human SHPS-1 and CD47 interaction observed at 37 °C (∼ 2 μM).36 Cross-species interaction between rat SHPS-1 and mouse CD47 was detected with a Kd value of ∼1.5 μM, but a significant interaction of mouse SIRPβ (mSIRPβ) with mCD47 was not detected under our assay conditions (Fig. 4c). Mutagenesis analysis of the CD47 binding sites To determine the amino acid residues responsible for mSHPS-1–mCD47 binding, we prepared a series of mSHPS-1 mutants on the basis of the crystal structure and the amino acid sequences (Table 2 and Fig. 5). We focused on the residues in mSHPS-1 that are conserved in rat SHPS-1 but not in the mSIRPβ. Since nonconserved residues between mSHPS-1 and mSIRPβ accumulate within the N-terminal 30 residues, we prepared chimera proteins consisting of the N-terminal mSIRPβ and the C-terminal mSHPS-1 (SIRPβ1–14 SHPS-115–115 and SIRPβ1–26 SHPS-127–115). Furthermore, we made single or multiple substitutions of mSHPS-1 residues with mSIRPβ
residues. The substitute residues of mSHPS-1 include Lys35 at the C-terminal end of the B′C loop (corresponding to CDR1); Ser50, Phe56, Val59 and Ala60 in the C″E loop (CDR2); Glu83 and Ala85 in the 310 module; Lys95 and Thr102 in the FG loop (CDR3); and Tyr115 near the C terminus (Fig. 5). These mutant proteins were produced in Escherichia coli, and the folding status of each purified protein was confirmed by CD spectroscopy analysis (data not shown). Most of the mutations did not significantly affect protein folding, but the substitution of Phe56 to Ser induced significant deviations in the structure, implicating a potential role for Phe56 in maintaining the overall structure. The binding affinities of these mutants to the mCD47-Fc fusion protein were estimated by SPR (Table 2). The Kd values of SIRPβ1–14 SHPS-115–115, SIRPβ1–26 SHPS-127–115, K36R, S50P, V59I, E83A, A85E, E83A/A85S, P97S, and Y115L were in a range close to that of wild-type mSHPS-1 (1∼2 μM), suggesting that these residues are not essential for the specific interaction between mSHPS-1 and mCD47. The T102I mutant showed a slightly higher affinity for mCD47-Fc (∼ 0.7 μM) than the wild-type molecule, indicating some inhibitory role of the Thr102 residue. In contrast, F56S, S50P/F56S, A65V, K95R, K95R/P97S and K95R/P97S/T102I showed significantly lower affinities with Kd values ranging from 3.42 to 18.0. Particularly, an Ala65Val substitution greatly affected the mSHPS-1–mCD47 binding. Cell adhesion assays of the full-length mSHPS-1 mutants The results of the SPR assays suggest that Phe56, Ala65, Lys95 and Thr102 are potentially required for the mSHPS-1–mCD47 binding specificity. To further evaluate the role of these residues in vivo, we expressed full-length mSHPS-1 having the substitutions of these residues in CHO-Ras cells and ana-
655
Structure of SHPS-1 Ligand-binding Domain
Discussion
Fig. 4. SPR assay of the SHPS-1–CD47 interaction. (a) Binding of mSHPS-1 LBD to immobilized mCD47-Fc fusion protein was detected in response units according to the protocol provided by the manufacturer (see Materials and Methods). Background responses obtained for the immobilized rat P-selectin-Fc fusion protein were subtracted from the total response. A diagram of the data obtained for a concentration gradient series of mSHPS-1 LBD is shown. (b) Dose response curve for the m SHPS-1– CD47 interaction. (c) Nonlinear curve fitting of the data using a 1:1 Langmuir binding isotherm yielded a Kd of 1.31 ± 0.27 μM for mSHPS-1 LBD. Similarly, the rate constant for rSHPS-1 was calculated to be 1.46 ± 0.06. However, mSIRPβ did not have a significant binding capacity for mCD47. N.S., not significant.
lyzed their binding abilities to mCD47 (Fig. 6a). Cell binding assays on immobilized mCD47-Fc fusion protein showed that the cells expressing mSHPS-1 mutants, F56S, A65V, K95R, S50P/F56S and K95R/P97S/T102I, were less adhesive compared to cells expressing wild type, indicating that Phe56, Ala65 and Lys95 were critical for binding (Fig. 6b). Similar findings were obtained when the cell adhesion assay was performed on CHO-Ras cells expressing full-length mCD47 (Fig. 6c). These results demonstrate that Phe56, Ala65 and Lys95 residues play critical roles in the mSHPS-1–mCD47 binding in vivo.
Our crystallographic analysis of mouse and rat SHPS-1 LBD has provided basic information required to understand the ligand, species and/or strain specificity of SIRP family receptors. The structure-based homology search revealed a striking homology between mSHPS-1 LBD and IgV domains of antigen receptors such as T-cell receptor and Igs, except that mSHPS-1 LBD has a unique loop structure (C′E loop), which forms a dimer interface in the crystal. These structural features are well consistent with the recently presented human SHPS1 LBD structure,29 confirming that SHPS-1 is evolutionarily and functionally a close relative of the antigen receptor-type Ig superfamily.16 Site-directed mutagenesis studies uncovered the mCD47-binding surface of mSHPS-1. We identified three residues, Phe56, Ala65 and Lys95, which are critical for specific recognition of mCD47 (Fig. 7a and b). Ala65 is located in the middle of the C′E loop that forms dimer interface in the crystal. In the mSHPS-1 dimer, Ala65 makes stable hydrogen bonds to Ser75 and Arg68 residues of the other molecule (Fig. 2a). However, the monomeric feature of mSHPS-1 LBD in solution suggests that dimer formation is potentially due to the flexibility and protein-binding ability of the C′E loop. Although it is currently unclear whether Ala65 directly interacts with the ligand or contributes to maintaining the conformation of mSHPS-1, it is likely that the C′E loop region containing Ala65 is involved in ligand recognition. However, it is known that the ligandbinding domain of SHPS-1 has substantial sequence diversity even among mouse strains and that Ala65 is a hot spot of polymorphism.37 The 129/sv strain type of mSHPS-1 has Ala65, whereas Ala65 is replaced by a Thr in Balb/c and C57BL/6 strains. On the other hand, mSIRPβ in C57BL/6 strain has a Val in this position, although Balb/c strain has a Thr. In the human SIRP family, SIRPα/SHPS-1 has a Ser at this site, and is replaced by a Leu in SIRPβ and Table 2. Effects of site-directed mutations on the mSHPS1–mCD47 binding mSHPS-1 mutants SIRPβ1–14 SHPS-115–115 SIRPβ1–26 SHPS-127–115 K36R S50P F56S S50P/F56S V59I A65V E83A A85E E83A/A85S K95R P97S K95R/P97S T102I Y115L K95R/P97S/T102I
Kd (μM) 1.39 ± 0.11 1.82 ± 0.06 1.18 ± 0.04 1.76 ± 0.12 4.16 ± 0.11 10.7 ± 0.80 1.00 ± 0.15 18.0 ± 0.40 1.17 ± 0.12 1.29 ± 0.03 1.11 ± 0.01 5.29 ± 0.98 1.00 ± 0.04 3.42 ± 0.16 0.76 ± 0.05 1.21 ± 0.02 3.62 ± 0.47
656
Structure of SHPS-1 Ligand-binding Domain
Fig. 5. Structure-based sequence alignment of mSHPS-1, rSHPS-1, SIRPβ and hSHPS-1. Numbering is for mSHPS-1 LBD protein used for structural analysis in this study. In mSHPS-1, rSHPS-1 and SIRPβ, residues replaced in the mutagenesis study are highlighted. Significant and irrelevant residues are boxed in red and green, respectively. In hSHPS-1, residues shown to be critical for ligand binding are boxed in orange.29 Locations of the β-strands are indicated by bold arrows above the alignment. The D strand unique to hSHPS-1 is indicated by a bold arrow under the alignment. The 310 module is also indicated by a small cylinder. Regions corresponding to CDR1, CDR2 and CDR3 in the antigen receptors are shown by blue, green and red boxes, respectively.
SIRPγ. The site-directed mutagenesis of Ser to Asp in this position of human SIRPα/SHPS-1 destroyed ligand-binding activity, confirming the critical role of this site.29 Thus, it is likely that the residue equivalent to Ala65 of mSHPS-1 is critical in determining the species, strain and/or isoform specificity in ligand recognition. Another critical residue identified by our mutagenesis analysis is Lys95, which is exposed on the molecular surface in close proximity to CDR3 (Fig. 7a and b). The substitution of Lys95 with an Arg in mSHPS-1 substantially abated the mCD47 binding. mSIRPβ that does not bind to mCD47 has an Arg residue at this site in all mouse strains examined. The Lys95 residue is also conserved in the human SHPS-1, and mutagenesis at this site destroyed its ligand-binding activity.29 However, mSHPS-1 in Balb/c strain has an Arg residue at the same site, suggesting that Lys residue is not essential for CD47 binding or that the CD47 binding activity of mSHPS-1 in Balb/c strain may be perturbed. Although additional analysis is required to elucidate its precise roles, it seems that this site is critical for constructing the species/strain-specific binding surface, as is the case for Ala65. The substitution of Phe56 to a Ser also reduced the mSHPS-1–mCD47 binding affinity, but circular dichroism spectroscopy analysis showed that the substitution of Phe56 significantly altered the folding status of the recombinant protein, indicating that Phe56 is required for proper protein folding of recombinant mSHPS-1. Both Ala65 and Lys95 are positioned in the ligandbinding surface, which contains CDR-like loop structures. Several residues critical for ligand binding were also mapped onto similar loops of human SHPS-1.29 These findings support the theory that SHPS-1 uses these CDR-like loop structures for ligand binding, rather than the faces of β-sheet that are often utilized by other Ig superfamily receptors such as CD238 and B7.39 Given that CDR-like regions are hot spots for polymorphism,2,37 the use of these regions for ligand recognition allows SIRP family receptors to acquire diverse ligand-binding specificities by only small changes in amino acids. The CDR1 and CDR3 loops in SHPS-1 are well
conserved among different species, while the loops equivalent to CDR2 (C′E loop of rodent SHPS-1 and DE loop of human SHPS-1) are significantly divergent; human SHPS-1 has an additional D strand29 (Fig. 7c). These observations suggest that deviations in the CDR2-like regions would be important for species specificity of SHPS-1. Recently an NMR structure of the membrane distal Ig domain of human SIRPβ became available (PDB code 2D9C). The NMR structure reveals the striking similarity in the backbone structures of SHPS-1 and SIRPβ (rmsd 1.2 Å) (Fig. 7d), which further supports the above hypothesis that a small modulation of the binding loops are sufficient for the generation of diverse ligand specificities in the SIRP family receptors. The SIRP gene family is thought to have evolved rapidly, with variable numbers of genes in different species and distinct genes, such as SIRPα/β/γ, that have evolved by gene duplication.16 Polymorphisms have been frequently identified in the SIRP gene family. For example, SIRPβ proteins from two mouse strains have 22 amino acid differences,40 and SHPS-1 accumulates polymorphisms in its ligandbinding domain.2,37 Additional genes related to the SIRP family have also been identified.41 The fulllength extracellular domain of SHPS-1 consists of a membrane-distal Ig domain that structurally resembles the V-set Ig fold of antigen receptors, a joining (J)-like region that is present in the putative antigen receptor predecessor, and two additional Ig domains that have the highest sequence similarity to the C1-set Ig domains characteristic of antigen receptors. Based on these structural features, it has been suggested that the SIRP family may have originated from a precursor of the rearranging antigen receptors.16 The speed of evolution may have allowed SHPS-1 to acquire functional diversity and species- and/or tissue-specific cell–cell recognition systems. Considering the implications of SHPS-1– CD47 signaling in innate immune response, autoimmune disease, hemolytic anemia and tumor progression, our structure-based functional analysis of SHPS-1 provides essential information for the development of therapeutic agents that could target SHPS-1 and/or CD47.
657
Structure of SHPS-1 Ligand-binding Domain
Materials and Methods Protein expression, purification and crystallization The ligand-binding domain of mouse SHPS-1 (mSHPS-1 LBD), containing residues 31–147, was subcloned into the BamHI and NdeI restriction sites of the bacterial expression vector pGEX-6P-1 (GE Healthcare) and transformed into the E. coli strain Rosetta-gami B (Novagen). The expressed soluble proteins were purified as described previously.31 Crystallization of mSHPS-1 LBD was performed by hanging-drop vapor diffusion in 0.1 M Pipes, pH 6.5, and 32% polyethylene glycol 4000. The crystal of rat SHPS-1 (rSHPS-1) LBD was obtained as described.31
Structure determination and refinement Diffraction data sets of rSHPS-1 at 2.8 Å and mSHPS-1 at 1.4 Å were collected on the beam line BL44XU at SPring-8 (Hyogo, Japan) and processed using MOSFLM42 and SCALA.43 The initial structure of rSHPS-1 was solved by six-dimensional molecular replacement search with Monte Carlo method,30 using the V-set Ig domain structure of the anti-testosterone Fab fragment (PDB code 1L7T) as an initial model. The rSHPS-1 structure was refined to an R value of 0.28 and an Rfree value of 0.36 at 2.8-Å resolution by REFMAC5 but the refinement is still in progress (unpublished result). A molecular replacement solution was found with the program MOLREP44 using the current rSHPS-1 structure as a search model. The structure of mSHSP-1 was refined using REFMAC543,45 and ARP/wARP46 combined with manual model rebuilding with COOT.47 The stereochemical properties of the structure were assessed by PROCHECK,48 which showed no residues in the disallowed regions on the Ramachandran plot. Diffraction data collection and refinement statistics of mSHPS-1 are listed in Table 1. Analytical ultracentrifugation Sedimentation equilibrium experiments were performed at 20 °C in an analytical ultracentrifuge using an An-60 Ti rotor and 1.2-cm six-channel charcoal-Epon centerpieces (Beckman Coulter). mSHPS-1 LBD was concentrated and exchanged into 10 mM Hepes, pH 7.4, 1 mM ethylenediaminetetraacetic acid (EDTA) and 150 mM NaCl. Data were collected at protein concentrations of 54, 31, and 18 μM (0.70, 0.47, and 0.23 mg/ml, respectively). Centrifugation was performed at 25,000 rpm, and interference scans were taken after 20 h when equilibrium was accomplished. Data analysis was performed using Beckman analytical software with the protein partial specific volume of 0.7391 cm3g− 1 and the buffer density 1.005 cm3g− 1. Site-directed mutagenesis All site-directed mutagenesis of mSHPS-1 were performed using two polymerase chain reactions. In the first
Fig. 6. The effect of mSHPS-1 mutations on cell adhesion mediated by mSHPS-1–mCD47 binding. (a) Expression of various mSHPS-1 mutants and mSIRPβ in CHORas cells. CHO-Ras cells transfected with the expression plasmids for wild-type or mutant mSHPS-1 and mSIRPβ were lysed and subjected to Western blot analysis using an anti-myc antibody. Actin was detected with an anti-actin antibody as an internal control. (b) The binding of mSHPS1 transfectants to mCD47-Fc fusion protein. Cell Tracker Green CMFDA-labeled mSHPS-1 transfectants were added to the wells coated with mCD47-Fc fusion protein. After washing, bound cells were measured by fluorescence. The data represent the means and standard deviations of four independent experiments. (c) The binding of mSHPS-1 transfectants to mCD47-transfected CHO monolayers. Cell Tracker Green CMFDA-labeled mSHPS-1 transfectants preincubated with (open bars) or without (filled bars) 50 μg/ml mCD47-Fc fusion protein were added to mCD47-transfected CHO monolayer cultures. The bound cells were measured by fluorescence. The data represent the means and the standard deviations of four independent experiments.
658
Structure of SHPS-1 Ligand-binding Domain
Fig. 7. The critical residues of mSHPS-1 required for mCD47 binding. (a) The positions of Phe56, Ala65 and Lys95 are shown on the ribbon diagram of mSHPS-1 LBD. (b) Phe56, Ala65 and Lys95 are marked in magenta. The residues whose mutations had no significant effect on ligand binding are indicated in gray. (c) LBDs of mSHPS-1 (cyan), rSHPS-1 (pink) and hSHPS-1 (yellow) are superimposed. (d) N-terminal Ig domains of hSHPS-1 (yellow) and hSIRPβ (gray) (PDB code 2D9C) are superimposed. reaction, the N-terminal and the C-terminal fragments were amplified separately, using the primers containing mutation sites. In the second reaction, these two amplified fragments were mixed together and amplified using the N-terminal forward and the C-terminal reverse primers. The PCR products were sequenced and then subcloned into pGEX-6P-1 (GE Healthcare) for bacterial expression, and into the retrovirus expression vector pCX4-myc. Proper folding of the bacterially expressed proteins were confirmed by circular dichroism by a JASCO J720WI spectropolarimeter (JASCO, Tokyo, Japan) with a spectral bandwidth of 1 nm at a sample concentration of 0.2 mg/ ml in phosphate-buffered saline (PBS). The expression levels of the full-length mutants of mSHPS-1 in COS cells were determined by Western blotting using an anti-myc antibody (MBL, Nagoya, Japan). SPR binding analysis SPR binding assays were carried out with a Biacore 2000 instrument (GE Healthcare) at 25 °C in HBS-EP running buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 3 mM EDTA and 0.005% surfactant P20). Mouse CD47-Fc fusion protein (mCD47-Fc) was purified from the culture medium of CHO-Ras cells stably expressing mCD47-Fc using Protein A Sepharose (GE Healthcare), and directly immobilized to the sensor surface of the experimental flow cell by amine coupling, which typically resulted in immobilization levels of 1000 RU. A sensor chip coupled to a rat P-selectin Ig fusion protein was used as a negative control for the binding assays and recorded levels of ∼ 1400 RU. For determining the binding affinity of the wild-type mSHPS-1 for mCD47-Fc, 11 serial dilutions of the mSHPS-1 were prepared in HBS-EP. Background responses from both the control flow cell and buffer alone were subtracted to obtain the final response
levels. Each experiment was repeated more than three times and the average was taken as the final result. Equilibrium dissociation constants were obtained by fitting the data using a 1:1 Langmuir binding model with the BIAevaluation software (GE Healthcare). The constants of mouse SHPS-1 mutants were measured using the same protocol. Retroviral expression and cell adhesion assays pCX4bsr and pCX4bsr-myc, two Moloney murine leukemia virus-based retrovirus vectors containing the blasticidin S resistance gene (bsr) as selectable markers were kindly provided by Dr. T. Akagi.49 pCX4-myc-SHPS1 and pCX4-CD47 were constructed by inserting wildtype or mutant full-length mouse SHPS-1 and wild-type full-length mouse CD47 into the multicloning site of pCX4bsr-myc and pCX4bsr, respectively. CHO cells stably expressing H-Ras-V12 (CHO-Ras cells), kindly provided by Dr. Akagi, were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum. CHO-Ras cells seeded in six-well plates at 5 × 105 cells per well were transfected with 10 μg of pCX4bsr-myc containing SHPS-1 using Lipofectamine LTX (Invitrogen). CHO-Ras cells stably expressing mCD47 were established by selection in the presence of blasticidin. To determine the ability of cell–cell adhesion mediated by mSHPS-1–mCD47 interaction, mCD47-transfected CHO-Ras cells were cultured on 96-well plates and fixed with 4% paraformaldehyde for 10 min. After the wells were washed with PBS, the cultures were incubated with 2% bovine serum albumin in PBS for 30 min to block nonspecific binding. Wild-type and mutant SHPS-1 transfectants were labeled with 5 mM Cell Tracker Green CMFDA (Invitrogen) and preincubated with or
659
Structure of SHPS-1 Ligand-binding Domain without 50 μg/ml mCD47-Fc fusion protein. The labeled mSHPS-1 transfectants (1 × 105 cells per well) were applied to the fixed mCD47-transfected monolayers and allowed to bind for 40 min at room temperature. The wells were then filled with DMEM containing 10% FCS, and the plate was inverted and placed for 5 min. After removal of the unbound cells by gentle aspiration, 1% NP-40 in PBS was added to each well and fluorescence was measured using 485- and 538-nm excitation and emission filter, respectively, in a Labosystems Fluoroscan II.
9.
10.
Protein Data Bank accession number The atomic coordinates of mSHPS-1 have been deposited in the Protein Data Bank (PDB code 2YZ1).
11.
12.
Acknowledgements We thank Drs A. Ogawa, T. Matsuura, and E. Yamashita for assistance with crystal data collection and analysis, M. Sakai for analytical ultracentrifugation experiments, T. Akagi for providing the pCX4 vector, and Y. Segawa and F. Akita for help with sample preparation. This work was supported by the National Project on Protein Structural and Functional Analyses from the Ministry of Education, Sports, Culture, Science and Technology of Japan.
References 1. Fujioka, Y., Matozaki, T., Noguchi, T., Iwamatsu, A., Yamao, T., Takahashi, N. et al. (1996). A novel membrane glycoprotein, SHPS-1, that binds the SH2domain-containing protein tyrosine phosphatase SHP2 in response to mitogens and cell adhesion. Mol. Cell. Biol. 16, 6887–6899. 2. Kharitonenkov, A., Chen, Z., Sures, I., Wang, H., Schilling, J. & Ullrich, A. (1997). A family of proteins that inhibit signalling through tyrosine kinase receptors. Nature, 386, 181–186. 3. Sano, S., Ohnishi, H., Omori, A., Hasegawa, J. & Kubota, M. (1997). BIT, an immune antigen receptorlike molecule in the brain. FEBS Lett. 411, 327–334. 4. Oldenborg, P. A., Zheleznyak, A., Fang, Y. F., Lagenaur, C. F., Gresham, H. D. & Lindberg, F. P. (2000). Role of CD47 as a marker of self on red blood cells. Science, 288, 2051–2054. 5. Gardai, S. J., McPhillips, K. A., Frasch, S. C., Janssen, W. J., Starefeldt, A., Murphy-Ullrich, J. E. et al. (2005). Cell-surface calreticulin initiates clearance of viable or apoptotic cells through trans-activation of LRP on the phagocyte. Cell, 123, 321–334. 6. Okazawa, H., Motegi, S., Ohyama, N., Ohnishi, H., Tomizawa, T., Kaneko, Y. et al. (2005). Negative regulation of phagocytosis in macrophages by the CD47SHPS-1 system. J. Immunol. 174, 2004–2011. 7. Liu, Y., Buhring, H. J., Zen, K., Burst, S. L., Schnell, F. J., Williams, I. R. & Parkos, C. A. (2002). Signal regulatory protein (SIRPalpha), a cellular ligand for CD47, regulates neutrophil transmigration. J. Biol. Chem. 277, 10028–10036. 8. Motegi, S., Okazawa, H., Ohnishi, H., Sato, R., Kaneko, Y., Kobayashi, H. et al. (2003). Role of the
13.
14.
15.
16. 17.
18.
19.
20.
21.
22.
23.
CD47–SHPS-1 system in regulation of cell migration. EMBO J. 22, 2634–2644. Inagaki, K., Yamao, T., Noguchi, T., Matozaki, T., Fukunaga, K., Takada, T. et al. (2000). SHPS-1 regulates integrin-mediated cytoskeletal reorganization and cell motility. EMBO J. 19, 6721–6731. Latour, S., Tanaka, H., Demeure, C., Mateo, V., Rubio, M., Brown, E. J. et al. (2001). Bidirectional negative regulation of human T and dendritic cells by CD47 and its cognate receptor signal-regulator proteinalpha: down-regulation of IL-12 responsiveness and inhibition of dendritic cell activation. J. Immunol. 167, 2547–2554. Seiffert, M., Brossart, P., Cant, C., Cella, M., Colonna, M., Brugger, W. et al. (2001). Signal-regulatory protein alpha (SIRPalpha) but not SIRPbeta is involved in T-cell activation, binds to CD47 with high affinity, and is expressed on immature CD34(+)CD38(−) hematopoietic cells. Blood, 97, 2741–2749. Piccio, L., Vermi, W., Boles, K. S., Fuchs, A., Strader, C. A., Facchetti, F. et al. (2005). Adhesion of human T cells to antigen-presenting cells through SIRPbeta2– CD47 interaction costimulates T-cell proliferation. Blood, 105, 2421–2427. Nakahata, Y., Okumura, N., Shima, T., Okada, M. & Nagai, K. (2000). Light-induced tyrosine phosphorylation of BIT in the rat suprachiasmatic nucleus. J. Neurochem. 74, 2436–2444. Murata, T., Ohnishi, H., Okazawa, H., Murata, Y., Kusakari, S., Hayashi, Y. et al. (2006). CD47 promotes neuronal development through Src- and FRG/Vav2mediated activation of Rac and Cdc42. J. Neurosci. 26, 12397–12407. Oshima, K., Machida, K., Ichigotani, Y., Nimura, Y., Shirafuji, N., Hamaguchi, M. & Matsuda, S. (2002). SHPS-1: a budding molecule against cancer dissemination. Cancer Res. 62, 3929–3933. Barclay, A. N. & Brown, M. H. (2006). The SIRP family of receptors and immune regulation. Nat. Rev. Immunol. 6, 457–464. Lienard, H., Bruhns, P., Malbec, O., Fridman, W. H. & Daeron, M. (1999). Signal regulatory proteins negatively regulate immunoreceptor-dependent cell activation. J. Biol. Chem. 274, 32493–32499. Veillette, A., Thibaudeau, E. & Latour, S. (1998). High expression of inhibitory receptor SHPS-1 and its association with protein-tyrosine phosphatase SHP- 1 in macrophages. J. Biol. Chem. 273, 22719–22728. Dietrich, J., Cella, M., Seiffert, M., Buhring, H. J. & Colonna, M. (2000). Cutting edge: signal-regulatory protein beta 1 is a DAP12-associated activating receptor expressed in myeloid cells. J. Immunol. 164, 9–12. Tomasello, E., Cant, C., Buhring, H. J., Vely, F., Andre, P., Seiffert, M. et al. (2000). Association of signal-regulatory proteins beta with KARAP/DAP-12. Eur. J. Immunol. 30, 2147–2156. Vernon-Wilson, E. F., Kee, W. J., Willis, A. C., Barclay, A. N., Simmons, D. L. & Brown, M. H. (2000). CD47 is a ligand for rat macrophage membrane signal regulatory protein SIRP (OX41) and human SIRPalpha 1. Eur. J. Immunol. 30, 2130–2137. Brown, E., Hooper, L., Ho, T. & Gresham, H. (1990). Integrin-associated protein: a 50-kD plasma membrane antigen physically and functionally associated with integrins. J. Cell. Biol. 111, 2785–2794. Lindberg, F. P., Gresham, H. D., Schwarz, E. & Brown, E. J. (1993). Molecular cloning of integrin-associated protein: an immunoglobulin family member with multiple membrane-spanning domains implicated in
660
24.
25.
26.
27.
28.
29.
30.
31.
32. 33. 34.
35.
alpha v beta 3-dependent ligand binding. J. Cell. Biol. 123, 485–496. Mawby, W. J., Holmes, C. H., Anstee, D. J., Spring, F. A. & Tanner, M. J. (1994). Isolation and characterization of CD47 glycoprotein: a multispanning membrane protein which is the same as integrin-associated protein (IAP) and the ovarian tumour marker OA3. Biochem. J. 304, 525–530. Liu, Y., O'Connor, M. B., Mandell, K. J., Zen, K., Ullrich, A., Buhring, H. J. & Parkos, C. A. (2004). Peptide-mediated inhibition of neutrophil transmigration by blocking CD47 interactions with signal regulatory protein alpha. J. Immunol. 172, 2578–2585. Liu, Y., Soto, I., Tong, Q., Chin, A., Buhring, H. J., Wu, T. et al. (2005). SIRPbeta1 is expressed as a disulfidelinked homodimer in leukocytes and positively regulates neutrophil transepithelial migration. J. Biol. Chem. 280, 36132–36140. Subramanian, S., Boder, E. T. & Discher, D. E. (2007). Phylogenetic divergence of CD47 interactions with human signal regulatory protein alpha reveals locus of species specificity. Implications for the binding site. J. Biol. Chem. 282, 1805–1818. Liu, Y., Tong, Q., Zhou, Y., Lee, H. W., Yang, J. J., Buhring, H. J. et al. (2007). Functional elements on SIRPalpha IgV domain mediate cell surface binding to CD47. J. Mol. Biol. 365, 680–693. Hatherley, D., Harlos, K., Dunlop, D. C., Stuart, D. I. & Barclay, A. N. (2007). The structure of the macrophage SIRPalpha inhibitory receptor reveals a binding face reminiscent of that used by T cell receptors. J. Biol. Chem. 282, 14567–14575. Yoshimura, M., Nagata, A., Matsuura, T., Kondou, Y., Nakagawa, A. & Tsukihara, T. (2005). Monte Carlo search with many CPUs: application to 6 dim. molecular replacement. Acta Crystallogr. A, 61, C161–C162. Nagata, A., Ohnishi, H., Yoshimura, M., Ogawa, A., Ujita, S., Adachi, H. et al. (2006). Crystallization and preliminary X-ray analysis of rat SHPS-1. Acta Crystallogr. F, 62, 189–191. Williams, A. F. & Barclay, A. N. (1988). The immunoglobulin superfamily-domains for cell surface recognition. Annu. Rev. Immunol. 6, 381–405. Bork, P., Holm, L. & Sander, C. (1994). The immunoglobulin fold. Structural classification, sequence patterns and common core. J. Mol. Biol. 242, 309–320. Stamper, C. C., Zhang, Y., Tobin, J. F., Erbe, D. V., Ikemizu, S., Davis, S. J. et al. (2001). Crystal structure of the B7-1/CTLA-4 complex that inhibits human immune responses. Nature, 410, 608–611. Schwartz, J. C., Zhang, X., Fedorov, A. A., Nathenson,
Structure of SHPS-1 Ligand-binding Domain
36.
37. 38.
39.
40.
41. 42. 43. 44. 45.
46. 47. 48.
49.
S. G. & Almo, S. C. (2001). Structural basis for costimulation by the human CTLA-4/B7-2 complex. Nature, 410, 604–608. Brooke, G., Holbrook, J. D., Brown, M. H. & Barclay, A. N. (2004). Human lymphocytes interact directly with CD47 through a novel member of the signal regulatory protein (SIRP) family. J. Immunol. 173, 2562–2570. Sano, S., Ohnishi, H. & Kubota, M. (1999). Gene structure of mouse BIT/SHPS-1. Biochem. J. 344, 667–675. Bodian, D. L., Jones, E. Y., Harlos, K., Stuart, D. I. & Davis, S. J. (1994). Crystal structure of the extracellular region of the human cell adhesion molecule CD2 at 2.5 A resolution. Structure, 2, 755–766. Collins, A. V., Brodie, D. W., Gilbert, R. J., Iaboni, A., Manso-Sancho, R., Walse, B. et al. (2002). The interaction properties of costimulatory molecules revisited. Immunity, 17, 201–210. Hayashi, A., Ohnishi, H., Okazawa, H., Nakazawa, S., Ikeda, H., Motegi, S. et al. (2004). Positive regulation of phagocytosis by SIRPbeta and its signaling mechanism in macrophages. J. Biol. Chem. 279, 29450–29460. van Beek, E. M., Cochrane, F., Barclay, A. N. & van den Berg, T. K. (2005). Signal regulatory proteins in the immune system. J. Immunol. 175, 7781–7787. Leslie, A. G. (2006). The integration of macromolecular diffraction data. Acta Crystallogr., Sect. D: Biol. Crystallogr. 62, 48–57. Collaborative Computing Project, Number 4 (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr. 50, 760–763. Vagin, A. & Teplyakov, A. (2000). An approach to multi-copy search in molecular replacement. Acta Crystallogr., Sect. D: Biol. Crystallogr. 56, 1622–1624. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr., Sect. D: Biol. Crystallogr. 53, 225–240. Lamzin, V. S. & Wilson, K. S. (1993). Automated refinement of protein models. Acta Crystallogr., Sect. D: Biol. Crystallogr. 49, 129–147. Emsley, P. & Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2126–2132. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. App. Crystallogr., 26. Akagi, T., Shishido, T., Murata, K. & Hanafusa, H. (2000). v-Crk activates the phosphoinositide 3-kinase/ AKT pathway in transformation. Proc. Natl Acad. Sci. USA, 97, 7290–7295.
J. Mol. Biol. (2008) 375, 650–660
Available online at www.sciencedirect.com
Structural Insight into the Specific Interaction between Murine SHPS-1/SIRPα and Its Ligand CD47 Aki Nakaishi 1 , Mayumi Hirose 1 , Masato Yoshimura 1 , Chitose Oneyama 2 , Kazunobu Saito 2 , Nobuharu Kuki 1 , Makoto Matsuda 1 , Nakayuki Honma 3 , Hiroshi Ohnishi 4 , Takashi Matozaki 4 , Masato Okada 2 ⁎ and Atsushi Nakagawa 1 1
Laboratory of Supramolecular Crystallography, Research Center for Structural and Functional Proteomics, Institute for Protein Research, Osaka University, Suita, Osaka 565-0871, Japan 2
Department of Oncogene Research, Research Institute for Microbial Diseases, Osaka University, 3-1 Yamadaoka, Suita, Osaka 565-0871, Japan 3 Pharmaceutical Research Laboratories, KIRIN Brewery Co., Ltd., Takasaki, Japan 4
Laboratory of Biosignal Sciences, Institute for Molecular and Cellular Regulation, Gunma University, Maebashi, Gunma 371-8512, Japan Received 12 June 2007; received in revised form 22 October 2007; accepted 29 October 2007 Available online 7 November 2007 Edited by J. Karn
SRC homology 2 domain-containing protein tyrosine phosphatase substrate 1 (SHPS-1 or SIRPα/BIT) is an immunoglobulin (Ig) superfamily transmembrane receptor and a member of the signal regulatory protein (SIRP) family involved in cell–cell interaction. SHPS-1 binds to its ligand CD47 to relay an inhibitory signal for cellular responses, whereas SIRPβ, an activating member of the same family, does not bind to CD47 despite sharing a highly homologous ligand-binding domain with SHPS-1. To address the molecular basis for specific CD47 recognition by SHPS-1, we present the crystal structure of the ligand-binding domain of murine SHPS1 (mSHPS-1). Folding topology revealed that mSHPS-1 adopts an I2-set Ig fold, but its overall structure resembles IgV domains of antigen receptors, although it has an extended loop structure (C′E loop), which forms a dimer interface in the crystal. Site-directed mutagenesis studies of mSHPS-1 identified critical residues for CD47 binding including sites in the C′E loop and regions corresponding to complementarity-determining regions of antigen receptors. The structural and functional features of mSHPS-1 are consistent with the human SHPS-1 structure except that human SHPS-1 has an additional β-strand D. These results suggest that the variable complementarity-determining region-like loop structures in the binding surface of SHPS-1 are generally required for ligand recognition in a manner similar to that of antigen receptors, which may explain the diverse ligand-binding specificities of SIRP family receptors. © 2007 Elsevier Ltd. All rights reserved.
Keywords: SHPS-1; SIRP family; immunogloblin superfamily; CD47; complementarity-determining regions
*Corresponding author. E-mail address: [email protected]. Abbreviations used: SHPS-1, SRC homology 2 domain-containing protein tyrosine phosphatase substrate 1; Ig, immunoglobulin; SIRP, signal regulatory protein; mSHPS-1, murine SHPS-1; CDR, complementarity-determining region; ITIM, immunoreceptor tyrosine-based inhibition motif; ITAM, immunoreceptor tyrosine-based activation motif; LBD, ligand-binding domain; rSHPS-1, rat SHPS-1; mSIRPβ, mouse SIRPβ; PBS, phosphate-buffered saline; DMEM, Dulbecco's modified Eagle's medium.
Introduction Cell–cell interaction, mediated by specific cell surface receptors, is crucial to support tissue integrity and intercellular communication in multicellular organisms. SRC homology 2 domain-containing protein tyrosine phosphatase substrate 1 (SHPS-1),1 also known as SIRPα2 or BIT,3 is a member of the immunoglobulin (Ig) superfamily involved in cell–cell communications. SHPS-1 contains an extracellular region composed of three Ig domains with multiple
0022-2836/$ - see front matter © 2007 Elsevier Ltd. All rights reserved.
651
Structure of SHPS-1 Ligand-binding Domain
N-linked glycosylation sites and a cytoplasmic region with two immunoreceptor tyrosine-based inhibition motifs (ITIMs). It is abundantly expressed in leukocytes and neural cells and plays a crucial role in various immune functions, including phagocytosis,4–6 cell motility7–9 and T cell function10–12 as well as additional functions in neural cells.13,14 SHPS-1 has also been implicated in the regulation of cancer progression.15 SHPS-1 belongs to the signal regulatory protein (SIRP) family, whose members all contain an N-terminal extracellular region, a single transmembrane domain and a C-terminal intracellular region.16 Based on the primary structure of its cytoplasmic region and its potential role in signal transduction, SIRP proteins are classified into three subfamilies, SIRPα (SHPS-1), SIRPβ and SIRPγ. SIRPα has an ITIM in its cytoplasmic region and, upon phosphorylation, relays inhibitory signals that regulate cell responses by recruiting tyrosine phosphatases such as SHP-1/2.1,17,18 In contrast, both SIRPβ and γ have minimal intracellular tails. SIRPβ1, an isoform of SIRPβ identified in humans and mice, binds to the immunoreceptor tyrosine-based activation motif (ITAM)-containing adaptor protein DAP12 and induces a positive regulatory signal by recruiting the cytoplasmic tyrosine kinases ZAP-70 and Syk.19,20 The signaling mechanism mediated by SIRPγ remains elusive. Based on these differential signaling pathways downstream of SHPS-1 and SIRPβ, the SIRP family proteins have been classified as inhibitory/activating paired receptors.11,16 CD47 (or IAP) is another cell surface Ig superfamily protein that serves as a ligand for SHPS-1.21–24 The SHPS-1–CD47 interaction has been confirmed by direct protein binding and cell adhesion assays,7 and the membrane distal Ig domain of SHPS-1 has been identified as the CD47 binding domain.21,25 In contrast, SIRPβ does not bind to CD47, despite sharing a highly homologous Ig domain with SHPS-1 (∼90% identity). It has been shown that SHPS-1 and SIRPβ1 are coexpressed on the same cell surfaces of granulocyte and mononuclear cells but mediate opposite cell signaling responses26, suggesting that these paired receptors must precisely recognize their specific binding partners. The specificity of the SHPS-1– CD47 interaction has been addressed by examining the interactions of various combinations of SHPS-1 and CD47 from distinct species,27 which demonstrated that SHPS-1–CD47 interaction is highly species-specific. To elucidate the molecular mechanism for the specific SHPS-1–CD47 interaction, candidate amino acid residues required for the human SHPS-1–CD47 interaction have been determined by mutagenesis analyses.28 Very recently, the potential CD47 binding surface on the human SHPS-1 Ig domain has been identified on the basis of its crystal structure.29 However, to generalize this observation and to address the molecular basis for the species-specific SHPS-1–CD47 interaction, comparative analysis of SHPS-1 structure from various species would be required. Here we present the crystal structure of the ligand-binding domain (LBD) of murine SHPS-1
(mSHPS-1). The crystal structure reveals that the Ig domain of mSHPS-1 has a striking similarity to the IgV domains of T-cell receptor and Igs, although it is topologically classified as an I2 set of Ig fold. Furthermore, site-directed mutagenesis studies identified two essential residues involved in the specific binding of mSHPS-1 to mCD47. These residues reside in the unique loop structure and the regions corresponding to the complementaritydetermining regions (CDRs) of antigen receptor. Together with the comparative analysis to the tentative rat SHPS-1 structure and the recently solved structure of human SHPS-129 and SIRPβ, our structure-based studies suggest that the binding surface of SHPS-1 is composed of variable CDR-like loop structures in a manner similarly observed in antigen receptors.
Results Molecular structure of mouse SHPS-1 The ligand-binding domain of mSHPS-1 (mSHPS1 LBD) was expressed and purified from bacteria according to the methods used for the preparation of rat SHPS-1 (rSHPS-1).31 The crystal structure of mSHPS-1 was determined at 1.4-Å resolution (Table 1) based on the tentative structure of rSHPS-131, which was solved by a molecular replacement search using the V-set Ig domain structure of the anti-testosterone Fab fragment [Protein Data Bank (PDB) code 1L7T] as an initial model. mSHPS-1 LBD contains an Ig fold characteristic of the Ig superfamily, identified by a two-layer β-sandwich structure (Fig. 1a and c). In addition to the β strands, mSHPS-1 contains a 310 helix at the periphery of the E–F loop. The core packing residues of Ig fold are also functional in mSHPS-1: a canonical disulfide bond (Cys24 and Cys90) packs against the consensus Table 1. Data collection and refinement statistics Data collection Space group Unit cell parameters (Å) Resolution (Å)a Observed reflectionsa Unique reflectionsa Rmergea,b 〈I/σ(I)〉a Completeness (%)a Multiplicitya Refinement Rcrystc Rfreec rmsd Bonds (Å) Angles (°) a
P212121 a = 48.6, b = 57.7, c = 69.2 39.75–1.40 (1.40–1.48) 281,790 (40,529) 39,022 (5624) 0.055 (0.340) 7.3 (2.1) 100 (100) 7.2 (7.2) 0.174 0.199 0.022 2.04
The highest resolution shell statistics are shown in parentheses. Rmerge = ∑hk∑| i Ii(hkl) − 〈I(hkl)〉|/∑hkl∑iIi(hkl). R cryst and R free = ∑| |F obs|−|F calc||/∑|Fobs|; R free was calculated for 5% of reflections randomly chosen for cross validation. b c
652 Trp residue (Trp37) (Fig. 1a). The mSHPS-1 LBD framework exhibits a folding topology that resembles both V- and I2-set folds.32,33 Similar to these Ig folds, mSHPS-1 has a typical kink between the A and A′ strands mediated by a cis-proline (Pro8). This allows the A and A′ strands to hydrogen-bond to the B and G strands, respectively. The strand composition of mSHPS-1 (ABCEGF) is categorized as an I2-set Ig domain, whereas mSHPS-1 has bulges of C′ and G strands characteristic of V-set Ig domains. Furthermore, the B strand breaks into shorter strands mediated by Asn23, a potential glycosylation site, and the C′E loop is atypically arranged. In the V-set Ig domains, the regions corresponding to the C′E loop are integrated into the β-sandwich structure via the C″ and D strands, while the C′E loop of mSHPS-1 is unpacked and protrudes from the core β-sandwich structure (Fig. 1c). rSHPS-1 LBD has a structure quite similar to that of mSHPS-1 (rmsd 0.90 Å), except for substantial deviations in the C′E loop structure (Fig. 1b). Considering that the amino acid sequences in the C′E loop are almost identical between the two species (Fig. 5), it seems that the C′E loop structure is structurally flexible or variable. In the crystal of mSHPS-1 LBD, there are two monomers in the asymmetric unit, with a low
Structure of SHPS-1 Ligand-binding Domain
rmsd of 0.75 Å between equivalent Cα atoms (Fig. 2a and b). The C′E loop contributes to the formation of binding surfaces that are stabilized by hydrogen bonds between the two molecules (Fig. 2a, blue box). Similarly, rSHPS-1 LBD can also form a dimer via the C′E loop interface, although its mode of interaction is apparently different, reflecting the unique conformations of the C′E loop in the two species (Fig. 2c). Notably, the rmsd value of the C′E loop between rSHPS-1 LBD dimers (2.7 Å) is significantly higher than the resting β-sandwich structure (0.35 Å) (Fig. 2d), supporting the flexible nature of the C′E loop in rSHPS-1. However, gel filtration chromatography analysis revealed that the major population of either mSHPS-1 or rSHPS-1 LBD eluted as a single peak of ∼ 13 kDa, suggesting that these molecules exist as a monomer in solution (data not shown). The monomeric feature of SHPS-1 LBD was further confirmed by sedimentation equilibrium experiments (Fig. 2e). Fitting data to a single-component model provided a theoretically monomeric molecular weight for mSHPS-1 LBD and rSHPS-1 LBD. In addition, the dimer is antiparallel with the C terminus of each monomer pointing in opposite directions. This configuration is unlikely to be formed by a transmembrane protein on a single cell surface. Taken together, it is likely
Fig. 1. Structure of mSHPS-1 LBD. (a) Ribbon diagram of the mSHPS-1 LBD monomer. The front (cyan) and back (light blue) sheets are composed of A′GGFCC′ and ABBE strands, respectively. The disulfide bond is shown in yellow. Locations of Pro8, Cys24, Trp37 and Cys90 are indicated by red dots. (b) Superimposed mSHPS-1 (cyan) on rSHPS-1 (pink) structures. Regions corresponding to the C′E loop are indicated. (c) Folding topologies of mSHPS-1 LBD (I2 set) and T-cell receptor Vα (V set).
Structure of SHPS-1 Ligand-binding Domain
653
Fig. 2. (a) The configuration of the two monomers in the asymmetric unit of mSHPS-1 LBD. The diagram to the right is a higher magnification of the region boxed in blue. Hydrogen bonds from Ala65 are indicated by broken red lines. (b) The two monomers of mSHPS-1 LBD are superimposed. (c) The configuration of the two monomers in the asymmetric unit of rSHPS-1 LBD. (d) The monomers of rSHPS-1 LBD are superimposed. (e) Sedimentation equilibrium analysis of mSHPS-1 LBD. Open circles represent data collected at 25,000 rpm (lower diagram). The curve represents the best fit to a singlecomponent sedimentation model, yielding an estimated molecular mass of 13.7 kDa. The residuals of the corresponding points are randomly distributed, indicating that the sample consists of only monomers (upper diagram).
that SHPS-1 LBD is present as a monomer in solution, although it is possible that the full-length SHPS-1 could form a dimer under certain physiological conditions. Structural comparison of SHPS-1 with T-cell receptor A structure-based homology search of the PDB by the DALI program revealed that mSHPS-1 LBD is highly homologous to V-set Ig proteins rather than I2-set proteins. T-cell receptor Vα (PDB code 1TCR) shows the highest similarity to mSHPS-1 LBD (rmsd 2.6 Å for 96 equivalent Cα atoms), despite only 28% amino acid sequence identity. Several other V-set Ig proteins, including Igs (PDB code 1MFA; rmsd 2.1 Å), sialoadhesion (PDB code 1QFO;
rmsd 2.2 Å) and NKp44 (PDB code 1HKF; rmsd 2.4 Å), are also significantly similar to mSHPS-1. The core secondary structure elements of mSHPS-1 LBD are tightly superimposed with those of T-cell receptor Vα (Fig. 3a). The major structural differences between mSHPS-1 and T-cell receptor are present in the loops of B′C, C′E and FG, which correspond to the complementarity-determining regions CDR1, CDR2 and CDR3 of T-cell receptor, respectively (Fig. 1c and Fig. 3a). Furthermore, electrostatic potential analysis revealed that the mSHPS-1 LBD includes a number of solventaccessible hydrophobic residues surrounding the regions corresponding to the CDR loops (Fig. 3b). These structural features suggest that SHPS-1 contains the binding surface similar to those of antigen receptor Ig superfamily proteins.34,35
654
Structure of SHPS-1 Ligand-binding Domain
Fig. 3. Comparison of mSHPS-1 and T-cell receptor. (a) mSHPS-1 LBD (cyan) and T-cell receptor Vα (yellow) structure (PDB code 1TCR) are superimposed. Regions corresponding to CDR1, CDR2 and CDR3 in T-cell receptor are marked in blue, green and red, respectively. (b) Electrostatic potential surfaces of mSHPS-1 LBD and T-cell receptor Vα. Potential binding surfaces are shown in rotated views (45°). Negative and positive potentials are colored in red and blue, respectively.
Surface plasmon resonance assay for the mSHPS-1–mCD47 interaction In order to locate the ligand-binding surface on mSHPS-1, we first confirmed the direct interaction between the recombinant mSHPS-1 LBD and the mCD47-Fc fusion protein by surface plasmon resonance (SPR) assay (Fig. 4). Interaction was detected at 25 °C as response units reached equilibrium in a dose-dependent manner (Fig. 4a). These analytical data reveal that recombinant mSHPS-1 LBD can specifically bind to the mCD47-Fc fusion protein. The equilibrium dissociation constant (Kd) of the mSHPS-1–mCD47 interaction was calculated from a dose–response curve (Fig. 4b and c). Nonlinear Langmuir fitting of the binding data gave a Kd of ∼ 1.3 μM, which is almost equivalent to the Kd values reported for the human SHPS-1 and CD47 interaction observed at 37 °C (∼ 2 μM).36 Cross-species interaction between rat SHPS-1 and mouse CD47 was detected with a Kd value of ∼1.5 μM, but a significant interaction of mouse SIRPβ (mSIRPβ) with mCD47 was not detected under our assay conditions (Fig. 4c). Mutagenesis analysis of the CD47 binding sites To determine the amino acid residues responsible for mSHPS-1–mCD47 binding, we prepared a series of mSHPS-1 mutants on the basis of the crystal structure and the amino acid sequences (Table 2 and Fig. 5). We focused on the residues in mSHPS-1 that are conserved in rat SHPS-1 but not in the mSIRPβ. Since nonconserved residues between mSHPS-1 and mSIRPβ accumulate within the N-terminal 30 residues, we prepared chimera proteins consisting of the N-terminal mSIRPβ and the C-terminal mSHPS-1 (SIRPβ1–14 SHPS-115–115 and SIRPβ1–26 SHPS-127–115). Furthermore, we made single or multiple substitutions of mSHPS-1 residues with mSIRPβ
residues. The substitute residues of mSHPS-1 include Lys35 at the C-terminal end of the B′C loop (corresponding to CDR1); Ser50, Phe56, Val59 and Ala60 in the C″E loop (CDR2); Glu83 and Ala85 in the 310 module; Lys95 and Thr102 in the FG loop (CDR3); and Tyr115 near the C terminus (Fig. 5). These mutant proteins were produced in Escherichia coli, and the folding status of each purified protein was confirmed by CD spectroscopy analysis (data not shown). Most of the mutations did not significantly affect protein folding, but the substitution of Phe56 to Ser induced significant deviations in the structure, implicating a potential role for Phe56 in maintaining the overall structure. The binding affinities of these mutants to the mCD47-Fc fusion protein were estimated by SPR (Table 2). The Kd values of SIRPβ1–14 SHPS-115–115, SIRPβ1–26 SHPS-127–115, K36R, S50P, V59I, E83A, A85E, E83A/A85S, P97S, and Y115L were in a range close to that of wild-type mSHPS-1 (1∼2 μM), suggesting that these residues are not essential for the specific interaction between mSHPS-1 and mCD47. The T102I mutant showed a slightly higher affinity for mCD47-Fc (∼ 0.7 μM) than the wild-type molecule, indicating some inhibitory role of the Thr102 residue. In contrast, F56S, S50P/F56S, A65V, K95R, K95R/P97S and K95R/P97S/T102I showed significantly lower affinities with Kd values ranging from 3.42 to 18.0. Particularly, an Ala65Val substitution greatly affected the mSHPS-1–mCD47 binding. Cell adhesion assays of the full-length mSHPS-1 mutants The results of the SPR assays suggest that Phe56, Ala65, Lys95 and Thr102 are potentially required for the mSHPS-1–mCD47 binding specificity. To further evaluate the role of these residues in vivo, we expressed full-length mSHPS-1 having the substitutions of these residues in CHO-Ras cells and ana-
655
Structure of SHPS-1 Ligand-binding Domain
Discussion
Fig. 4. SPR assay of the SHPS-1–CD47 interaction. (a) Binding of mSHPS-1 LBD to immobilized mCD47-Fc fusion protein was detected in response units according to the protocol provided by the manufacturer (see Materials and Methods). Background responses obtained for the immobilized rat P-selectin-Fc fusion protein were subtracted from the total response. A diagram of the data obtained for a concentration gradient series of mSHPS-1 LBD is shown. (b) Dose response curve for the m SHPS-1– CD47 interaction. (c) Nonlinear curve fitting of the data using a 1:1 Langmuir binding isotherm yielded a Kd of 1.31 ± 0.27 μM for mSHPS-1 LBD. Similarly, the rate constant for rSHPS-1 was calculated to be 1.46 ± 0.06. However, mSIRPβ did not have a significant binding capacity for mCD47. N.S., not significant.
lyzed their binding abilities to mCD47 (Fig. 6a). Cell binding assays on immobilized mCD47-Fc fusion protein showed that the cells expressing mSHPS-1 mutants, F56S, A65V, K95R, S50P/F56S and K95R/P97S/T102I, were less adhesive compared to cells expressing wild type, indicating that Phe56, Ala65 and Lys95 were critical for binding (Fig. 6b). Similar findings were obtained when the cell adhesion assay was performed on CHO-Ras cells expressing full-length mCD47 (Fig. 6c). These results demonstrate that Phe56, Ala65 and Lys95 residues play critical roles in the mSHPS-1–mCD47 binding in vivo.
Our crystallographic analysis of mouse and rat SHPS-1 LBD has provided basic information required to understand the ligand, species and/or strain specificity of SIRP family receptors. The structure-based homology search revealed a striking homology between mSHPS-1 LBD and IgV domains of antigen receptors such as T-cell receptor and Igs, except that mSHPS-1 LBD has a unique loop structure (C′E loop), which forms a dimer interface in the crystal. These structural features are well consistent with the recently presented human SHPS1 LBD structure,29 confirming that SHPS-1 is evolutionarily and functionally a close relative of the antigen receptor-type Ig superfamily.16 Site-directed mutagenesis studies uncovered the mCD47-binding surface of mSHPS-1. We identified three residues, Phe56, Ala65 and Lys95, which are critical for specific recognition of mCD47 (Fig. 7a and b). Ala65 is located in the middle of the C′E loop that forms dimer interface in the crystal. In the mSHPS-1 dimer, Ala65 makes stable hydrogen bonds to Ser75 and Arg68 residues of the other molecule (Fig. 2a). However, the monomeric feature of mSHPS-1 LBD in solution suggests that dimer formation is potentially due to the flexibility and protein-binding ability of the C′E loop. Although it is currently unclear whether Ala65 directly interacts with the ligand or contributes to maintaining the conformation of mSHPS-1, it is likely that the C′E loop region containing Ala65 is involved in ligand recognition. However, it is known that the ligandbinding domain of SHPS-1 has substantial sequence diversity even among mouse strains and that Ala65 is a hot spot of polymorphism.37 The 129/sv strain type of mSHPS-1 has Ala65, whereas Ala65 is replaced by a Thr in Balb/c and C57BL/6 strains. On the other hand, mSIRPβ in C57BL/6 strain has a Val in this position, although Balb/c strain has a Thr. In the human SIRP family, SIRPα/SHPS-1 has a Ser at this site, and is replaced by a Leu in SIRPβ and Table 2. Effects of site-directed mutations on the mSHPS1–mCD47 binding mSHPS-1 mutants SIRPβ1–14 SHPS-115–115 SIRPβ1–26 SHPS-127–115 K36R S50P F56S S50P/F56S V59I A65V E83A A85E E83A/A85S K95R P97S K95R/P97S T102I Y115L K95R/P97S/T102I
Kd (μM) 1.39 ± 0.11 1.82 ± 0.06 1.18 ± 0.04 1.76 ± 0.12 4.16 ± 0.11 10.7 ± 0.80 1.00 ± 0.15 18.0 ± 0.40 1.17 ± 0.12 1.29 ± 0.03 1.11 ± 0.01 5.29 ± 0.98 1.00 ± 0.04 3.42 ± 0.16 0.76 ± 0.05 1.21 ± 0.02 3.62 ± 0.47
656
Structure of SHPS-1 Ligand-binding Domain
Fig. 5. Structure-based sequence alignment of mSHPS-1, rSHPS-1, SIRPβ and hSHPS-1. Numbering is for mSHPS-1 LBD protein used for structural analysis in this study. In mSHPS-1, rSHPS-1 and SIRPβ, residues replaced in the mutagenesis study are highlighted. Significant and irrelevant residues are boxed in red and green, respectively. In hSHPS-1, residues shown to be critical for ligand binding are boxed in orange.29 Locations of the β-strands are indicated by bold arrows above the alignment. The D strand unique to hSHPS-1 is indicated by a bold arrow under the alignment. The 310 module is also indicated by a small cylinder. Regions corresponding to CDR1, CDR2 and CDR3 in the antigen receptors are shown by blue, green and red boxes, respectively.
SIRPγ. The site-directed mutagenesis of Ser to Asp in this position of human SIRPα/SHPS-1 destroyed ligand-binding activity, confirming the critical role of this site.29 Thus, it is likely that the residue equivalent to Ala65 of mSHPS-1 is critical in determining the species, strain and/or isoform specificity in ligand recognition. Another critical residue identified by our mutagenesis analysis is Lys95, which is exposed on the molecular surface in close proximity to CDR3 (Fig. 7a and b). The substitution of Lys95 with an Arg in mSHPS-1 substantially abated the mCD47 binding. mSIRPβ that does not bind to mCD47 has an Arg residue at this site in all mouse strains examined. The Lys95 residue is also conserved in the human SHPS-1, and mutagenesis at this site destroyed its ligand-binding activity.29 However, mSHPS-1 in Balb/c strain has an Arg residue at the same site, suggesting that Lys residue is not essential for CD47 binding or that the CD47 binding activity of mSHPS-1 in Balb/c strain may be perturbed. Although additional analysis is required to elucidate its precise roles, it seems that this site is critical for constructing the species/strain-specific binding surface, as is the case for Ala65. The substitution of Phe56 to a Ser also reduced the mSHPS-1–mCD47 binding affinity, but circular dichroism spectroscopy analysis showed that the substitution of Phe56 significantly altered the folding status of the recombinant protein, indicating that Phe56 is required for proper protein folding of recombinant mSHPS-1. Both Ala65 and Lys95 are positioned in the ligandbinding surface, which contains CDR-like loop structures. Several residues critical for ligand binding were also mapped onto similar loops of human SHPS-1.29 These findings support the theory that SHPS-1 uses these CDR-like loop structures for ligand binding, rather than the faces of β-sheet that are often utilized by other Ig superfamily receptors such as CD238 and B7.39 Given that CDR-like regions are hot spots for polymorphism,2,37 the use of these regions for ligand recognition allows SIRP family receptors to acquire diverse ligand-binding specificities by only small changes in amino acids. The CDR1 and CDR3 loops in SHPS-1 are well
conserved among different species, while the loops equivalent to CDR2 (C′E loop of rodent SHPS-1 and DE loop of human SHPS-1) are significantly divergent; human SHPS-1 has an additional D strand29 (Fig. 7c). These observations suggest that deviations in the CDR2-like regions would be important for species specificity of SHPS-1. Recently an NMR structure of the membrane distal Ig domain of human SIRPβ became available (PDB code 2D9C). The NMR structure reveals the striking similarity in the backbone structures of SHPS-1 and SIRPβ (rmsd 1.2 Å) (Fig. 7d), which further supports the above hypothesis that a small modulation of the binding loops are sufficient for the generation of diverse ligand specificities in the SIRP family receptors. The SIRP gene family is thought to have evolved rapidly, with variable numbers of genes in different species and distinct genes, such as SIRPα/β/γ, that have evolved by gene duplication.16 Polymorphisms have been frequently identified in the SIRP gene family. For example, SIRPβ proteins from two mouse strains have 22 amino acid differences,40 and SHPS-1 accumulates polymorphisms in its ligandbinding domain.2,37 Additional genes related to the SIRP family have also been identified.41 The fulllength extracellular domain of SHPS-1 consists of a membrane-distal Ig domain that structurally resembles the V-set Ig fold of antigen receptors, a joining (J)-like region that is present in the putative antigen receptor predecessor, and two additional Ig domains that have the highest sequence similarity to the C1-set Ig domains characteristic of antigen receptors. Based on these structural features, it has been suggested that the SIRP family may have originated from a precursor of the rearranging antigen receptors.16 The speed of evolution may have allowed SHPS-1 to acquire functional diversity and species- and/or tissue-specific cell–cell recognition systems. Considering the implications of SHPS-1– CD47 signaling in innate immune response, autoimmune disease, hemolytic anemia and tumor progression, our structure-based functional analysis of SHPS-1 provides essential information for the development of therapeutic agents that could target SHPS-1 and/or CD47.
657
Structure of SHPS-1 Ligand-binding Domain
Materials and Methods Protein expression, purification and crystallization The ligand-binding domain of mouse SHPS-1 (mSHPS-1 LBD), containing residues 31–147, was subcloned into the BamHI and NdeI restriction sites of the bacterial expression vector pGEX-6P-1 (GE Healthcare) and transformed into the E. coli strain Rosetta-gami B (Novagen). The expressed soluble proteins were purified as described previously.31 Crystallization of mSHPS-1 LBD was performed by hanging-drop vapor diffusion in 0.1 M Pipes, pH 6.5, and 32% polyethylene glycol 4000. The crystal of rat SHPS-1 (rSHPS-1) LBD was obtained as described.31
Structure determination and refinement Diffraction data sets of rSHPS-1 at 2.8 Å and mSHPS-1 at 1.4 Å were collected on the beam line BL44XU at SPring-8 (Hyogo, Japan) and processed using MOSFLM42 and SCALA.43 The initial structure of rSHPS-1 was solved by six-dimensional molecular replacement search with Monte Carlo method,30 using the V-set Ig domain structure of the anti-testosterone Fab fragment (PDB code 1L7T) as an initial model. The rSHPS-1 structure was refined to an R value of 0.28 and an Rfree value of 0.36 at 2.8-Å resolution by REFMAC5 but the refinement is still in progress (unpublished result). A molecular replacement solution was found with the program MOLREP44 using the current rSHPS-1 structure as a search model. The structure of mSHSP-1 was refined using REFMAC543,45 and ARP/wARP46 combined with manual model rebuilding with COOT.47 The stereochemical properties of the structure were assessed by PROCHECK,48 which showed no residues in the disallowed regions on the Ramachandran plot. Diffraction data collection and refinement statistics of mSHPS-1 are listed in Table 1. Analytical ultracentrifugation Sedimentation equilibrium experiments were performed at 20 °C in an analytical ultracentrifuge using an An-60 Ti rotor and 1.2-cm six-channel charcoal-Epon centerpieces (Beckman Coulter). mSHPS-1 LBD was concentrated and exchanged into 10 mM Hepes, pH 7.4, 1 mM ethylenediaminetetraacetic acid (EDTA) and 150 mM NaCl. Data were collected at protein concentrations of 54, 31, and 18 μM (0.70, 0.47, and 0.23 mg/ml, respectively). Centrifugation was performed at 25,000 rpm, and interference scans were taken after 20 h when equilibrium was accomplished. Data analysis was performed using Beckman analytical software with the protein partial specific volume of 0.7391 cm3g− 1 and the buffer density 1.005 cm3g− 1. Site-directed mutagenesis All site-directed mutagenesis of mSHPS-1 were performed using two polymerase chain reactions. In the first
Fig. 6. The effect of mSHPS-1 mutations on cell adhesion mediated by mSHPS-1–mCD47 binding. (a) Expression of various mSHPS-1 mutants and mSIRPβ in CHORas cells. CHO-Ras cells transfected with the expression plasmids for wild-type or mutant mSHPS-1 and mSIRPβ were lysed and subjected to Western blot analysis using an anti-myc antibody. Actin was detected with an anti-actin antibody as an internal control. (b) The binding of mSHPS1 transfectants to mCD47-Fc fusion protein. Cell Tracker Green CMFDA-labeled mSHPS-1 transfectants were added to the wells coated with mCD47-Fc fusion protein. After washing, bound cells were measured by fluorescence. The data represent the means and standard deviations of four independent experiments. (c) The binding of mSHPS-1 transfectants to mCD47-transfected CHO monolayers. Cell Tracker Green CMFDA-labeled mSHPS-1 transfectants preincubated with (open bars) or without (filled bars) 50 μg/ml mCD47-Fc fusion protein were added to mCD47-transfected CHO monolayer cultures. The bound cells were measured by fluorescence. The data represent the means and the standard deviations of four independent experiments.
658
Structure of SHPS-1 Ligand-binding Domain
Fig. 7. The critical residues of mSHPS-1 required for mCD47 binding. (a) The positions of Phe56, Ala65 and Lys95 are shown on the ribbon diagram of mSHPS-1 LBD. (b) Phe56, Ala65 and Lys95 are marked in magenta. The residues whose mutations had no significant effect on ligand binding are indicated in gray. (c) LBDs of mSHPS-1 (cyan), rSHPS-1 (pink) and hSHPS-1 (yellow) are superimposed. (d) N-terminal Ig domains of hSHPS-1 (yellow) and hSIRPβ (gray) (PDB code 2D9C) are superimposed. reaction, the N-terminal and the C-terminal fragments were amplified separately, using the primers containing mutation sites. In the second reaction, these two amplified fragments were mixed together and amplified using the N-terminal forward and the C-terminal reverse primers. The PCR products were sequenced and then subcloned into pGEX-6P-1 (GE Healthcare) for bacterial expression, and into the retrovirus expression vector pCX4-myc. Proper folding of the bacterially expressed proteins were confirmed by circular dichroism by a JASCO J720WI spectropolarimeter (JASCO, Tokyo, Japan) with a spectral bandwidth of 1 nm at a sample concentration of 0.2 mg/ ml in phosphate-buffered saline (PBS). The expression levels of the full-length mutants of mSHPS-1 in COS cells were determined by Western blotting using an anti-myc antibody (MBL, Nagoya, Japan). SPR binding analysis SPR binding assays were carried out with a Biacore 2000 instrument (GE Healthcare) at 25 °C in HBS-EP running buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 3 mM EDTA and 0.005% surfactant P20). Mouse CD47-Fc fusion protein (mCD47-Fc) was purified from the culture medium of CHO-Ras cells stably expressing mCD47-Fc using Protein A Sepharose (GE Healthcare), and directly immobilized to the sensor surface of the experimental flow cell by amine coupling, which typically resulted in immobilization levels of 1000 RU. A sensor chip coupled to a rat P-selectin Ig fusion protein was used as a negative control for the binding assays and recorded levels of ∼ 1400 RU. For determining the binding affinity of the wild-type mSHPS-1 for mCD47-Fc, 11 serial dilutions of the mSHPS-1 were prepared in HBS-EP. Background responses from both the control flow cell and buffer alone were subtracted to obtain the final response
levels. Each experiment was repeated more than three times and the average was taken as the final result. Equilibrium dissociation constants were obtained by fitting the data using a 1:1 Langmuir binding model with the BIAevaluation software (GE Healthcare). The constants of mouse SHPS-1 mutants were measured using the same protocol. Retroviral expression and cell adhesion assays pCX4bsr and pCX4bsr-myc, two Moloney murine leukemia virus-based retrovirus vectors containing the blasticidin S resistance gene (bsr) as selectable markers were kindly provided by Dr. T. Akagi.49 pCX4-myc-SHPS1 and pCX4-CD47 were constructed by inserting wildtype or mutant full-length mouse SHPS-1 and wild-type full-length mouse CD47 into the multicloning site of pCX4bsr-myc and pCX4bsr, respectively. CHO cells stably expressing H-Ras-V12 (CHO-Ras cells), kindly provided by Dr. Akagi, were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum. CHO-Ras cells seeded in six-well plates at 5 × 105 cells per well were transfected with 10 μg of pCX4bsr-myc containing SHPS-1 using Lipofectamine LTX (Invitrogen). CHO-Ras cells stably expressing mCD47 were established by selection in the presence of blasticidin. To determine the ability of cell–cell adhesion mediated by mSHPS-1–mCD47 interaction, mCD47-transfected CHO-Ras cells were cultured on 96-well plates and fixed with 4% paraformaldehyde for 10 min. After the wells were washed with PBS, the cultures were incubated with 2% bovine serum albumin in PBS for 30 min to block nonspecific binding. Wild-type and mutant SHPS-1 transfectants were labeled with 5 mM Cell Tracker Green CMFDA (Invitrogen) and preincubated with or
659
Structure of SHPS-1 Ligand-binding Domain without 50 μg/ml mCD47-Fc fusion protein. The labeled mSHPS-1 transfectants (1 × 105 cells per well) were applied to the fixed mCD47-transfected monolayers and allowed to bind for 40 min at room temperature. The wells were then filled with DMEM containing 10% FCS, and the plate was inverted and placed for 5 min. After removal of the unbound cells by gentle aspiration, 1% NP-40 in PBS was added to each well and fluorescence was measured using 485- and 538-nm excitation and emission filter, respectively, in a Labosystems Fluoroscan II.
9.
10.
Protein Data Bank accession number The atomic coordinates of mSHPS-1 have been deposited in the Protein Data Bank (PDB code 2YZ1).
11.
12.
Acknowledgements We thank Drs A. Ogawa, T. Matsuura, and E. Yamashita for assistance with crystal data collection and analysis, M. Sakai for analytical ultracentrifugation experiments, T. Akagi for providing the pCX4 vector, and Y. Segawa and F. Akita for help with sample preparation. This work was supported by the National Project on Protein Structural and Functional Analyses from the Ministry of Education, Sports, Culture, Science and Technology of Japan.
References 1. Fujioka, Y., Matozaki, T., Noguchi, T., Iwamatsu, A., Yamao, T., Takahashi, N. et al. (1996). A novel membrane glycoprotein, SHPS-1, that binds the SH2domain-containing protein tyrosine phosphatase SHP2 in response to mitogens and cell adhesion. Mol. Cell. Biol. 16, 6887–6899. 2. Kharitonenkov, A., Chen, Z., Sures, I., Wang, H., Schilling, J. & Ullrich, A. (1997). A family of proteins that inhibit signalling through tyrosine kinase receptors. Nature, 386, 181–186. 3. Sano, S., Ohnishi, H., Omori, A., Hasegawa, J. & Kubota, M. (1997). BIT, an immune antigen receptorlike molecule in the brain. FEBS Lett. 411, 327–334. 4. Oldenborg, P. A., Zheleznyak, A., Fang, Y. F., Lagenaur, C. F., Gresham, H. D. & Lindberg, F. P. (2000). Role of CD47 as a marker of self on red blood cells. Science, 288, 2051–2054. 5. Gardai, S. J., McPhillips, K. A., Frasch, S. C., Janssen, W. J., Starefeldt, A., Murphy-Ullrich, J. E. et al. (2005). Cell-surface calreticulin initiates clearance of viable or apoptotic cells through trans-activation of LRP on the phagocyte. Cell, 123, 321–334. 6. Okazawa, H., Motegi, S., Ohyama, N., Ohnishi, H., Tomizawa, T., Kaneko, Y. et al. (2005). Negative regulation of phagocytosis in macrophages by the CD47SHPS-1 system. J. Immunol. 174, 2004–2011. 7. Liu, Y., Buhring, H. J., Zen, K., Burst, S. L., Schnell, F. J., Williams, I. R. & Parkos, C. A. (2002). Signal regulatory protein (SIRPalpha), a cellular ligand for CD47, regulates neutrophil transmigration. J. Biol. Chem. 277, 10028–10036. 8. Motegi, S., Okazawa, H., Ohnishi, H., Sato, R., Kaneko, Y., Kobayashi, H. et al. (2003). Role of the
13.
14.
15.
16. 17.
18.
19.
20.
21.
22.
23.
CD47–SHPS-1 system in regulation of cell migration. EMBO J. 22, 2634–2644. Inagaki, K., Yamao, T., Noguchi, T., Matozaki, T., Fukunaga, K., Takada, T. et al. (2000). SHPS-1 regulates integrin-mediated cytoskeletal reorganization and cell motility. EMBO J. 19, 6721–6731. Latour, S., Tanaka, H., Demeure, C., Mateo, V., Rubio, M., Brown, E. J. et al. (2001). Bidirectional negative regulation of human T and dendritic cells by CD47 and its cognate receptor signal-regulator proteinalpha: down-regulation of IL-12 responsiveness and inhibition of dendritic cell activation. J. Immunol. 167, 2547–2554. Seiffert, M., Brossart, P., Cant, C., Cella, M., Colonna, M., Brugger, W. et al. (2001). Signal-regulatory protein alpha (SIRPalpha) but not SIRPbeta is involved in T-cell activation, binds to CD47 with high affinity, and is expressed on immature CD34(+)CD38(−) hematopoietic cells. Blood, 97, 2741–2749. Piccio, L., Vermi, W., Boles, K. S., Fuchs, A., Strader, C. A., Facchetti, F. et al. (2005). Adhesion of human T cells to antigen-presenting cells through SIRPbeta2– CD47 interaction costimulates T-cell proliferation. Blood, 105, 2421–2427. Nakahata, Y., Okumura, N., Shima, T., Okada, M. & Nagai, K. (2000). Light-induced tyrosine phosphorylation of BIT in the rat suprachiasmatic nucleus. J. Neurochem. 74, 2436–2444. Murata, T., Ohnishi, H., Okazawa, H., Murata, Y., Kusakari, S., Hayashi, Y. et al. (2006). CD47 promotes neuronal development through Src- and FRG/Vav2mediated activation of Rac and Cdc42. J. Neurosci. 26, 12397–12407. Oshima, K., Machida, K., Ichigotani, Y., Nimura, Y., Shirafuji, N., Hamaguchi, M. & Matsuda, S. (2002). SHPS-1: a budding molecule against cancer dissemination. Cancer Res. 62, 3929–3933. Barclay, A. N. & Brown, M. H. (2006). The SIRP family of receptors and immune regulation. Nat. Rev. Immunol. 6, 457–464. Lienard, H., Bruhns, P., Malbec, O., Fridman, W. H. & Daeron, M. (1999). Signal regulatory proteins negatively regulate immunoreceptor-dependent cell activation. J. Biol. Chem. 274, 32493–32499. Veillette, A., Thibaudeau, E. & Latour, S. (1998). High expression of inhibitory receptor SHPS-1 and its association with protein-tyrosine phosphatase SHP- 1 in macrophages. J. Biol. Chem. 273, 22719–22728. Dietrich, J., Cella, M., Seiffert, M., Buhring, H. J. & Colonna, M. (2000). Cutting edge: signal-regulatory protein beta 1 is a DAP12-associated activating receptor expressed in myeloid cells. J. Immunol. 164, 9–12. Tomasello, E., Cant, C., Buhring, H. J., Vely, F., Andre, P., Seiffert, M. et al. (2000). Association of signal-regulatory proteins beta with KARAP/DAP-12. Eur. J. Immunol. 30, 2147–2156. Vernon-Wilson, E. F., Kee, W. J., Willis, A. C., Barclay, A. N., Simmons, D. L. & Brown, M. H. (2000). CD47 is a ligand for rat macrophage membrane signal regulatory protein SIRP (OX41) and human SIRPalpha 1. Eur. J. Immunol. 30, 2130–2137. Brown, E., Hooper, L., Ho, T. & Gresham, H. (1990). Integrin-associated protein: a 50-kD plasma membrane antigen physically and functionally associated with integrins. J. Cell. Biol. 111, 2785–2794. Lindberg, F. P., Gresham, H. D., Schwarz, E. & Brown, E. J. (1993). Molecular cloning of integrin-associated protein: an immunoglobulin family member with multiple membrane-spanning domains implicated in
660
24.
25.
26.
27.
28.
29.
30.
31.
32. 33. 34.
35.
alpha v beta 3-dependent ligand binding. J. Cell. Biol. 123, 485–496. Mawby, W. J., Holmes, C. H., Anstee, D. J., Spring, F. A. & Tanner, M. J. (1994). Isolation and characterization of CD47 glycoprotein: a multispanning membrane protein which is the same as integrin-associated protein (IAP) and the ovarian tumour marker OA3. Biochem. J. 304, 525–530. Liu, Y., O'Connor, M. B., Mandell, K. J., Zen, K., Ullrich, A., Buhring, H. J. & Parkos, C. A. (2004). Peptide-mediated inhibition of neutrophil transmigration by blocking CD47 interactions with signal regulatory protein alpha. J. Immunol. 172, 2578–2585. Liu, Y., Soto, I., Tong, Q., Chin, A., Buhring, H. J., Wu, T. et al. (2005). SIRPbeta1 is expressed as a disulfidelinked homodimer in leukocytes and positively regulates neutrophil transepithelial migration. J. Biol. Chem. 280, 36132–36140. Subramanian, S., Boder, E. T. & Discher, D. E. (2007). Phylogenetic divergence of CD47 interactions with human signal regulatory protein alpha reveals locus of species specificity. Implications for the binding site. J. Biol. Chem. 282, 1805–1818. Liu, Y., Tong, Q., Zhou, Y., Lee, H. W., Yang, J. J., Buhring, H. J. et al. (2007). Functional elements on SIRPalpha IgV domain mediate cell surface binding to CD47. J. Mol. Biol. 365, 680–693. Hatherley, D., Harlos, K., Dunlop, D. C., Stuart, D. I. & Barclay, A. N. (2007). The structure of the macrophage SIRPalpha inhibitory receptor reveals a binding face reminiscent of that used by T cell receptors. J. Biol. Chem. 282, 14567–14575. Yoshimura, M., Nagata, A., Matsuura, T., Kondou, Y., Nakagawa, A. & Tsukihara, T. (2005). Monte Carlo search with many CPUs: application to 6 dim. molecular replacement. Acta Crystallogr. A, 61, C161–C162. Nagata, A., Ohnishi, H., Yoshimura, M., Ogawa, A., Ujita, S., Adachi, H. et al. (2006). Crystallization and preliminary X-ray analysis of rat SHPS-1. Acta Crystallogr. F, 62, 189–191. Williams, A. F. & Barclay, A. N. (1988). The immunoglobulin superfamily-domains for cell surface recognition. Annu. Rev. Immunol. 6, 381–405. Bork, P., Holm, L. & Sander, C. (1994). The immunoglobulin fold. Structural classification, sequence patterns and common core. J. Mol. Biol. 242, 309–320. Stamper, C. C., Zhang, Y., Tobin, J. F., Erbe, D. V., Ikemizu, S., Davis, S. J. et al. (2001). Crystal structure of the B7-1/CTLA-4 complex that inhibits human immune responses. Nature, 410, 608–611. Schwartz, J. C., Zhang, X., Fedorov, A. A., Nathenson,
Structure of SHPS-1 Ligand-binding Domain
36.
37. 38.
39.
40.
41. 42. 43. 44. 45.
46. 47. 48.
49.
S. G. & Almo, S. C. (2001). Structural basis for costimulation by the human CTLA-4/B7-2 complex. Nature, 410, 604–608. Brooke, G., Holbrook, J. D., Brown, M. H. & Barclay, A. N. (2004). Human lymphocytes interact directly with CD47 through a novel member of the signal regulatory protein (SIRP) family. J. Immunol. 173, 2562–2570. Sano, S., Ohnishi, H. & Kubota, M. (1999). Gene structure of mouse BIT/SHPS-1. Biochem. J. 344, 667–675. Bodian, D. L., Jones, E. Y., Harlos, K., Stuart, D. I. & Davis, S. J. (1994). Crystal structure of the extracellular region of the human cell adhesion molecule CD2 at 2.5 A resolution. Structure, 2, 755–766. Collins, A. V., Brodie, D. W., Gilbert, R. J., Iaboni, A., Manso-Sancho, R., Walse, B. et al. (2002). The interaction properties of costimulatory molecules revisited. Immunity, 17, 201–210. Hayashi, A., Ohnishi, H., Okazawa, H., Nakazawa, S., Ikeda, H., Motegi, S. et al. (2004). Positive regulation of phagocytosis by SIRPbeta and its signaling mechanism in macrophages. J. Biol. Chem. 279, 29450–29460. van Beek, E. M., Cochrane, F., Barclay, A. N. & van den Berg, T. K. (2005). Signal regulatory proteins in the immune system. J. Immunol. 175, 7781–7787. Leslie, A. G. (2006). The integration of macromolecular diffraction data. Acta Crystallogr., Sect. D: Biol. Crystallogr. 62, 48–57. Collaborative Computing Project, Number 4 (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr. 50, 760–763. Vagin, A. & Teplyakov, A. (2000). An approach to multi-copy search in molecular replacement. Acta Crystallogr., Sect. D: Biol. Crystallogr. 56, 1622–1624. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr., Sect. D: Biol. Crystallogr. 53, 225–240. Lamzin, V. S. & Wilson, K. S. (1993). Automated refinement of protein models. Acta Crystallogr., Sect. D: Biol. Crystallogr. 49, 129–147. Emsley, P. & Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2126–2132. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. App. Crystallogr., 26. Akagi, T., Shishido, T., Murata, K. & Hanafusa, H. (2000). v-Crk activates the phosphoinositide 3-kinase/ AKT pathway in transformation. Proc. Natl Acad. Sci. USA, 97, 7290–7295.