|Home | About | Journals | Submit | Contact Us | Français|
T cell/transmembrane, Ig, and mucin (TIM) proteins, identified using a congenic mouse model of asthma, critically regulate innate and adaptive immunity. TIM-1 and TIM-4 are receptors for phosphatidylserine (PtdSer), exposed on the surfaces of apoptotic cells. Herein, we show with structural and biological studies that TIM-3 is also a receptor for PtdSer that binds in a pocket on the N-terminal IgV domain in coordination with a calcium ion. The TIM-3/PtdSer structure is similar to that of TIM-4/PtdSer, reflecting a conserved PtdSer binding mode by TIM family members. Fibroblastic cells expressing mouse or human TIM-3 bound and phagocytosed apoptotic cells, with the BALB/c allelic variant of mouse TIM-3 showing a higher capacity than the congenic C.D2 Es-Hba–allelic variant. These functional differences were due to structural differences in the BC loop of the IgV domain of the TIM-3 polymorphic variants. In contrast to fibroblastic cells, T or B cells expressing TIM-3 formed conjugates with but failed to engulf apoptotic cells. Together these findings indicate that TIM-3–expressing cells can respond to apoptotic cells, but the consequence of TIM-3 engagement of PtdSer depends on the polymorphic variants of and type of cell expressing TIM-3. These findings establish a new paradigm for TIM proteins as PtdSer receptors and unify the function of the TIM gene family, which has been associated with asthma and autoimmunity and shown to modulate peripheral tolerance.
The T cell/transmembrane, Ig, and mucin (TIM) genes comprise a family with diverse functions in regulating immunity in both mice and humans (1, 2). Using a congenic mouse model of allergen-induced airway hyperreactivity (AHR), the eight murine TIM family members were identified by positional cloning in the T cell airway phenotype regulatory (Tapr) locus (3). The human TIM (hTIM) gene family orthologues include TIM-1, TIM-3, and TIM-4 (3, 4). TIM-1, an important susceptibility gene for asthma and allergy (5), is preferentially expressed on Th2 cells and functions as a potent costimulatory molecule for T cell activation (6). TIM-3 is preferentially expressed on Th1 cells but is also expressed on dendritic cells (DCs) (1) and generates an inhibitory signal resulting in the apoptosis of Th1 cells (7). TIM-4 is exclusively expressed on APCs and may play an important role in mediating tolerance (8, 9).
TIM proteins are type I cell surface glycoproteins with common structural features including Ig-like, mucin, transmembrane, and cytoplasmic domains. The crystal structures of murine TIM-1, TIM-2, TIM-3, and TIM-4 Ig-like domains have been determined (10-12). All are IgV domains with a long CC′ loop folded back onto the GFC β-sheet, which is held in place by two disulfide bonds formed by the four noncanonical cysteine residues found in all of the TIM proteins. Structures of the mouse TIM-4 (mTIM-4) protein showed that the CC′ and FG loops form a narrow cavity or pocket in which phosphatidylserine (PtdSer) binds (11). The phosphate of PtdSer coordinates at the base of the cavity with a metal ion that is also bound to conserved residues in the FG loop in all of the TIM proteins except mTIM-2, which does not bind PtdSer (11). This conserved binding pocket has been termed the metal ion-dependent ligand binding site (MILIBS).
The function of TIM-1 and TIM-4 as receptors for PtdSer confers important properties to cells expressing these molecules. PtdSer is normally localized to the inner leaflet of the plasma membrane, but it is redistributed and exposed on the outer membrane when the cell is activated or undergoes apoptosis. PtdSer on apoptotic cells provides a key signal that triggers cell engulfment (13, 14). The TIM gene family plays a critical role in regulating tolerance and immune responses (6, 15-19). Because rapid removal of apoptotic cells by phagocytes is critical for the maintenance of tolerance and prevention of autoimmunity (20, 21), TIM proteins may play critical roles in regulating tolerance by mediating the clearance of apoptotic cells. Indeed, we and others showed that TIM-1 and TIM-4 bind PtdSer exposed on apoptotic cells, and when expressed on certain cell types such as a peritoneal macrophage or 3T3 cells, both TIM-1 and TIM-4 facilitate engulfment of apoptotic cells (8, 9, 22).
The purpose of the current study was to examine the detailed interaction of TIM-3 with PtdSer and determine whether TIM-3 polymorphic variants differed in recognition of apoptotic cells. mTIM-3 was initially reported not to bind PtdSer (9), but a recent study found that TIM-3 binds PtdSer and that CD8+ DCs expressing TIM-3 phagocytose apoptotic cells (23). We found that both hTIM-3 and mTIM-3 are receptors for PtdSer, based on binding studies and a crystal structure, and that TIM-3–expressing cells bound or engulfed apoptotic cells expressing PtdSer, or both. Furthermore, allelic variants of TIM-3 had distinct binding and phagocytic properties, with the BALB/c allele demonstrating stronger binding and phagocytosis of apoptotic cells compared with the C.D2 Es-Hba (HBA) allele. Moreover, although fibroblasts or APCs expressing TIM-3 phagocytosed apoptotic cells, T cells expressing TIM-3 bound but did not engulf apoptotic cells. Together these findings suggest that TIM-3–expressing DCs, macrophages, and T cells can sense the presence of apoptotic cells but that the consequence of TIM-3 engagement of PtdSer depends on the specific polymorphic variants of and the type of cell expressing TIM-3.
All of the mouse experiments were approved by the Animal Care and Use Committee of the Dana-Farber Cancer Institute or Children’s Hospital (Boston, MA). BALB/cByJ mice were purchased from The Jackson Laboratory (Bar Harbor, ME).
To collect resident peritoneal cells, mice were sacrificed and the peritoneal cavities were infused with 3 ml ice-cold PBS with 0.3% BSA and 0.1 mM EDTA and washed in PBS containing 2% FBS. Cells were enriched for adherent cells by incubation in tissue culture dishes for 2 h and rinsed to remove nonadherent cells. Cells were detached using 0.5 mM EDTA in PBS.
Bone marrow-derived DCs (BMDCs) were generated from wild-type mice using GM-CSF as previously described with some modifications (24, 25). On day 7, some DCs were treated at 1.0 × 106 DCs/ml with cholera toxin (5 μg/ml). DCs were collected after 24 h and analyzed by flow cytometry.
Mouse primary airway epithelial cells were prepared as follows. Briefly, lungs were digested for 25 min in dissociation buffer containing pronase (1 mg/ml; Streptomyces griseus protease; Sigma-Aldrich, St. Louis, MO). Cells were filtered to remove debris, and macrophages were depleted by adherence to tissue culture dishes for 1 h at 37°C. Nonadherent cells were plated on plates precoated with BD Matrigel (BD Biosciences, San Jose, CA) in Airway Epithelial Cell Growth Medium (Promo Cell, Heidelberg, Germany). After 2 wk, >95% of the cells were cytokeratin-positive when stained with Cytokeratin 5 + 8 Ab RCK102 (Abcam, Cambridge, MA) (data not shown).
TIM-3 mAbs 7D11, 10G12, and 11G8 (all mouse IgG1) for hTIM-3 were made as described elsewhere (M. Pichavant, A.P.G. Silva, H.Y. Kim, H.-H. Lee, R.E. Sells, H. Nagumo, N. Kobayashi, S.E. Umetsu, Y.-L.E. Chim, V. Shaw, D.M. Dorfman, G.J. Freeman, D.T. Umetsu, and R.H. DeKruyff, submitted for publication) and recognized hTIM-3 but not hTIM-1 or hTIM-4 on transfected 300.19 cells and did not react with untransfected cells (Supplemental Fig. 1).
The human macrophage cell lines THP-1, MD, and KMA and the mouse macrophage cell lines RAW264.7, PMJ2R, and MH-S were obtained from American Type Culture Collection (Manassas, VA) and maintained in media recommended by American Type Culture Collection.
NIH 3T3, 300.19, or DO11.10 hybridoma cells were stably transfected by electroporation with a pEF6 plasmid containing mTIM-3 (HBA allele), mTIM-3 (BALB/c allele), mTIM-1 (BALB/c allele), or hTIM-3 or with pEF6 vector only. Cells were selected with blasticidin or puromycin and sorted twice by flow cytometry with specific Abs for mTIM-3 (RMT3-23-PE), hTIM-3 (11G8), or mTIM-1 (3B3).
Thymocytes isolated from 4- to 6-wk-old wild-type BALB/c mice were incubated with 10 μM dexamethasone (Sigma-Aldrich) in RPMI 1640 without FCS for 3 h. Cells were washed, and apoptosis was confirmed by annexin V-FITC and propidium iodide staining (BD Pharmingen, San Diego, CA). Eryptotic mouse RBCs were prepared as described (8).
Apoptotic thymocytes or transfected 300.19 cells were labeled with 4 μM PKH67 (Sigma-Aldrich; FL1 channel) or PKH26 (Sigma-Aldrich; FL2 channel) according to the manufacturer’s instructions. TIM-transfected 300.19 cells (10 × 104) or control human PD-L1–transfected 300.19 cells were incubated with apoptotic thymocytes (30 × 104) in RPMI 1640 containing 10% FBS at 37°C for 30–120 min, gently resuspended and analyzed by flow cytometry gating on the transfected cell population. For blocking experiments, transfected cells were preincubated with TIM mAb or isotype control mAb for 15 min at room temperature. Apoptotic thymocytes were added, and incubation was continued at 37°C for the indicated time. In some experiments, coincubation was performed in media containing EGTA (0.5–5 mM).
For measurement of phagocytosis by TIM-transfected 3T3 cells, 1.0 × 105 3T3 cells were seeded in six-well plates and cultured overnight. Apoptotic thymocytes were labeled with pHrodo (1.5 μM; Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions and added to the 3T3 cells in the ratio of 20:1. After 2–3 h of incubation at 37°C, wells were quickly washed three times with PBS containing 2% FBS and 0.1 mM EDTA to remove nonadherent apoptotic cells. 3T3 cells were detached with PBS containing 0.5 mM EDTA, resuspended in PBS containing 2% FBS and 0.1 mM EDTA, and analyzed by flow cytometry, gating on 3T3 cells by forward and side scatter. To evaluate phagocytosis by 300.19 or DO11.10 hybridoma cells, cells were cocultured with pHrodo-labeled apoptotic thymocytes for 2 h at 37°C, washed, and analyzed by flow cytometry.
For confocal microscopy, 3T3 cells were labeled with CellTracker Green (CMFDA) (2–4 μM) (Invitrogen), incubated with pHrodo-labeled apoptotic thymocytes, and washed as described above. Cells were fixed with 3.7% formaldehyde in PBS for 10 min at room temperature, washed once with PBS, and analyzed by confocal microscopy. DO11.10 hybridoma or 300.19 cells were labeled with CMFDA or CellTracker Orange (CMTMR) (Invitrogen), cocultured with apoptotic thymocytes or eryptotic RBCs labeled with CMTMR or PKH26 for 60 min at 37°C as indicated in the figure legend, and analyzed by confocal microscopy. Confocal micrographs in Figs. 6 and and77 were acquired by a C1 Plus-equipped TE2000-U microscope (Nikon, Melville, NY) using 488- and 543-nm lasers with 525/550- and 590/550-nm filter sets. Z-stack images were taken using a 60× plan apochromatic objective. Plane- and volume-rendered images were processed with Adobe Photoshop (version 11; Adobe, San Jose, CA) and EZ-C1 (version 3.8; Nikon Instruments, Melville, NY) software. Confocal images in Fig. 8 were acquired using a TE-2000 inverted microscope (Nikon) with attached Orca AG camera (Hamamatsu Photonics, Hamamatsu City, Japan) at ×100 magnification and 37°C in DMEM. Slidebook software (Intelligent Imaging Innovations, Philadelphia, PA) was used to create XZ and XY views from the confocal image stacks and to adjust the brightness and contrast for display.
Analytical flow cytometry was carried out using a FACSCanto (BD Biosciences), and the data were processed using the FlowJo software (Tree Star, Ashland, OR). For characterization of mAbs, untransfected or TIM-1–, TIM-3–, or TIM-4–transfected 300.19 cells were incubated with mAbs and stained with PE-conjugated goat anti-mouse or anti-rat Ig (Southern Biotechnology Associates, Birmingham, AL). For detection of TIM-3 expression on freshly isolated cells, cells were isolated and Fc receptors were blocked with excess anti-Fc receptor (mAb 2.4G2, purified from hybridoma HB197; American Type Culture Collection) for 20 min at 4°C prior to staining. Peritoneal cells were stained with Alexa Fluor 700-conjugated CD11b Ab (eBioscience, San Diego, CA) and costained with PE-conjugated TIM-3 mAb RMT3-23 (eBioscience), TIM-1 mAb 3B3 (6), TIM-4 mAb 21H12 (8), or isotype control (eBioscience).
The isolated mTIM-3 IgV domain was prepared by in vitro refolding of inclusion bodies produced in bacteria as described for other mTIMs (11, 12). The refolded protein has an N-terminal Met residue, residues 21–133 of the precursor protein (3), a thrombin recognition site, and two epitope tags at the C-terminal end. For crystallization, the recombinant mTIM-3 protein was thrombin-treated to release C-terminal tags and further purified by ion-exchange chromatography using 25 mM sodium acetate buffer pH 5.6. Complexes of the mTIM-3 IgV domain and the phospholipid were prepared by incubation of the protein (15 mg/ml) with 5 mM 1,2-dicaproyl-sn-glycero-3-phospho-l-serine (DCPtdSer) (Avanti Polar Lipids, Alabaster, AL) in 25 mM HEPES buffer (pH 7.0) with 5 mM CaCl2 and 100 mM NaCl. The complex was crystallized using sitting drops under paraffin oil at a protein concentration of 15 mg/ml and 22% polyethylene glycol 4000, 50 mM NaF, 5 mM CaCl2, and 0.1 M sodium citrate (pH 5.6). The crystals have two molecules in the asymmetric unit and 50% solvent content and belong to the R32 space group (Table I).
The crystal structure was determined by the molecular replacement method using the program Phaser (26) and the mTIM-3 structure without ligand as search model (Protein Data Bank [PDB] ID: 2OYP) (10). The structure was refined using the program PHENIX 1.4 (www.phenix-online.org). Models were adjusted manually during the refinement process. The N- and C-terminal five residues of the mTIM-3 protein were poorly defined and are missing in the final structure. All of the residues in the structure are in allowed regions of the Ramachandran plot. Fig. 2A was prepared with RIBBONS (www.cbse.uab.edu/ribbons), and the remaining figures of the structure were prepared with PYMOL (www.pymol.org).
Liposomes containing PtdSer and phosphatidylcholine (PC) were prepared as described elsewhere (11). They were diluted in PBS to a final phospholipid concentration of 10 mM and added (50 μl) to wells of 96-well plates for binding to plastic. Preparation of TIM-Fc fusion proteins and analysis of protein binding to plastic-bound PtdSer and PC liposomes were carried out for 1 h at 37°C as previously described (11). Background binding signal to wells lacking liposomes was subtracted from the OD determined in wells with liposomes. Specific binding of the proteins to wells coated with PtdSer was determined by subtracting the OD obtained in wells coated with PC. A range of protein concentrations as shown in Fig. 1B (0.5–20 μg/ml) was used for the analysis of PtdSer binding by mutant TIM-Fc proteins. Relative binding (mutant versus wild type) was determined from those protein concentrations giving a significant binding (OD > 0.1) and the average binding for all of those concentrations calculated in each experiment.
Structural and biochemical studies have shown that TIM-1 and TIM-4 bind to PtdSer (8, 11). Both TIM-1 and TIM-4 have MILIBS motifs with the amino acids Asn and Asp coordinating a metal ion, which in turn coordinates with the charged head of PtdSer (Fig. 1A, red) (11). Because these amino acids are conserved in TIM-3, as are the adjacent hydrophobic amino acids in the FG loop (WF in TIM-1 and TIM-4 and LM in TIM-3) (Fig. 1A, magenta), we asked whether the MILIBS in TIM-3 might also bind PtdSer.
We prepared mTIM-1-Fc, mTIM-2-Fc, mTIM-3-Fc, and mTIM-4-Fc fusion proteins and compared their binding to PtdSer in liposomes immobilized on plastic. As previously reported, TIM-2 does not bind PtdSer and served as a negative control (8, 9, 11). The mTIM-3 (BALB/c) protein bound to PtdSer but with a lower affinity than either mTIM-1 or mTIM-4 (Fig. 1B). Compared with TIM-1 or TIM-4, mTIM-3 had ~50% PtdSer binding affinity at longer incubation times and equilibrium conditions (Fig. 1C). All three mTIM proteins showed rapid kinetics in binding to PtdSer.
To gain further insight into the PtdSer binding site, we crystallized the isolated N-terminal IgV domain of BALB/c mTIM-3 in complex with a PtdSer compound having a short fatty acid moiety (DCPtdSer). Diffraction data extending to 3.0 Å resolution were used for structure determination (Table I and Materials and Methods; PDB ID: 3KAA, www.rcsb.org/pdb/home/home.do). As described for the other TIM protein structures (10-12), the IgV domain has longer β-strands in the GFC β-sheet than in the BED β-sheet, (Fig. 2A) and the folded conformation of the CC′ loop onto the GFC β-sheet is preserved (Fig. 2A, 2B). The bottom of the CC′ loop is bridged to the β-sheet by two disulfide bonds, whereas the tip is hydrogen-bonded to the conserved Arg and Lys residues on the F and G β-strands, respectively (Fig. 2A). The overall folding of the mTIM-3 polypeptide backbone with DCPtdSer is highly similar to the reported structure without ligand (root-mean-square deviation of 1.05 Å for Cα comparison of 108 residues) (Fig. 2B). However, there are relevant differences in some of the residues at the PtdSer binding site. The Trp residue at the tip of the CC′ loop is rotated and distant from the FG loop (Fig. 2B), allowing accommodation of the PtdSer molecule (Fig. 2C, left), which can then fit between the CC′ and FG loops that build the MILIBS cavity. In the structure of mTIM-3 without the ligand, this Trp faces toward the FG loop (Fig. 2B) and connects the CC′ and FG loops, closing the MILIBS cavity (Fig. 2C, right). There are also minor changes in the conformations of FG loop residues (Fig. 2B). The shift in the side chain of the Asn allows its coordination to the metal ion coordinated with the phosphate of PtdSer (see below).
There were marked differences in the electron density filling the MILIBS between the two molecules of the asymmetric unit (data not shown), and a clearly bound DCPtdSer molecule appeared only in molecule B (shown in Fig. 3A). Electron density contiguous to the phosphate of DCPtdSer in molecule B was modeled as a metal ion, also seen in the mTIM-4/DCPtdSer complex structure solved at a higher resolution (11). The two fatty acid side chains linked to the glycerol of PtdSer were poorly defined in the structure, and they have not been included in the refined structure (Fig. 3A). The hydrophilic moiety of PtdSer penetrated into the pocket. The hydrophobic regions of the phospholipid were between the aromatic Trp residue in the tip of the CC′ loop and the hydrophobic Leu and Met residues in the FG loop. The binding mode of PtdSer to mTIM-3 was very similar to that described for mTIM-4 (11), although mTIM-4 has a polar Asn instead of a Trp residue in the tip of the CC′ loop and two hydrophobic (WF) residues in the tip of the FG loop (Fig. 1A). Moreover, we determined with the Computed Atlas of Surface Topography of Proteins server (http://sts-fw.bioengr.uic.edu/castp/) that the volume of the binding pocket in mTIM-3 (170 Å3) was smaller and thus less accessible than that in the mTIM-4 structure (200 Å3). These differences in the MILIBS may account for the observed differences in PtdSer binding affinity among the TIMs. The phosphate group of PtdSer was coordinated to a metal ion (green sphere in Fig. 3A) that was also bound to two side chain oxygen atoms of Asn and Asp residues as well as two polypeptide backbone carbonyl atoms in the FG loop. The serine moiety of PtdSer was less defined, likely because of contact with a neighboring molecule in the crystal (data not shown). The amino group of the serine interacts with the conserved Asp residue that coordinates with the metal ion, whereas the carboxylate of PtdSer bound to the serine residue in the tip of the CC′ loop, similar to the reported mTIM-4/PtdSer complex structure (11).
To verify the contribution of the MILIBS residues in the CC′ and FG loops in PtdSer recognition, we carried out site-directed mutagenesis of mTIM-3 and compared PtdSer binding activities (Fig. 3B). The substitution of hTIM-3 residues (VFE) at the tip of the CC′ loop for those of the mouse protein (WSQ) significantly decreased the PtdSer binding activity (hCC′ mutant). The minimal binding seen with the W/A mutant showed that this is a key PtdSer binding residue at the CC′ loop of mTIM-3 (Fig. 3B). Alanine substitution of FG loop residues engaged in metal ion coordination (ND/AA) or interacting with the hydrophobic moiety of PtdSer (LM/AA) abolished mTIM-3 binding to PtdSer, as was observed with mutation of the homologous residues in mTIM-1 and mTIM-4 (11). Therefore, both metal ion coordination and the interaction of the hydrophobic residues with the fatty acid moiety of PtdSer are critical for efficient recognition of PtdSer by TIM proteins.
The critical contribution of the FG loop hydrophobic residues to PtdSer binding shown here for mTIM-3 (Fig. 3) and previously for mTIM-1 and mTIM-4 (8, 11) as well as the contacts of these residues with the fatty acid moiety of PtdSer suggested that they must penetrate the lipid bilayer as modeled for the IgV-like domain of mTIM-4 in Fig. 4A. The polar CC′ loop of mTIM-4 and mTIM-1 may interact with the hydrophilic region of the phospholipids, whereas the aromatic Trp residue at the tip of the CC′ loop of mTIM-3 would penetrate deeper into the bilayer. Indeed, deletion of the Trp side chain almost abolished mTIM-3 binding to PtdSer (Fig. 3B), whereas substitution of the Asn residue at the tip of the mTIM-4 CC′ loop for an Asp residue significantly decreased binding to PtdSer (Fig. 4B).
This model of TIM binding to PtdSer in a membrane suggests that regions of the TIM molecule to each side of the MILIBS must interact with the surface of the cell membrane. The model predicts that the BC loop would be proximal to the membrane surface, and we noticed the presence of an Arg residue in the BC loop of mTIM-1 and mTIM-4, which could mediate electrostatic interactions with the charged phosphate head of membrane phospholipids (Fig. 4A). Substitution of this Arg residue in mTIM-1 (R22A) or mTIM-4 (R25A) decreased PtdSer binding activity ~50%, showing a contribution of the BC loop to recognition of PtdSer in a membrane (Fig. 4B). Thus, residues in the BC loop adjacent to the MILIBS appear to modulate the affinity of the TIM proteins for PtdSer in a membrane structure.
There are seven amino acid differences between the BALB/c and the HBA alleles of TIM-3 (3), clustered in either the β-strand A or the BC loop (Fig. 1A). To analyze the contribution of those polymorphisms to PtdSer binding, we compared the binding activities of HBA and BALB/c TIM-3 proteins with those of BALB/c TIM-3 proteins containing either the four polymorphic amino acids of the HBA β-strand A (called HBA1) or the three polymorphic amino acids of the HBA BC loop (called HBA2) to PtdSer in liposomes immobilized on plastic (Fig. 4C). Binding activity of the HBA TIM-3 was ~50% lower than that of the BALB/c TIM-3. Analysis of the HBA1 and HBA2 mutants showed that the reduced binding was accounted for by the polymorphisms in the BC loop (HBA2) of the IgV domain. The BC loop is also engaged in mTIM-1 and mTIM-4 binding to PtdSer in a membrane (Fig. 4B) and thus is important in all of these TIM proteins. The approximate affinities of the TIM-3-Fc proteins for PtdSer in liposomes, determined from the protein concentration at 50% of binding saturation (data not shown), were 45, 80, and 130 nM, respectively, for BALB/c, HBA, and human TIM-3. The affinity of mTIM-3-Fc for PtdSer, determined on microtiter plates coated with PtdSer without liposomes, carried out as previously reported for TIM-4-Fc (9), was ~10 nM (data not shown) as compared with an affinity of 2 nM for TIM-4 (9). In the absence of liposomes, there was no difference in binding of HBA and BALB/c TIM-3-Fc to PtdSer (data not shown). This is consistent with the location of the TIM-3 polymorphisms in the BC loop, a region that interacts with the membrane surface but not directly with PtdSer.
To determine whether the BALB/c or HBA alleles of TIM-3 differed in recognition of apoptotic cells, we transfected BALB/c or HBA TIM-3 into 300.19 cells (Fig. 5A) and examined binding to apoptotic cells. Following coincubation, flow cytometry was used to measure binding of PKH67-labeled apoptotic thymocytes to PKH26-labeled TIM-3 transfectants. Apoptotic thymocytes and TIM-3–transfected 300.19 cells could be distinguished by both fluorescence and size (Supplemental Fig. 1). The percentage of TIM-3 transfectants that bind one or more apoptotic thymocytes was determined by gating on PKH26-labeled TIM-3–transfected 300.19 cells (>99% FL-2+) and analyzing PKH67 fluorescence (bound apoptotic thymocytes) (Fig. 5B and Supplemental Fig. 1). BALB/c TIM-3–transfected 300.19 cells were more effective in binding apoptotic thymocytes than HBA TIM-3–transfected 300.19 cells (Fig. 5B, 5C). Neither BALB/c nor HBA TIM-3–transfected 300.19 cells bound live thymocytes (Fig. 5C). BALB/c TIM-3–transfected cells consistently demonstrated greater binding of apoptotic cells than HBA TIM-3 transfectants over a period of 30–120 min (Fig. 5D). Control PD-L1–transfected 300.19 cells showed minimal binding of apoptotic cells (Fig. 5C–E). TIM-3 mAb RMT3-23 inhibited the binding of apoptotic thymocytes by HBA TIM-3–transfected cells to the level of control transfected cells and reduced the binding of apoptotic thymocytes by BALB/c TIM-3–transfected cells by 60%, confirming the role of TIM-3 in this process (Fig. 5E).
The structure revealed the presence of a metal ion coordinated to conserved FG loop residues and the phosphate of DCPtdSer (Fig. 3A), where the side chain of Asn98 adopts a conformation suited for coordination of the metal ion (Figs. 2B, ,3A).3A). Binding of apoptotic cells by BALB/c TIM-3–transfected 300.19 cells was greatly reduced in the presence of 0.5 mM EGTA and eliminated in media containing ≥1 mM EGTA (Fig. 5F). Similarly, binding of apoptotic cells by hTIM-3–transfected 300.19 cells was also eliminated in the presence of 1 mM EGTA (Supplemental Fig. 1). Because EGTA specifically binds calcium and not sodium or other divalent cations, this indicates that calcium is the metal ion required for binding of TIM-3 to PtdSer.
To determine whether the HBA and BALB/c TIM-3 variants differed in their capacities to phagocytose apoptotic cells, we prepared stably transfected HBA and BALB/c TIM-3 3T3 cells (Fig. 6A). Flow cytometry was used to quantify the engulfment of pHrodo-labeled, apoptotic thymocytes. The pH-sensitive dye pHrodo strongly fluoresces red in acidic conditions, such as those in a phagosome or endosome, but very weakly at neutral pH. Therefore, red fluorescence indicates true engulfment as opposed to cell surface binding. Apoptotic thymocytes were more efficiently phagocytosed by the BALB/c TIM-3–transfected 3T3 cells (90.4% FL-2+) than by the HBA TIM-3–transfected 3T3 cells (45.7% FL-2+) and were not phagocytosed by vector-transfected 3T3 cells (p < 0.01) (Fig. 6B, 6C). The mean fluorescence intensity of BALB/c TIM-3 3T3 cells after phagocytosis was also significantly higher than that of HBA TIM-3 3T3 cells (p < 0.02) (Fig. 6C). In addition, engulfment was confirmed using confocal microscopy imaging of CMFDA-labeled BALB/c and HBA TIM-3–transfected 3T3 cells (green) and pHrodo-labeled thymocytes (red) (Fig. 6D), indicating localization in an acidic compartment. BALB/c TIM-3–transfected cells contained 8.8 apoptotic thymocytes per 3T3 cell compared with 3.8 apoptotic thymocytes per HBA TIM-3–transfected 3T3 cell (Fig. 6E). To confirm that PtdSer is the molecule recognized by TIM-3 for phagocytosis of apoptotic cells, we prepared phospholipid liposomes containing various phospholipids and tested their capacities to inhibit phagocytosis by TIM-3 transfectants. Liposomes containing equal amounts of PtdSer and PC reduced phagocytosis in a concentration-dependent manner, with complete inhibition at 100 μM total phospholipids (Fig. 6F). Liposomes containing PC did not inhibit phagocytosis. These data show that BALB/c TIM-3 binds PtdSer and apoptotic cells better than HBA TIM-3 and phagocytoses apoptotic cells more efficiently (Figs. 4--66).
We then examined the capacity of hTIM-3 to bind and mediate phagocytosis of apoptotic cells. Fig. 7A shows surface expression of hTIM-3 and hTIM-1 on transfected 300.19 cells and reactivity with specific mAbs. hTIM-3–transfected 300 cells bound apoptotic cells, and this binding was inhibited by hTIM-3 mAb 7D11 (Fig. 7B, 7C). hTIM-1–transfected 300.19 cells bound apoptotic cells as expected (8), and this binding was more robust, as demonstrated by a higher percentage of apoptotic cells bound to hTIM-1– than hTIM-3–transfected cells (Fig. 7B). hTIM-3–transfected cells did not bind live cells (Supplemental Fig. 1). Liposomes containing equal amounts of PtdSer and PC reduced hTIM-3–mediated binding of apoptotic thymocytes in a concentration-dependent manner with complete inhibition at 100 μM total phospholipids (Fig. 7D). Liposomes containing PC did not inhibit binding, confirming the specific recognition of PtdSer. We compared the binding of apoptotic cells by mTIM-3–allelic variants and hTIM-3. BALB/c TIM-3–transfected cells demonstrated the highest binding of apoptotic cells, followed by hTIM-3 and HBA TIM-3 (Fig. 7E and data not shown).
Apoptotic thymocytes were efficiently phagocytosed by hTIM-3–transfected 3T3 cells (expression shown in Supplemental Fig. 2) but not by vector-transfected control cells (Fig. 7F). Live thymocytes were not phagocytosed by hTIM-3–transfected 3T3 cells (data not shown). Engulfment was confirmed using confocal microscopy imaging of CMFDA-labeled hTIM-3–transfected 3T3 cells (green) and pHrodo-labeled thymocytes (red) (Fig. 7G, 7H).
We identified five coding region polymorphisms in hTIM-3 (Genbank), one in the IgV domain and four in the mucin domain (Fig. 7I). The IgV domain polymorphism is predicted to be at the beginning of the BC loop, a region shown here to be adjacent to the MILIBS and involved in interaction with the membrane surface.
Although the phagocytic abilities of cells such as macrophages, DCs, or TIM-transfected 3T3 cells are well documented, the capacity of lymphoid cells to function as phagocytic cells is less well characterized. Because TIM-1 and TIM-3 are expressed on CD4+ T cell subsets (3, 15), we investigated the capacity of TIM-expressing lymphoid cells to phagocytose apoptotic cells. The hTIM-3 transfected 300.19 cells (a pre-B cell line) formed conjugates with but did not phagocytose apoptotic thymocytes (Fig. 8A, 8B). Untransfected 300.19 cells did not bind apoptotic thymocytes (Fig. 5C–E and data not shown). We also tested whether T cells expressing TIM-3 have the capacity to phagocytose apoptotic cells. CMFDA-labeled BALB/c TIM-3–transfected DO11.10 T hybridoma cells (Supplemental Fig. 2) avidly formed conjugates with but did not phagocytose CMTMR-labeled apoptotic thymocytes or eryptotic RBCs (Fig. 8C, 8D).
We examined stable TIM-1 transfectants of DO11.10 hybridoma cells (Supplemental Fig. 2) to determine whether the inability of TIM-3–transfected 300.19 or DO11.10 hybridoma cells to phagocytose was specific for TIM-3 or a general property of lymphoid cells. TIM-1–transfected DO11.10 hybridoma cells (red) avidly formed conjugates with but did not phagocytose apoptotic thymocytes (green) as demonstrated using confocal microscopy (Fig. 8E). TIM-1 mAb 3B3 blocked the binding of apoptotic cells by TIM-1–transfected DO11.10 cells, indicating that binding was mediated by TIM-1 recognition of PtdSer-expressing apoptotic cells (Fig. 8F). Untransfected DO11.10 T hybridoma cells, which did not express detectable levels of cell surface TIM-1 or TIM-3, showed minimal binding of apoptotic cells by flow cytometry (Fig. 8F). We also showed that normal CD4+ T cells from a TIM-1 transgenic mouse formed conjugates with but did not phagocytose apoptotic cells (Fig. 8G, 8H). Together these data indicate that TIM-1– or TIM-3–expressing lymphoid cells conjugate with but do not phagocytose apoptotic cells and suggest that recognition of apoptotic cells may play a role in T cell effector function.
We next examined TIM-3 expression on a variety of different cells. Mouse peritoneal resident macrophages expressed low levels of TIM-3, high levels of TIM-4, and no detectable TIM-1 (Fig. 9A). BMDCs demonstrated low levels of surface TIM-3 expression but not detectable levels of TIM-4 or TIM-1 (Fig. 9B and data not shown). Exposure of BMDCs to cholera toxin for 24 h enhanced TIM-3 but not TIM-4 or TIM-1 expression (Fig. 9B). Cholera toxin-treated BMDCs phagocytosed apoptotic thymocytes, and blockade with TIM-3 mAb reduced phagocytosis by ~50% (Fig. 9C), indicating that the expression of TIM-3 contributed to the phagocytic capacity of the BMDCs.
Primary bronchial epithelial cell lines expressed moderate amounts of cell surface TIM-3 but not TIM-1 or TIM-4 (Fig. 9D). In mouse macrophage cell lines, TIM-3 was expressed strongly in RAW264.7 and very weakly in MH-S and was not detectable in PMJ2R (Fig. 9E). In human cell lines, TIM-3 was weakly expressed on the KMA and MD macrophage cell lines but not on the THP-1 monocytic leukemia cell line (Fig. 9F). PMA can induce THP-1 cells to differentiate from monocytic leukemia cells into macrophages (27). Stimulation of THP-1 with PMA (5 ng/ml) induced expression of TIM-3 by 24 h, and expression increased at 48 and 72 h (Fig. 9G). These results indicate that TIM-3 is expressed on a diverse range of macrophage and DC populations in mouse and human. TIM-3 expression can be upregulated following exposure to inflammatory stimuli and contributes to the phagocytic capacity of the cells.
The Tim gene family was identified using a congenic mouse model in which polymorphisms in both TIM-1 and TIM-3 were associated with Th1/Th2 differentiation and AHR between BALB/c and congenic HBA mice. The HBA alleles of TIM-1 and TIM-3 originated from a DBA/2 ancestor (28) and are identical to C57BL/6 alleles. Here, we show that the BALB/c allele of TIM-3 bound PtdSer and mediated the recognition and phagocytosis of apoptotic cells more efficiently than the HBA allele of TIM-3. Polymorphisms in the BC loop of the IgV domain were responsible for these differences. This is the first report of a functional difference between the BALB/c and HBA alleles of TIM-3 or TIM-1 and suggests that functional differences in the TIM-3 alleles may contribute to the Th1/Th2 phenotypic differences between the strains. Further support for a role for TIM-3 in these differences is provided by the observation that blockade of TIM-3 with mAb in BALB/c modulates Th1/Th2 responses and AHR (29). We also showed that cells expressing hTIM-3 bound and phagocytosed apoptotic cells. Similarly in humans, one study has linked asthma to both TIM-1 and TIM-3 polymorphisms (30), suggesting that TIM-1 and TIM-3 may contribute together to Th1 and Th2 differentiation. However, precise mechanisms whereby recognition of apoptotic cells by TIM-3 variants mediates differences in immune responsiveness require further investigation.
In this study, we determined the structure of the mTIM-3/PtdSer complex, which showed that except for mTIM-2, the TIM proteins have a MILIBS motif designed for recognition of PtdSer, thus demonstrating a conserved binding mode in TIM family members. In the two TIM/PtdSer structures solved to date, the hydrophobic region of PtdSer interacts with hydrophobic residues on the tips of loops building the MILIBS, whereas the hydrophilic moiety of PtdSer penetrates into the binding pocket. The metal ion coordination must provide the specificity for phospholipid recognition. The ligand binding cleft on the GFC β-sheet defined in the ligand-free mTIM-3 structure overlaps with the PtdSer binding pocket and is distinct from the galectin-9 binding site on an N-linked glycan on the opposite side of the IgV domain from the MILIBS (10). Regions near the MILIBS, such as the BC loop, also contribute to the binding of TIM proteins to PtdSer in a membrane. According to our model, this additional interaction site is likely to interact with the membrane surface and not directly with PtdSer, such as proposed for the binding of protein kinase C to PtdSer (31). Differences in the cavity size and the membrane-interacting residues, at both the tips of the CC′ and FG loops and the regions surrounding the MILIBS, could modulate the TIM binding affinity for PtdSer and account for the differences observed among TIM proteins. The highly conserved TIM-PtdSer interactions observed in the TIM/PtdSer complex structures prove that the TIM proteins are pattern recognition receptors specialized for recognition of the PtdSer death signal.
Identification of TIM-3 as an additional receptor for PtdSer adds to the versatility of the host in recognition of apoptotic cells. PtdSer is normally localized to the inner leaflet of the plasma membrane, but it is redistributed or exposed to the outer leaflet when the cell undergoes apoptosis, injury, or cell activation. Recognition of PtdSer on the surfaces of apoptotic cells provides a key signal to the phagocyte that triggers engulfment of apoptotic cells, which can result in potent anti-inflammatory effects (32, 33) and protection from autoimmune disease (34). However, phagocytosis of apoptotic cells in association with danger signals can induce immune responses and inflammation (35).
DCs, macrophages, and human monocytes (36), including microglial cells in the CNS, all express TIM-3. These APCs have been suggested to regulate immunity by phagocytosing apoptotic cells and cross-presenting Ags in tolerogenic pathways (37), which may be mediated in part by TIM-3 (23). Consistent with this idea, mice deficient in TIM-3 were resistant to development of transplant tolerance (38). In experimental autoimmune encephalomyelitis and multiple sclerosis, there is an expanded number of TIM-3+ microglial cells, which may function to down modulate disease, because treatment with a blocking TIM-3 mAb-enhanced demyelination and the severity of experimental autoimmune encephalomyelitis (15, 39) and induced TNF-α production by microglial cells (36). In some systems, engagement of TIM-3 on APCs may enhance inflammation by increasing inflammatory cytokine production; however, this may be due to coactivation with TLR agonists (36). Considering that myelin is a membrane with normal compartmentalization of PtdSer (40), future studies on multiple sclerosis will need to clarify the role of TIM-3+ microglial cells in the phagocytosis of apoptotic myelin and cross-presentation of myelin Ags to T cells.
In contrast to APCs, we showed that T cells that express TIM-3 form conjugates with but do not phagocytose apoptotic cells. The failure of T cells to engulf apoptotic cells was probably not due to the small size of the T cells, because small eryptotic RBCs or fragments of U937 cells were not phagocytosed and large T cell hybridoma cells or pre-B cells transfected with TIM-3 were still unable to phagocytose apoptotic cells. It is possible that lymphocytes lack some cellular machinery required for engulfment or alternatively have an active “do not eat” receptor system (41). Cross-linking of TIM-3 on T cells by apoptotic cells instead may provide a proapoptotic signal to the T cell, as is induced by binding of TIM-3 on Th1 or Th17 cells by galectin-9, another ligand of TIM-3 (7, 42). TIM-3 has an intracellular tyrosine phosphorylation motif, and cross-linking TIM-3 with mAb induces tyrosine phosphorylation. The induction of a proapoptotic signal by TIM-3 in T cells would also be consistent with a negative regulatory role for TIM-3 in immunity (23). This effect may depend on the affinity of the interaction, which we have shown is different for BALB/c versus HBA alleles.
In contrast, TIM-1 is expressed on Th2 cells, invariant NK T cells, and mast cells (43) and costimulates T cell activation and cytokine production (6, 44). TIM-1–expressing T cells avidly form conjugates with apoptotic cells, suggesting that this interaction provides a signal to the T cell that could result in T cell expansion, cytokine production, or both.
The finding that TIM gene family members TIM-1, TIM-3, and TIM-4, which have distinct patterns of expression on distinct cell types or on cells at specific stages of activation or differentiation, are a family of pattern recognition receptors for PtdSer suggests that the TIM proteins provide a functional repertoire for recognition of apoptotic cells. For example, TIM-1 may bind PtdSer on apoptotic cells and mediate T cell activation, whereas TIM-3 on T cells may mediate T cell elimination, and TIM-4 on APCs may mediate apoptotic cell clearance resulting in tolerance. Previous work that identified several receptors mediating apoptotic cell recognition and clearance by phagocytes (45) led to the speculation that the repertoire might provide specificity in the phagocyte response. However, PtdSer receptors such as milk fat globule–EGF-factor 8 and growth arrest-specific gene 6 (46) are widely expressed in somatic cells and do not appear to specify phagocyte behavior following phagocytosis of apoptotic cells (21). We suggest that the TIM molecules evolved as a family of pattern recognition receptors for PtdSer that determine whether apoptotic cell recognition leads to immune activation or tolerance, depending on the TIM molecule engaged and the cell type on which it is expressed.
In summary, we have shown using structural and functional approaches that TIM-3 is a receptor for PtdSer and that polymorphic variants of TIM-3 differ functionally in their recognition of PtdSer and clearance of apoptotic cells. These findings establish a new paradigm for TIM proteins as PtdSer receptors and unify the function of the TIM gene family, which has been associated with asthma and autoimmunity and shown to modulate peripheral tolerance. Demonstration of functional differences in TIM-3 alleles has important implications for understanding of autoimmune disease mechanisms and development of therapeutic approaches.
We thank Sheena Tin for excellent technical assistance, the expert help of Jessica Wagner and The Harvard Digestive Diseases Center Imaging Core, and Barbara Osborne for the DO11.10 hybridoma cells. We acknowledge the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities through the MX-861 BAG project.
This work was supported by National Institutes of Health Grants P01 AI054456 (to D.T.U., G.J.F., R.H.D., and G.G.K.) and HL069507 (to R.H.D.) and Ministerio de Ciencia e Innovación Grants BFU2005-05972 and BFU2008-00971 (to J.M.C.).
Disclosures The authors have no financial conflicts of interest.
The structure presented in this article has been submitted to the Protein Data Bank under accession number 3KAA.
The online version of this article contains supplemental material.