TIM-4 structures identify a Metal Ion-dependent Ligand Binding Site where phosphatidylserine binds

doi:10.1016/j.immuni.2007.11.008

Immunity. Author manuscript; available in PMC 2008 December 1.

Published in final edited form as:

Immunity. 2007 December; 27(6): 941–951.

doi: 10.1016/j.immuni.2007.11.008

PMCID: PMC2330274

NIHMSID: NIHMS36632

TIM-4 structures identify a Metal Ion-dependent Ligand Binding Site where phosphatidylserine binds

Cesar Santiago,¹ Angela Ballesteros,¹ Laura Martinez-Muñoz,¹ Mario Mellado,¹ Gerardo G. Kaplan,² Gordon J. Freeman,³ and José M. Casasnovas¹^*

¹ Centro Nacional de Biotecnologia, CSIC, Campus Universidad Autonoma, 28049 Madrid, Spain

² Center for Biologics Evaluation and Research, Food and Drug Administration, Bethesda, MD 20892, USA

³ Department of Medical Oncology, Dana-Farber Cancer Institute, Department of Medicine, Harvard Medical School, Boston, MA 02115, USA

* Corresponding author, e-mail: jcasasnovas/at/cnb.uam.es, Phone: 34 915854917, Fax: 34 915854506

The publisher's final edited version of this article is available at Immunity

See other articles in PMC that cite the published article.

Publisher's Disclaimer

Abstract

The T-cell immunoglobulin and mucin domain (TIM) proteins are important regulators of T cell responses. They have been linked to autoimmunity and cancer. Structures of the murine TIM-4 identified a Metal Ion-dependent Ligand Binding Site (MILIBS) in the immunoglobulin (Ig) domain of the TIM family. The characteristic CC’ loop of the TIM domain and the hydrophobic FG loop shaped a narrow cavity where acidic compounds penetrate and coordinate to a metal ion bound to conserved residues in the TIM proteins. The structure of phosphatidylserine bound to the Ig domain showed that the hydrophilic head penetrates into the MILIBS and coordinates with the metal ion, while the aromatic residues on the tip of the FG loop interacted with the fatty acid chains and could insert into the lipid bilayer. Our results also revealed a significant role of the MILIBS in trafficking of TIM-1 to the cell surface.

Keywords: asthma, autoimmunity, metal ions, phospholipids, crystallography

INTRODUCTION

The TIM family is involved in the regulation of immune responses by modulating effector Th1 and Th2 cell functions (Kuchroo et al., 2006). Whereas TIM-1 and TIM-2 are found in Th2 cells, TIM-3 is preferentially expressed in Th1 cells. TIM-4 has been detected in antigen presenting cells and may function as a natural ligand of TIM-1 (Meyers et al., 2005). Crosslinking of murine mTIM-1 on the cell surface enhanced T-cell activation and proliferation (Meyers et al., 2005; Umetsu et al., 2005). Diverging from mTIM-1, mTIM-2 inhibits Th2 responses (Chakravarti et al., 2005; Rennert et al., 2006) and binds to heterologous ligands such as semaphorin 4A and ferritin H (Chen et al., 2005; Kumanogoh et al., 2002). mTIM-3 provides a negative costimulatory signal that leads to immune tolerance (Sabatos et al., 2003; Sanchez-Fueyo et al., 2003), while binding of mTIM-3 to galectin-9 triggers Th1 apoptosis and the inhibition of Th1 responses (Zhu et al., 2005).

The TIM family has been associated with immune-related diseases, cancer and viral infections. TIM-1 has been usurped by the hepatitis A virus (HAV) for cell entry in monkeys and humans (Feigelstock et al., 1998; Kaplan et al., 1996). HAVCR1/TIM-1 is an important asthma determinant gene in humans (McIntire et al., 2003) and its expression is upregulated in acute kidney diseases and renal carcinoma (Han and Bonventre, 2004; Vila et al., 2004). The TIM gene family is located in a genomic locus linked to autoimmune disease and asthma both in mouse and humans (Kuchroo et al., 2006). These genes were linked to an airway hyperreactivity regulatory locus and certain genetic variants of mTIM-1 and mTIM-3 were associated with the development of asthma in mouse models (McIntire et al., 2001).

The TIM are type I membrane proteins with an N-terminal immunoglobulin domain followed by a heavily glycosylated mucin domain in the extracellular region, a single transmembrane region and a cytoplasmic tail with tyrosine phosphorylation motifs except in the human and mouse TIM-4. Whereas sequence identity among the Ig domains is high (40–60%), there are large differences in the length of the mucin domains (Kuchroo et al., 2003). Crystal structures of the Ig domains of the murine mTIM-1, mTIM-2 and mTIM-3 revealed an Ig fold belonging to the V set (Cao et al., 2007; Santiago et al., 2007). A folded CC’ loop disulphide linked to the GFC β-sheet by four Cys residues characteristic of the TIM proteins is a distinctive structural feature of the IgV domain in the family. This loop appeared structurally connected to the FG loop in the mTIM-1 and mTIM-3 structures, building up a CC’FG epitope onto the GFC β-sheet (Cao et al., 2007; Santiago et al., 2007). In contrast, in mTIM-2 the CC’ and FG loops have a distinct conformation, so that the domain lacks the structural epitope (Santiago et al., 2007). The structures provided insights on ligand recognition by the TIM family. HAV virus and an unidentified mTIM-3 ligand appear to bind to the GFC face of the IgV domain (Cao et al., 2007; Santiago et al., 2007), while the opposite BED face was used in intercellular interactions among receptors of the TIM family (Santiago et al., 2007). Even though homophilic TIM-1 binding and homotypic TIM-TIM receptor interactions engaged the N-terminal IgV domains, high affinity binding of TIM-1 to either TIM-1 or TIM-4 required the mucin domains (Meyers et al., 2005; Santiago et al., 2007). It has been proposed that TIM-TIM receptor binding required a combinatorial epitope built by both IgV and mucin domains (Wilker et al., 2007). Both domains were also needed for efficient neutralization of HAV by soluble monkey TIM-1 (Silberstein et al., 2003).

Crystal structures of the mTIM-4 IgV domain presented here identified a distinctive ligand binding pocket with a metal ion coordination site. The physiological ligand phosphatidylserine bound to the cavity and coordinated to the metal. Mutation of protein residues engaged in metal ion coordination or building up the binding pocket in the IgV domain affected TIM functions and proved the relevance of this site in ligand recognition.

RESULTS

The crystal structure of the N-terminal IgV domain of mTIM-4

The isolated N-terminal domain of mTIM-4 was obtained essentially as described for the homologous mTIM-1 and mTIM-2 domains (Santiago et al., 2007). An initial crystal form (X1) diffracting at about 3 Å resolution was prepared with the crystallization conditions used for mTIM-1 (Experimental Procedures and Table 1). The crystal structure was solved by the molecular replacement method (Experimental Procedures). The IgV domain had significantly longer β-strands in the GFC than in the BED β-sheet (Fig. 1A). The long CC’-loop folded back onto the GFC β-sheet as in the structures of the homologous TIM domains (Cao et al., 2007; Santiago et al., 2007). The bottom of the loop was bridged to the β-sheet by two disulphide bonds, while the tip was hydrogen bonded to the conserved Arg and Lys residue on F and G β-strand, respectively (Fig. 1A). The interactions of the tip of the loop with the basic residues were also observed in the mTIM-1 and mTIM-3 structures and they were absent in mTIM-2 (Cao et al., 2007; Santiago et al., 2007). Therefore, the conformation of the CC’-loop in the mTIM-4 IgV domain structure was identical to that in mTIM-1 (Fig. 1B), whereas some structural variability was observed in the BC and FG loops.

Table 1

Data Collection and Refinement Statistics.

Figure 1

Crystal structure of the N-terminal domain of mTIM-4 and conformation of the CC’FG loop epitope

In the mTIM-1 structure the CC’ and FG loops were connected by a Phe residue that is conserved in the FG-loop of the mTIM-4 IgV domain (Fig. 1B, 1C). Hydrophobic residues also joined the loops in the mTIM-3 domain (Cao et al., 2007). However, the loops appeared disconnected in the mTIM-4 structure because of a conformational switch in the Phe side chain, which was distant from the CC’-loop. The conformation of the Asn residue conserved in the FG loop of the TIM proteins was also switched (Fig. 1B). A narrow cavity built by the tip of the CC’ loop and the interacting basic residue at the bottom and the aromatic Trp and Phe residues at the top of the FG loop was revealed by a surface representation of the mTIM-4 IgV domain structure (Fig. 1C). Some electron density was observed inside the cavity, which could come from either an alternate conformation of the Phe residue or a bound ion from the crystallization buffer. Thus, the structure of the X1 crystal form of mTIM-4 revealed a distinct conformation for the CC’FG loop epitope from that shown by the mTIM-1 structure, but the relevance of those changes was unclear.

A metal ion coordination site in the FG-loop of the mTIM-4 IgV domain

To get further insights on the conformation of the CC’FG epitope in the TIM IgV domains from crystal structures, we searched for new crystal forms with both the mTIM-1 and the mTIM-4 proteins. Crystals diffracting at 2.2 Å resolution were prepared with the mTIM-4 IgV domain using potassium sodium tartrate as precipitant (crystal form X2, Table 1). The structure was almost identical to that of the X1 crystals (Supplementary Fig. 1). The Phe residue on the FG-loop pointed away from the CC’-loop, so that both loops were disconnected (Fig. 2A). The high resolution of the structure allowed the identification of an electron density filling up the cavity as a tartrate ion present at high concentration in the crystallization conditions. The tartrate ion lay below the Trp residue of the FG-loop and was hydrogen bonded to several protein residues (Fig. 2B). A second electron density near the carboxylate group of the tartrate was identified as a sodium or calcium metal ion by the WaSP program (Nayal and Cera, 1996). The metal ion was coordinated to the tartrate, two main chain and two side chain oxygens of an Asn and Asp residues in the FG loop, and to a water molecule (Fig. 2B). The coordination number (6) and the average coordination distance (2.5 Å) are usual for both sodium and calcium ions, while the distances to potassium are over 2.7 Å (Harding, 2002). A sodium ion was modelled because of the lack of calcium in the crystallization conditions and the lower B factor obtained with sodium in the structure refinement. Temperature B factor for potassium was similar to that obtained with calcium (28) and higher than that of the refined sodium ion (15), which was close to the tartrate B factor (Table 1).

Figure 2

A Methal Ion coordination site in the FG loop of the mTIM-4 IgV domain

The Asn and Asp protein residues engaged in metal coordination are conserved in all TIM proteins except in mTIM-2 (highlighted in red in Fig. 2C), which had a distinct CC’ and FG loop conformation and therefore lacked the pocket where the tartrate ion penetrates in the mTIM-4 IgV domain (Santiago et al., 2007). The high conservation of the residues involved in metal ion coordination as well as the conserved conformation of the CC’ loop suggested that this Metal Ion-dependent Ligand Binding Site (MILIBS) will be conserved in the TIM family. Coordination of the metal ion and ligand binding to the pocket required an open conformation of the CC’FG loop epitope, where both loops must be disconnected (Fig. 2). No metal ions were identified by the WaSP program in the mTIM-1 and mTIM-3 structures (pdb access codes 2OR8 and 2OYP, respectively), where the cavity was closed and the Asn residue at the FG loop was in a conformation not suited for coordination of the metal ion (Fig. 1B) (Cao et al., 2007; Santiago et al., 2007).

Binding of phosphatidylserine to the MILIBS

TIM-1 and TIM-4 proteins bind to phosphatidylserine (PS) exposed on the surface of apoptopic cells (accompanying paper from Kobayashi et al.). We used X-ray crystallography to identify the PS binding site in the IgV domain of murine TIM proteins by co-crystallization of the isolated N-terminal domain of mTIM-4 with a PS compound having a short fatty acid moiety (C6, Experimental Procedures). Crystals (X3 form) were prepared with the mTIM-4 IgV domain and the structure of the complex was determined at 2.5 Å resolution (Experimental Procedures and Table 1).

The structure of the complex revealed that the PS ligand bound to the MILIBS (Fig. 3), with its hydrophilic moiety penetrating in the cavity built by the CC’ and FG loops. The acidic phosphate group of PS coordinated with the metal ion that was linked to the same mTIM-4 residues and with the same average coordination distance (2.5 Å) seen in the mTIM-4 structure with tartrate (Fig. 3B). The bipyramidal metal ion coordination was consistent with the presence of either a sodium or calcium ion according to the WaSP program (Nayal and Cera, 1996), although temperature B factor was lower for sodium (34) than for calcium (58). A sodium ion was included in the final pdb file because of the lower B factor and Rfree values given by the structure refinement, even though calcium was included in the preparation of the mTIM-4-PS complex at low concentration (Experimental Procedures). The structure of the metal ion coordination site is similar to that seen in a PS annexin-V complex, where the phosphate coordinates with a calcium ion bound mostly to main chain oxygens of a protein loop and the carboxylate of a glutamic acid residue (Swairjo et al., 1995)

Figure 3

Structure of the mTIM-4 IgV domain with a bound phosphatidylserine (PS) ligand

The Ser residue of PS fitted between the metal ion and the tip of the CC’ loop. Its amine group made specific interactions with the conserved Asp residue involved in metal ion coordination (Fig. 3B), while the carboxylate of the Ser was hydrogen bonded with a main chain amino and the hydroxyl group of a Ser residue in the CC’-loop of the mTIM-4 IgV domain. These interactions will be specific for the L stereoisomer of PS used in the crystallization. In the D isomer the amino and carboxylate groups of the Ser will interchange their positions, so that the carboxylate will face a repulsive electrostatic environment. The size of the phosphate-linked Ser and its interactions with the protein will be unique for PS and they must be responsible for the restricted phospholipid binding specificity of the TIM proteins (accompanying paper from Kobayashi et al.). The fatty acid moiety that anchors the phospholipid to the lipid bilayer interacts with the hydrophobic side chains of the Trp and Tyr residues, so that they must penetrate into the lipid bilayer upon TIM binding to PS on the surface of apoptotic cells (accompanying paper from Kobayashi et al.). Residues on the BC loop such as the Arg found in mTIM-1 and mTIM-4 will be close to the membrane (Fig. 3A), and they could interact with the lipid bilayer as well. The BC loop is largely variable among IgV domains of the TIM family (Fig. 2C) and bears mTIM-3 polymorphic residues (McIntire et al., 2001). This loop could regulate the PS binding affinity of TIM proteins. Residue variation in either BC, CC’ or FG loops could provide specificity to phospholipid binding by the TIM proteins. Even though TIM-3 conserves the residues involved in metal ion coordination it does not have any aromatic residue in the FG loop (Fig. 2C), which are critical for binding of mTIM-1 and mTIM-4 to PS (see below).

mTIM-1 and mTIM-4 soluble proteins bound to liposomes containing PS, while no binding was observed with mTIM-2 (Fig. 4A). The TIM proteins did not bind to phosphatidylcholine in liposomes (not shown). mTIM-1 and mTIM-4 shared high sequence similarity in the FG and CC’ loops building up the MILIBS and bound similarly to PS, which it is indicative of a nearly identical binding mode. Soluble molecules having the isolated IgV domain bound to PS liposomes, while no binding was detected with the mucin domain alone (Supplementary Fig 2 and accompanying paper from Kobayashi et al.). However it appears that the mucin domain facilitated binding of the bivalent mTIM soluble molecules to PS liposomes and could be important for presentation of the IgV domain on the cell surface.

Figure 4

Binding of murine TIM proteins to liposomes having PS and contribution of the MILIBS

The single (N/A) and double mutation (ND/AA) of the metal ion coordination residues decreased significantly (85%) and similarly binding of mTIM-1 and mTIM-4 proteins to PS in liposomes (Fig. 4B), suggesting that both residues shared metal ion coordination required for PS binding (Fig. 3B). However, single deletion of the aromatic residues in the FG loop (W/A or F/A) decreased about 70% binding of the TIM proteins to PS in liposomes, while the double mutation had an additive effect and abolished binding (Fig.4B). The dramatic effect of this double mutation indicated that the FG loop aromatic residues provided most of the binding energy. In good correlation with these results, the mutations inhibited also TIM-4 binding to PS on apoptotic cells and subsequent cellular phagocytosis of eryptotic red blood cells (see accompanying paper from Kobayashi et al.).

Binding of the mTIM proteins to PS liposomes was inhibited with the dicaproyl PS compound used for crystallization (Fig. 4C). The inhibition observed at low concentration of EDTA (Fig. 4C) or EGTA (not shown) indicated requirement of divalent calcium ions for high affinity binding. However, purified mTIM-1-Fc protein bound to PS liposomes without addition of divalent cations (not shown) and the structure of the mTIM4-PS complex showed that the most likely metal ion coordinated at the MILIBS was sodium (Fig. 2, ,3).3). Therefore, we can not exclude that the chelating agents blocked binding by penetrating into the binding pocket rather than by chelating divalent ions. The observed inhibitory effect of tartrate was consistent with the binding of this compound to the MILIBS (Fig. 2), while the moderate inhibitory effect observed at high concentrations of sodium chloride must be electrostatic (Fig. 4C).

These results proved a highly conserved PS recognition mode in the mTIM-1 and mTIM-4 proteins, consistent with the high similarity of the PS binding pocket revealed by the structure (Fig. 2C, ,3).3). Moreover, this binding mode must be shared by human TIM-1 and TIM-4, which have identical PS binding residues at the FG loop than the mouse proteins and bound to the phospholipid with similar affinities (see accompanying paper from Kobayashi et al.).

The MILIBS regulates trafficking of mTIM-1 to the cell surface

Recently we reported that most of the mTIM-1 protein (C57BL/6J strain variant) expressed in transfected lymphocytes was intracellular and that efficient trafficking of the receptor to the cell surface required stimulation with phorbol esters or ionomycin (Santiago et al., 2007). To determine the role of the MILIBS in the trafficking of mTIM-1 to the cell surface, we analyzed the cellular distribution of fluorescent tagged mTIM-1 protein mutants by confocal microscopy (Fig. 5A). We observed that receptor mutants lacking either the cation binding (N/A or ND/AA) or the aromatic residues (WF/AA) at the FG-loop were well expressed on the cell surface, such as we observed with the mTIM-2 protein (Santiago et al., 2007). The wild type mTIM-1 had a very distinct cellular distribution from the mutant proteins even though the expression of the proteins was similar (Fig. 5A). These results indicated that the MILIBS regulates trafficking of mTIM-1 to the cell surface.

Figure 5

Regulation of mTIM-1 intracellular trafficking by the MILIBS

We analyzed the cell surface presentation of two distinct IgV domain antibody epitopes (T1.4 and T1.10) mapped into the IgV domain of wild type and mutant mTIM-1 soluble proteins (not shown). The antibodies recognized cell surface expressed mutant proteins (Fig. 5B). However, we observed that the T1.10 epitope in the wild type protein was greatly diminished with respect to the T1.4 epitope and was detected in just 11% of the transfected cells. The lower mean fluorescence of the T1.4 antibody on the wild type compared to the mutant mTIM-1 transfected cells showed lower amount of wild type mTIM-1 molecules on the cell surface, as also shown by confocal microscopy (Fig. 5A). The differential display of IgV epitopes between the wild type and mutant mTIM-1 proteins indicates two distinct conformations for the membrane bound mTIM-1.

The increase in the concentration of intracellular calcium related to cell activation induces appearance of PS on the outer leaflet of the cell membrane (Bratton et al., 1997; Williamson et al., 1992; Zwaal and Schroit, 1997) and mediates also efficient trafficking of mTIM-1 to the cell surface (Santiago et al., 2007). This correlation in the cellular distribution of mTIM-1 and its physiological ligand PS as well as the cell surface localization of mTIM-1 mutants with a disrupted PS binding suggested that the mTIM-1 cellular distribution could be dependent on its PS binding activity. After translocation and folding in the ER lumen, type I membrane proteins are linked to the ER membrane by both the N-terminal signal peptide and the transmembrane region. The proximity of the MILIBS to the ER membrane prior to the proteolysis of the signal peptide could allow binding to PS, which would link the tip of the IgV domain to the lipid bilayer (Fig. 6). The structure of the PS bound to the IgV domain suggested that while the hydrophilic phospholipid head penetrated into the MILIBS, the aromatic residues on the tip of the FG loop would insert into the bilayer such as shown in Figure 6. The interaction of the loop with the membrane must be critical for high affinity binding of the TIM proteins to PS, such as suggested by the WF/AA mutant (Fig. 4B). So the N-terminal IgV domain of mTIM-1 could remain bound to the membrane after cleavage of the signal peptide in the ER (Fig. 6) and the protein would then be sequestered inside of the cells. It is known that increasing the intracellular calcium concentration induces transbilayer diffusion of membrane bound phospholipids (flip-flop) and the exposure of PS to the outer leaflet of the cell surface by the activation of the non-specific lipid scramblase (Verhoven et al., 1995; Williamson et al., 1992; Zwaal and Schroit, 1997). Therefore, an increase of the intracellular calcium concentration could enhance the exposure of PS-bound TIM to the outer leaflet of the membrane upon cell activation, with subsequent release of the receptor from the phospholipid on the cell surface by extracellular factors or by competition with TIM ligands (Fig. 6). This model implies the presence of two distinct conformations for the mTIM-1 molecule on the cell surface, supported by the distinct presentation of the T1.10 IgV domain epitope in the wild type and mutant proteins (Fig. 5B).

Figure 6

Model for TIM protein binding to PS and mTIM-1 trafficking to the cell surface

DISCUSSION

The structures of the mTIM-4 IgV domain identified a Metal Ion-dependent Ligand Binding Site (MILIBS) used for binding to phosphatidylserine by the TIM proteins. The ligand binding specificity is mediated by both the constrained size of the cavity and the requirement for an acidic group in the ligand for coordination with the metal ion. Although the most likely ion seen in the structures is sodium, other cations with similar coordination behaviour such as calcium could bind to this site depending on the ligand. The narrow cavity where the PS penetrates is built by two distinctive loops in the IgV domains of TIM proteins, so that the MILIBS is a specific signature of this receptor family. The lack of a ligand binding pocket in mTIM-2 due to its distinct CC’-loop and the absence of cation coordination residues in its FG loop further prove a divergent ligand recognition mode for this TIM family member (Santiago et al., 2007).

Comparison of the ligand-free mTIM-1 with the mTIM-4 structure in complex with PS showed certain requirements for residue rearrangement on the CC’FG loop epitope for ligand binding. In the open conformation of the MILIBS revealed by the mTIM-4 structure the hydrophobic Phe and Trp residues on the tip of the loop were primed for penetration into the lipid bilayer upon binding to PS on the cell surface. The hydrophobic tip of the FG loop in TIM-1 and TIM-4 resemble fusion loops described in class II virus membrane proteins (Harrison, 2005), which are known to penetrate into cellular membranes. The interaction of the hydrophobic residues with the cell membrane would strengthen the monomeric binding of the TIM proteins to PS, while the metal ion coordination and the conserved interactions of the Ser residue with the protein must provide the phospholipid binding specificity Even though the metal ion coordination residues and the conformation of the CC’ loop are conserved in all TIM but mTIM-2, there are certain differences in the FG and CC’ loops building up the binding pocket. These differences could have a significant influence on the binding affinity. Additional interaction of the IgV domain with the membrane could be mediated by residues at the tip of the BC loop, such as the Arg residues found in mTIM-1 and mTIM-4 or the residues that defined a polymorphism in the mTIM-3 BC loop (McIntire et al., 2001). The interaction model presented here for TIM binding to PS in cellular membranes is analogous to that proposed for binding of annexin-V and protein kinase C (Swairjo et al., 1995; Verdaguer et al., 1999). In those models hydrophobic residues near the PS binding site penetrated into the membrane and residues on neighbouring protein loops, such as a Lys in protein kinase C, defined a second interacting site with the bilayer.

Because the expression of TIM-4 is restricted to phagocytic cells, including macrophages and dendritic cells, the binding of PS to TIM-4 suggests that a major function of TIM-4 is the recognition and removal of apoptotic cells by such phagocytic cells (see accompanying manuscript by Kobayashi et al). Indeed, TIM-4 expressing cells rapidly and specifically phagocytized apoptotic cells, and this process was prevented by anti-TIM-4 blocking mAb and mutations in the MILIBS residues. On the other hand, since expression of TIM-1 is restricted to T cells and to ischemic kidney cells, the binding of PS to TIM-1 may mediate resolution of ischemic kidney damage by clearance of dying renal tubular epithelial cells and regulate TIM-1 intracellular trafficking in T cells.

The ability of mTIM-1 to bind phospholipids, its intracellular distribution and the absence of the T1.10 antibody epitope in the membrane bound wild type protein led us to propose two distinct conformations for the protein on the cell surface, that could be shared by TIM proteins binding to PS: One folded molecule with the tip of the IgV domain penetrating into the lipid bilayer and a second extended conformation where the IgV domain would be available for ligand binding on the cell surface. Moreover, TIM binding to PS on its own membrane could sequester the proteins inside of the cell, where PS resides, preventing their exposure on the cell surface for ligand binding and subsequent receptor-mediated intracellular signalling. According to this model, the efficient trafficking of the mTIM-1 protein to the cell surface observed upon cell activation could be related to either its intracellular release from PS or to the flopping of the PS bound TIM toward the outer leaflet of the cell membrane by changes in the lipid membrane polarity. It is well documented that the increase of the intracellular concentration of calcium during cell activation flops PS out of the cell membrane by the simultaneous activation of the enzyme scramblase and the inhibition of the aminophospholipid-specific translocase (Zwaal and Schroit, 1997). This regulatory mechanism for mTIM-1 trafficking must be specific for TIM proteins that bind PS, so that it might differ from other regulatory processes such as that described for CTLA-4 (Linsley et al., 1996).

According to the trafficking regulatory model presented here, the exposure of a functional mTIM-1 and perhaps other TIM proteins for ligand binding would be controlled and restricted to certain cellular conditions. Uncontrolled exposure of TIM molecules on the cell surface could come from changes in the IgV domain affecting its interaction with PS and the lipid bilayer, such as those seen here with the MILIBS mutants. Polymorphisms in mTIM-1 (McIntire et al., 2001), which modify the signal peptide and extend the mucin domain in the BALBc strain variant, could increase the distance between the tip of the domain and the lipid bilayer prior to the cleavage of the signal peptide, affecting binding of the IgV domain to PS and the amount of molecules on the cell surface. This uncontrolled exposure of TIM proteins could trigger undesired intracellular signals leading to lymphocyte activation upon binding to exogenous ligands.

We have shown here a unique Metal Ion-dependent Ligand Binding Site in the Ig domain of TIM proteins built up by distinctive structural motifs specifically found in the N-terminal domains of this family. We showed this is a phospatidylserine binding site and it could be also involved in binding to an identified mTIM-3 ligand and HAV, whose binding surface has been mapped on the GFC face of the IgV domain where the MILIBS locates (Cao et al., 2007; Santiago et al., 2007). Small molecules containing acidic groups could bind also to the MILIBS, while binding to multimeric ligands might crosslink TIM proteins on the cell surface and trigger intracellular signals. On the contrary, the critical contribution of the MILIBS to several TIM functions and its concave and well defined size opens up the rational design of small molecules targeting this site and modulating the functions of the TIM family in the immune system.

EXPERIMETAL PROCEDURES

Antibodies and cDNAs

mTIM-1 monoclonal antibodies (T1.4 and T1.10) and a mTIM-4 polyclonal antibody used were from eBioscience, Inc. The full-length cDNAs coding for mTIM-1 and mTIM-4 were from the mouse strain C57BL/6J strain and were used as templates for preparation of all the constructs.

Protein sample preparation

The mTIM-4 IgV domain was prepared by in vitro refolding of inclusion bodies produced in bacteria as described elsewhere (Santiago et al., 2007). The proteins had an N-terminal Met residue, residues 43 to 154 of the precursor protein, a thrombin recognition site and two epitope tags. The soluble proteins eluted from a Superdex-75 column (Amersham Biosciences) with a retention volume around 15–20 kDa. The recombinant proteins were thrombin treated to release C-terminal tags and further purified by ion-exchange chromatography using 25 mM Hepes buffer pH 7.5.

mTIM-4 crystallization and structure determination

Small crystals of the mTIM-4 domain diffracting at about 3Å were initially prepared (crystal form X1, Table 1) at a protein concentration of 15 mg/ml and crystallization conditions having 30% PEG-2000 methylether, 5% PEG-400, 0.1M ammonium sulphate, 0.1M sodium acetate pH 5.6 and 100 mM O-(n-Octyl)-phosphorylcholine. The X1 crystals belong to the hexagonal P321 space group, they have one molecule in the asymmetric unit and 55% solvent content. A second crystal form (X2) diffracting at 2.2Å resolution was prepared with purified protein at 10 mg/ml and 0.8 M potassium sodium tartrate having 0.1 M Hepes pH 7.5 (Table 1). The crystals have a molecule in the asymmetric unit, they belong to the hexagonal P6₃22 space group and have 63% of solvent content (Table 1). Complexes of the TIM-4 IgV domain and PS were prepared by incubation of the protein (15 mg/ml) with 5 mM 1,2-Dicaproyl-sn-Glycero-3-(Phospho-L-Serine) (Avanti) in 25 mM Hepes buffer pH 7.0 with 5 mM CaCl₂ and 100 mM NaCl. The complex was crystallized at a protein concentration of 15 mg/ml and conditions having 10% Jeffamine, 10 mM FeCl₃ and 0.1 M sodium citrate pH 5.6. The crystals (form X3) have a molecule in the asymmetric unit, 64% of solvent content and belong to the hexagonal P6₃22 space group (Table 1).

The structures of the mTIM-4 IgV domain were determined by the molecular replacement method using the program Phaser (Read, 2001) and the mTIM-2 structure lacking BC, CC’ and FG loops as search model. The X1 crystal form was refined using the program CNS (Brunger et al., 1998), while the structure of the X2 and X3 crystal forms were refined with Refmac (Collaborative Computational Project, 1994). Models were adjusted manually during the refinement process. The first two N-terminal and the last four C-terminal residues were poorly defined in the electron density maps. All residues in the structures are in allowed regions of Ramachandran plots. The ribbon representations were prepared with the program ribbons (Carson, 1987), while the remaining figures of the structure were done with PYMOL (http://www.pymol.org).

Protein expression in mammalian cells

DNAs coding for the indicated fragment of the proteins and having the hemagglutinin A epitope (TIM-HA) or the IgG1-Fc (TIM-Fc) at the C-terminus were prepared by transient expression in 293T cells and protein concentration determined by ELISA (Santiago et al., 2007). Recombinant constructs for expression of isolated mucin domains included the IgK leader sequence from the pDisplay vector (Invitrogen) at their 5’ end. Mutagenesis was done by the overlapping PCR technique and confirmed by DNA sequencing. All mutants bound to the specific TIM antibodies as the wild type soluble proteins in ELISA tests.

Fluorescent tagged proteins at their cytoplasmic tails were expressed on the surface of 293T or 300.19 pre-B cells by transfection with recombinant TIM cDNAs cloned in frame with a cyan or yellow fluorescent variants of GFP (CFP or YFP) in the pECFP-N1 or pEYFP-N1 vectors (Clontech), respectively. Expression was analyzed on a D-Eclipse C1 Leika microscope two days after transfection with a 435/10-nm and 500/20-nm excitation filters, 455-nm and 515-nm dichroic beam splitter and a 480/20-nm and 535/30 emission filter, respectively for CFP and YFP

Binding of TIM proteins to phospholipids in liposomes

Mixtures of phosphatidylserine:cholesterol (4:1) or phosphatidylcoline:cholesterol (4:1) (Sigma-Aldrich) were resuspended in chloroform:methanol (3:1,v/v). The solvent was removed using a rotary evaporator and finally dried under vacuum. PS and PC liposomes were resuspended in PBS to a final phospholipid concentration of 100 mM and sonicated to clarity. Binding of soluble Fc fusion proteins prepared in mammalian cells to plastic coated liposomes was carried out in duplicate wells of 96-well plates. Binding of liposomes to the plastic was done by adding 50 μl of the suspension at 0.01 mM to the wells and incubated overnight at 4ºC. Cell supernatants containing soluble Fc fusion proteins were diluted with DMEM (Invitrogen) having 5% FCS at the indicated protein concentration and binding assays were carried out essentially as performed for TIM-TIM protein binding (Santiago et al., 2007). Background binding signal to wells lacking liposomes was subtracted from optical density determined in wells with liposomes.

Acknowledgments

We acknowledge the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities through the MX-490 and MX-601 BAG projects. We are grateful to D.T. Umetsu and R. DeKruyff for helpful discussions. This work was supported by grants from the Ministerio de Educacion y Ciencia of Spain (BFU2005-05972) and CAM-CSIC (200520M028) to J.M.C. and NIH AI054456 to G.F and G.K.

Footnotes

Coordinates. The atomic coordinates and structure factors are being deposited in the Protein Data Bank.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

Bratton DL, Fadok VA, Richter DA, Kailey JM, Guthrie LA, Henson PM. Appearance of Phosphatidylserine on Apoptotic Cells Requires Calcium-mediated Nonspecific Flip-Flop and Is Enhanced by Loss of the Aminophospholipid Translocase. J Biol Chem. 1997;272:26159–26165. [PubMed]
Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges N, Pannu NS, et al. Crystallography and NMR system (CNS): A new software system for macromolecular structure determination. Acta Cryst D. 1998;54:905–921. [PubMed]
Cao E, Zang X, Ramagopal UA, Mukhopadhaya A, Fedorov A, Fedorov E, Zencheck WD, Lary JW, Cole JL, Deng H, et al. T Cell Immunoglobulin Mucin-3 Crystal Structure Reveals a Galectin-9-Independent Ligand-Binding Surface. Immunity. 2007;26:311–321. [PubMed]
Carson M. Ribbon models of macromolecules. J Mol Graph. 1987;5:103–106.
Collaborative Computational Project N. The CCP4 Suite: Programs for Protein Crystallography. Acta Cryst D. 1994;50:760–763. [PubMed]
Chakravarti S, Sabatos CA, Xiao S, Illes Z, Cha EK, Sobel RA, Zheng XX, Strom TB, Kuchroo VK. Tim-2 regulates T helper type 2 responses and autoimmunity. The Journal of Experimental Medicine. 2005;202:437–444. [PMC free article] [PubMed]
Chen TT, Li L, Chung DH, Allen CD, Torti SV, Torti FM, Cyster JG, Chen CY, Brodsky FM, Niemi EC, et al. TIM-2 is expressed on B cells and in liver and kidney and is a receptor for H-ferritin endocytosis. J Exp Med. 2005;202:955–965. [PMC free article] [PubMed]
Feigelstock D, Thompson P, Mattoo P, Zhang Y, Kaplan GG. The human homolog of HAVcr-1 codes for a hepatitis A virus cellular receptor. J Virol. 1998;72:6621–6628. [PMC free article] [PubMed]
Han WK, Bonventre JV. Biologic markers for the early detection of acute kidney injury. Curr Opin Crit Care. 2004;10:476–482. [PubMed]
Harding M. Metal-ligand geometry relevant to proteins and in proteins: sodium and potassium. Acta Crystallographica Section D. 2002;58:872–874.
Harrison SC. Mechanism of membrane fusion by viral envelope proteins. Adv Virus Res. 2005;64:231–261. [PubMed]
Kaplan G, Totsuka A, Thompson P, Akatsuka T, Moritsugu Y, Feinstone SM. Identification of a surface glycoprotein on african green monkey kidney cells as a receptor for hepatitis A virus. EMBO J. 1996;15:4282–4296. [PubMed]
Kuchroo VK, Meyers JH, Umetsu DT, DeKruyff RH. TIM family of genes in immunity and tolerance. Adv Immunol. 2006;91:227–249. [PubMed]
Kuchroo VK, Umetsu DT, DeKruyff RH, Freeman GJ. The TIM gene family: emerging roles in immunity and disease. Nat Rev Immunol. 2003;3:454–462. [PubMed]
Kumanogoh A, Marukawa S, Suzuki K, Takegahara N, Watanabe C, Ch'ng E, Ishida I, Fujimura H, Sakoda S, Yoshida K, Kikutani H. Class IV semaphorin Sema4A enhances T-cell activation and interacts with Tim-2. Nature. 2002;419:629–633. [PubMed]
Linsley PS, Bradshaw J, Greene J, Peach R, Bennett K, Mittler RS. Intracellular trafficking of CTLA-4 and focal localization towards sites of TCR engagement. Immunity. 1996;4:535–543. [PubMed]
McIntire JJ, Umetsu SE, Akbari O, Potter M, Kuchroo VK, Barsh GS, Freeman GJ, Umetsu DT, DeKruyff RH. Identification of Tapr (an airway hyperreactivity regulatory locus) and the linked Tim gene family. Nat Immunol. 2001;2:1109–1116. [PubMed]
McIntire JJ, Umetsu SE, Macaubas C, Hoyte EG, Cinnioglu C, Cavalli-Sforza LL, Barsh GS, Hallmayer JF, Underhill PA, Risch NJ, et al. Hepatitis A virus link to atopic disease. Nature. 2003;425:576–576. [PubMed]
Meyers JH, Chakravarti S, Schlesinger D, Illes Z, Waldner H, Umetsu SE, Kenny J, Zheng XX, Umetsu DT, DeKruyff RH, et al. TIM-4 is the ligand for TIM-1, and the TIM-1-TIM-4 interaction regulates T cell proliferation. Nat Immunol. 2005;6:455–464. [PubMed]
Nayal M, Cera ED. Valence Screening of Water in Protein Crystals Reveals Potential Na+Binding Sites. Journal of Molecular Biology. 1996;256:228–234. [PubMed]
Read RJ. Pushing the boundaries of molecular replacement with maximum likelihood. Acta Cryst D. 2001;57:1373–1382. [PubMed]
Rennert PD, Ichimura T, Sizing ID, Bailly V, Li Z, Rennard R, McCoon P, Pablo L, Miklasz S, Tarilonte L, Bonventre JV. T Cell, Ig Domain, Mucin Domain-2 Gene-Deficient Mice Reveal a Novel Mechanism for the Regulation of Th2 Immune Responses and Airway Inflammation. The Journal of Immunology. 2006;177:4311–4321. [PubMed]
Sabatos CA, Chakravarti S, Cha E, Schubart A, Sanchez-Fueyo A, Zheng XX, Coyle AJ, Strom TB, Freeman GJ, Kuchroo VK. Interaction of Tim-3 and Tim-3 ligand regulates T helper type 1 responses and induction of peripheral tolerance. Nat Immunol. 2003;4:1102–1110. [PubMed]
Sanchez-Fueyo A, Tian J, Picarella D, Domenig C, Zheng XX, Sabatos CA, Manlongat N, Bender O, Kamradt T, Kuchroo VK, et al. Tim-3 inhibits T helper type 1-mediated auto- and alloimmune responses and promotes immunological tolerance. Nat Immunol. 2003;4:1093–1101. [PubMed]
Santiago C, Ballesteros A, Tami C, Martinez-Munoz L, Kaplan GG, Casasnovas JM. Structures of T Cell Immunoglobulin Mucin Receptors 1 and 2 Reveal Mechanisms for Regulation of Immune Responses by the TIM Receptor Family. Immunity. 2007;26:299–310. [PubMed]
Silberstein E, Xing L, van de Beek W, Lu J, Cheng H, Kaplan GG. Alteration of hepatitis A virus (HAV) particles by a soluble form of HAV cellular receptor 1 containing the immunoglobin-and mucin-like regions. J Virol. 2003;77:8765–8774. [PMC free article] [PubMed]
Swairjo MA, Concha NO, Kaetzel MA, Dedman JR, Seaton BA. Ca2+-bridging mechanism and phospholipid head group recognition in the membrane-binding protein annexin V. Nat Struct Mol Biol. 1995;2:968–974.
Umetsu SE, Lee WL, McIntire JJ, Downey L, Sanjanwala B, Akbari O, Berry GJ, Nagumo H, Freeman GJ, Umetsu DT, DeKruyff RH. TIM-1 induces T cell activation and inhibits the development of peripheral tolerance. Nat Immunol. 2005;6:447–454. [PubMed]
Verdaguer N, Corbalan-Garcia S, Ochoa WF, Fita I, Gómez-Fernández JC. Ca2+ bridges the C2 membrane-binding domain of protein kinase C directly to phosphatidylserine. The EMBO Journal. 1999;18:6329–6338. [PubMed]
Verhoven B, Schlegel RA, Williamson P. Mechanisms of phosphatidylserine exposure, a phagocyte recognition signal, on apoptotic T lymphocytes. J Exp Med. 1995;182:1597–1601. [PMC free article] [PubMed]
Vila MR, Kaplan GG, Feigelstock D, Nadal M, Morote J, Porta R, Bellmunt J, Meseguer A. Hepatitis A virus receptor blocks cell differentiation and is overexpressed in clear cell renal cell carcinoma. Kidney International. 2004;65:1761–1773. [PubMed]
Wilker PR, Sedy JR, Grigura V, Murphy TL, Murphy KM. Evidence for carbohydrate recognition and homotypic and heterotypic binding by the TIM family. Int Immunol. 2007;19:763–773. [PubMed]
Williamson P, Kulick A, Zachowski A, Schlegel RA, Devaux PF. Ca2+ Induces Transbilayer Redistribution of All Major Phospholipids in Human Erythrocytes. Biochemistry. 1992;31:6355–6360. [PubMed]
Zhu C, Anderson AC, Schubart A, Xiong H, Imitola J, Khoury SJ, Zheng XX, Strom TB, Kuchroo VK. The Tim-3 ligand galectin-9 negatively regulates T helper type 1 immunity. Nat Immunol. 2005;6:1245–1252. [PubMed]
Zwaal RFA, Schroit AJ. Pathophysiologic Implications of Membrane Phospholipid Asymmetry in Blood Cells. Blood. 1997;89:1121–1132. [PubMed]