The Structure of the Extracellular Domain of Triggering Receptor Expressed on Myeloid Cells Like Transcript-1 and Evidence for a Naturally Occurring Soluble Fragment*

Triggering receptor expressed on myeloid cells like transcript-1 (TLT-1) is an abundant platelet-specific, type I transmembrane receptor. The extracellular fragment of TLT-1 consists of a single, immunoglobulin-like domain connected to the platelet cell membrane by a linker region called the stalk. Here we present evidence that a soluble fragment of the TLT-1 extracellular domain is found in serum of humans and mice and that an isoform of similar mass is released from platelets following activation with thrombin. We also report the crystal structure of the immunoglobulin domain of TLT-1 determined at the resolution of 1.19 Å. The structure of TLT-1 is similar to other immunoglobulin-like variable domains, particularly those of triggering receptor expressed on myeloid cells-1 (TREM-1), the natural killer cell-activating receptor NKp44, and the polymeric immunoglobulin receptor. Particularly interesting is a 17-amino acid segment of TLT-1, homologous to a fragment of murine TREM-1, which, in turn, showed activity in blocking the TREM-1-mediated inflammatory responses in mice. Structural similarity to TREM-1 and polymeric immunoglobulin receptor, and evidence for a naturally occurring soluble fragment of the TLT-1 extracellular domain, suggest that this immunoglobulin-like domain autonomously plays an as yet unidentified, functional role.

Triggering receptor expressed on myeloid cells like transcript-1 (TLT-1) 2 is a membrane-bound protein, abundant in the ␣-granules of resting platelets and on the surface of activated platelets (1). The gene that codes for TLT-1 is located in the TREM gene cluster on human chromosome 6, locus 6p21.1, and murine chromosome 17. Related genes in the human TREM gene cluster include TREMs 1 and 2, TLTs 1 through 5, and the natural killer cell-activating receptor NKp44 (2). In the murine genome, chromosome 17 includes genes encoding TREMs 1 through 5, and TLTs 1 and 2 (2). Gene products from the TREM gene cluster are expressed in specific populations of cells of the myeloid lineage. To date, only TREM-1, TREM-2, NKp44, and TLT-1 have been characterized.
The immunoglobulin-like domain of human TLT-1 (hTLT-1) consists of 105 residues and is attached to the membrane by a 37-amino acid stalk. The putative transmembrane segment of hTLT-1 is 20 amino acids long and lacks the positively charged amino acid that other TREM family members use to interact with a negatively charged side chain of the adaptor molecule DAP12 (3). Inside the membrane-spanning region, DNAX activation protein 12 contains a signaling motif known as an immunoreceptor tyrosine-based activation motif (3). Two cDNA sequences, which code for proteins with identical extracellular and membrane-spanning regions but different cytoplasmic domains (2), have been reported for hTLT-1. The shorter transcript has a cytoplasmic domain of 18 amino acids with no recognizable signaling motifs, and the cytoplasmic domain of the longer hTLT-1 transcript consists of 127 amino acids, including two tyrosine residues. Both cytoplasmic tyrosines are found in sequence motifs known as immunoreceptor tyrosine-based inhibitory motifs (4). TLT-1, carrying two immunoreceptor tyrosine-based inhibitory motifs, is the only product of the TREM gene cluster with known signaling motifs.
Studies using a murine model of septic shock have demonstrated an interesting paradigm for TREM-1 function: Expression of surface-bound TREM-1 is up-regulated in response to bacterial infection, resulting in increased production of pro-inflammatory cytokines and leading to an immune response against the potential pathogen. However, overreaction of the immune system can cause septic shock. A recent study showed that administration of either the intact immunoglobulin-like domain of murine TREM-1 or a specific 17-amino acid peptide fragment of mTREM-1 resulted in decreased production of pro-inflammatory cytokines and protected mice from death caused by septic shock (12). Thus, the mTREM-1 immunoglobulin-like domain apparently blocks the signal produced by surface-bound mTREM-1 (12).
While examining the expression patterns of murine TLT-1 (mTLT-1) during platelet activation, we observed the time-dependent production of a smaller form of mTLT-1. Here, we present evidence that soluble species of TLT-1 can be found in murine and human serum. To build a strong foundation for future characterization of TLT-1 function, we crystallized a recombinant fragment of the hTLT-1 immunoglobulinlike domain consisting of residues 20 -125, and determined its crystal structure at the resolution of 1.19 Å.

Cell Culture and Transfection
Human embryonic kidney (HEK) 293T cells were maintained and transfected as described earlier (13). Expression plasmids for mTLT-1 and characterization of full-length murine TLT-1 (mTLT-1) were previously described (13). The expression plasmid for mTLT-1 was tagged at the carboxyl terminus with a yellow fluorescent protein (YFP) fragment using PCR. Primers used for amplification included: forward, gtttaaacatgtgatggtgagcaagggcgaggag, and reverse, cggcacatgtggcagtcgtccatgccgagagtg. The PCR product was digested with PmeI and PvuII and blunt-end-ligated into an EcoRV site in the vector tag site to create mTLT-1YFP. hTLT-1 was generated by RT-PCR using Platinum HiFi-Supermix (Invitrogen) with human platelet cDNA as a template. Primers included: forward, atgggcctcaccctgctcttg, and reverse, gctggatggagtctgattg.

Preparation of Washed Platelets
Murine Platelets-Mice were bred and maintained under specific pathogen-free conditions at NCI, National Institutes of Health (NIH), Frederick, MD. Animal care was provided following the procedures outlined in A Guide for the Care and Use of Laboratory Animals (14). Peripheral blood collection via cardiac puncture and platelet purification from peripheral blood were performed as described earlier (15). Platelets were then washed in modified mouse Tyrode's buffer (10 mM HEPES, 138 mM NaCl, 5.5 mM glucose, 2.9 mM KCl, and 12 mM NaHCO 3 , pH 7.4).

Antibodies
Antimurine TLT-1 was generated by immunizing rabbits with a fusion protein containing residues 21-167 of mTLT-1 and a polyhistidine tag for purification (Fusion Antibodies, Belfast, Northern Ireland). An additional antimurine TLT-1 polyclonal antibody, termed mTLT1-c, as well as antihuman TLT-1, were generated by immunizing goats, and purified by affinity chromatography (R&D Systems, Minneapolis). Horseradish peroxidase-conjugated antibodies and controls were obtained from BD Pharmingen (San Diego, CA).

Synthetic Gene Construction
The cDNA sequence encoding full-length hTLT-1 was obtained from GenBank TM (BC100945) (16). The expression construct used for structure determination contains only the immunoglobulin-like domain (residues 20 -125). This fragment (20 -125) was chosen for structural studies based on alignment of the sequence of TLT-1 with the sequence used for structure determination of NKp44 (1HKF), and predicted secondary structure of hTLT-1 from the Jufo server (5,17). The DNA sequence was optimized for expression in Escherichia coli using the program DNAWorks (18). The gene assembly and amplification were performed as previously described (18), and then the synthetic gene was cloned into a pET-22b plasmid vector for expression in E. coli.

Expression and Purification of Recombinant Immunoglobulin-like Domain of hTLT-1
A pET-22b vector (Novagen) containing the gene encoding residues 20 -125 of hTLT-1 was transformed into the E. coli BL21DE3-RIL-competent cells (Invitrogen). Transformed cells were grown on Luria-Bertani (LB):agar plates supplemented with 34 g/ml chloramphenicol and 50 g/ml of ampicillin. Individual colonies were isolated and used to inoculate 50-ml cultures of LB with chloramphenicol and ampicillin, and incubated for 12-16 h at 37°C. The overnight cultures were resuspended in 950 ml of LB (with chloramphenicol and ampicillin) and incubated at 37°C until optical densities showed logarithmic growth. Expression was induced by addition of 1 mM isopropyl 1-thio-␤-Dgalactopyranoside and 1% glucose (m/v). After incubation with isopropyl 1-thio-␤-D-galactopyranoside for 3-5 h, cells were harvested by centrifugation at 5,000 ϫ g for 20 min. The cell pellet was resuspended in 50 mM Tris, pH 8, with lysozyme (1 mg/ml), deoxyribonuclease I (1 mg/ml), and ribonuclease A (1 mg/ml). After mechanical cell lysis and centrifugation (28,000 ϫ g for 40 min), the soluble lysate was discarded. The insoluble material containing TLT-1 was washed several times, and solubilized in 8 M urea as previously described (19). After centrifugation and filtering, the resulting solution was diluted with 50 mM Tris (pH 8) containing arginine (400 mM), reduced glutathione (5 mM), and oxidized glutathione (0.5 mM) to a final protein concentration between 0.1 and 0.01 mg/ml, and final urea concentration of 50 -100 mM. After stirring for 4 days at 4°C, the refolded and oxidized solution of hTLT-1 was concentrated (Amicon ultrafiltration cell, Millipore) and subjected to size-exclusion and ion-exchange chromatographies (Sephacryl S100 HR and DEAE-Sepharose, GE Biosciences). Purity and identity of the protein were established by SDS-PAGE, and by electrospray mass spectrometry, ESI-MS (Agilent Technologies, Hewlett-Packard).
The purified hTLT-1 sample was digested with trypsin and analyzed by ESI-MS to establish the connectivity of disulfide bonds. Sequencing grade-modified trypsin (Roche Diagnostics Gmbh, Mannheim, Germany) was dissolved in 1 mM HCl to a concentration of 1 mg/ml and mixed in a mass-ratio of 1:50 with a 1 mg/ml solution of hTLT-1 in 0.1 M Tris, pH 8.5. After overnight digestion at room temperature, samples were analyzed by ESI-MS.

Crystallization
Preparations of hTLT-1 immunoglobulin-like domain were subjected to crystallization trials by hanging drop vapor diffusion. Clusters of thin crystal plates initially appeared in the droplets containing 5-10 mg/ml hTLT-1, 0.8 M sodium/potassium phosphate (pH 5.6). Subsequent optimization resulted in thicker crystals. The final crystallization medium contained 1 l each of 10 mg/ml hTLT-1 in 20 mM Tris, pH 8, and mother liquor consisting of 15-19% (w/v) polyethylene glycol 6000, and 0.5 M sodium/potassium phosphate (pH 5.6). All crystallization experiments were completed at 15°C.

Data Collection and Processing
Diffraction data were collected at 100 K (Oxford Cryosystems, Oxford, UK) after brief soaking of crystals in a mother liquor, supple-

Structure of the TLT-1 Extracellular Domain
A complete set of diffraction intensities extending to the resolution of 1.7 Å was collected using X-radiation generated by rotating anode ( ϭ 1.542 Å) mounted on the Rigaku Ru200 x-ray generator (Rigaku/MSC) operated at 50 kV and 100 mA. Intensities were recorded using a Mar345 image plate detector (Mar Research). The high resolution data set (30 -1.19 Å) was collected at the SER-CAT beamline at the Advanced Photon Source (Argonne, USA) with x-ray wavelength tuned to 1.135 Å. The diffraction images were recorded using the 2 ϫ 2 mosaic chargecoupled device detector (Mar Research). All data were processed and scaled using HKL2000 (HKL Research) (20). Table 1 shows detailed data collection statistics.

Structure Determination and Refinement
The structure of the immunoglobulin-like domain of hTLT-1 was solved by molecular replacement, using the program Phaser (21), and the experimental data were collected on a conventional x-ray source. The assembly of superimposed models of hTREM-1 (PDB: 1SMO chain A) (6), NKp44 (PDB: 1HKF) (5), and the D1 domain of pIgR (PDB: 1XED chain E) (22) was used as a search probe. All variable loop regions were removed from the search model. After successful identification of the solution, hTREM-1 atoms from the search assembly were used as the initial model of the hTLT-1 structure. Through an iterative process of manual and automated structural refinement, the amino acid sequence was converted to that of hTLT-1 using the programs O (23) and CNS (24). This process was repeated until all residues were incorporated into the model. The resolution of x-ray data was gradually extended to 1.6 Å. At this stage, all well defined water molecules were included in the model, and the resolution of the experimental data was extended to the available limit of 1.19 Å. Subsequent structural refinement was carried out with the program Refmac (25). Additional water molecules were identified automatically with the program ARP/wARP (26). During the final stages of refinement, isotropic B factors were replaced with anisotropic terms. Values of crystallographic R factor and free R for the final model were 0.165 and 0.195, respectively. The model's geometric and stereochemical properties were monitored by the programs Procheck (27) and WhatCheck (28). Table 1 shows detailed refinement statistics. The final coordinates and the experimental structure factors have been deposited with the protein data bank (PDB code: 2FRG).

RESULTS
Soluble Fragment of TLT-1-TLT-1, found in the ␣-granules of resting platelets, is translocated to the platelet surface following activation by thrombin (1). A time-course experiment was performed using resting and activated mouse platelets to characterize activation-dependent changes in mTLT-1 expression. Following activation, platelet suspensions were combined with an equal volume of 2ϫ lysis buffer. The equivalent of 1.5 ϫ 10 7 platelets was resolved by polyacrylamide gel electrophoresis. During the 5-h period, the apparent mass of mTLT-1 shifted from almost exclusively a 40-kDa form to a nearly equal mixture of 40-and 25-kDa forms (Fig. 1A). This observation led us to test whether the smaller mTLT-1 isoform was released from platelets. In subsequent experiments, platelets were centrifuged prior to lysis to separate the cell-associated and -soluble fractions. Pelleted resting platelets and the resulting supernatant showed no trace of a 25-kDa isoform (Fig.  1B, lanes 1 and 2). However, following activation with thrombin, a soluble form of mTLT-1 could be detected in the supernatant after removal of platelets by centrifugation (Fig. 1B, lane 4). The 25-kDa form of mTLT-1 was detected in murine serum when probed with an antibody specific for the extracellular domain of mTLT-1 but not detected in serum using an antibody specific to the cytoplasmic domain of mTLT-1, suggesting that the species in serum includes the extracellular domain (Fig. 1C). We confirmed the identity of the 25-kDa band as TLT-1 by showing that two different antibodies directed toward the extracellular domain of mTLT-1 reacted with it, whereas a series of control antibodies did not (Fig. 1D). TLT-1 was not observed in murine plasma or in defibrinated plasma (Fig. 1D, lanes 2 and 3). To test whether mTLT-1 found in serum resulted from proteolytic cleavage of the platelet surface-associated variant, or from alternative transcription of soluble isoforms, we transfected HEK293 cells with cDNA encoding mTLT-1, or a mTLT1/YFP chimera (Fig. 1E). Similar results were obtained for both forms of recombinant mTLT-1 in that the protein is observed near the expected molecular masses in cell-associated fractions (lanes 1 and 3), and isoforms of identical size were detected in conditioned cell media, irrespective of the cDNA construct used for transfection (lanes 2 and 4). The isoform released from HEK293 cells was identical in apparent molecular mass to the form detected in murine serum (lane 5).
Although there is Ͼ70% homology between the murine and human orthologs, we repeated the experiments using human platelets, serum, and HEK293 cells transfected with hTLT-1 cDNA. Human platelets and serum probed with an antibody specific for the extracellular domain of hTLT-1 demonstrate similarities to the murine system (Fig. 2, A and B). Activated human platelets show various hTLT-1 isoforms, but human serum contains only a pair of smaller forms ( Fig. 2A). In human serum, two bands corresponding to apparent masses of 12 and 14 kDa were observed ( Fig. 2A), very similar to the doublet observed in supernatant following platelet activation (Fig. 2B). Human TLT-1 is absent from supernatants of resting platelets and from plasma (Fig. 2B). To test whether hTLT-1 fragments were in fact derived from cleavage of hTLT-1, or from alternative transcription, we transfected HEK293 cells with the hTLT-1 cDNA encoding the smaller of two known forms of hTLT-1 and evaluated the cell culture media for the presence of these fragments. No hTLT-1 was detected from wild-type HEK293 cells or conditioned culture media, but full-length (35 kDa) and smaller forms (12 and 14 kDa) are clearly observed in extracts from cells, which were transfected with the hTLT-1 gene (Fig. 2C).  3C). The overall structure of the immunoglobulin-like domain of hTLT-1 is very similar to antibody variable domains with the largest differences observed in CDR loop regions (Fig. 3D). hTLT-1 has two disulfide bonds, homologous to those seen previously in the structures of NKp44 and pIgR. Formation of the disulfide bonds was closely monitored during the purification of hTLT-1 by trypsin digestion and subsequent ESI-MS. One of the disulfide bonds (Cys 38 -Cys 104 ) in TLT-1 is relatively conserved in immunoglobulin-like domains, and it is found in murine and human TREM-1, NKp44, and pIgR. The second disulfide bond in hTLT-1, connecting Cys 52 located in strand ␤C and Cys 59 (strand ␤CЈ), is not present in murine or human TREM-1 but is found in NKp44 and pIgR.
Comparison with Known Structures-The model of hTLT-1 immunoglobulin-like domain was used as a probe for searching the protein data bank (PDB) using the Vector Search Alignment Tool from NCBI (29). The search identified 53 non-redundant protein structures with significant structural similarity to hTLT-1. The closest structural relatives to hTLT-1 are cell-surface receptors, including pIgR, TREM-1, NKp44, and the mouse myeloid cell receptor, Clm-1 (PDB code: 1ZOX). The immunoglobulin-like domain of hTLT-1 also shares significant structural homology with T-cell receptors, cell adhesion molecules, and other immune receptors. The root-mean-square deviation (r.m.s.d.) between the structurally equivalent C ␣ atoms in hTLT-1 and a T-cell receptor, V␣11 (PDB code: 1H5B chain A) is 1.7 Å, whereas the value calculated for the human junctional adhesion molecule type I (PDB  CDR-equivalent Loops-Complementarity determining regions (CDRs) have been identified in antibody V-domains as portions as antigen binding regions (30). In single-domain immunoglobulin-like molecules, the CDRequivalent loops are referred to as CDR1, -2, and -3. The structure of hTLT-1 demonstrates variability in these loop structures, and these are the locations of the largest structural differences between hTLT-1 and related immunoglobulin-like crystal structures.
The conformations of CDR-equivalent loops in hTLT-1 were clearly defined in electron density, due in part to crystal contacts with neighboring molecules, which stabilized the positions of CDR2-and CDR3equivalent loops. A section of the CDR1-equivalent loop forms a 3 10 helix (31) (Fig. 4A). In hTLT-1, the side chains of Tyr 40 and Lys 49 interact with the carboxylate oxygen atoms of D81, located on the D-E loop, probably stabilizing the orientations of strands ␤B and ␤C relative to strands ␤D and ␤E. Human TREM-1 and NKp44 are homologues at the positions equivalent to Tyr 40 , Lys 49 , and Asp 81 in hTLT-1. In pIgR the latter residue is asparagine, resulting in a similar pattern of interactions. In mTREM-1 the residues equivalent to Tyr 40 and Asp 81 (in hTLT-1) are phenylalanine (Phe 43 ) and histidine (His 89 ). Their relative orientations do not support a similar favorable interaction; however, the side chain of Lys 53 is within hydrogen bond distance from a nitrogen atom in the side chain of His 89 (Fig. 4).
The second hypervariable region in the TREM family includes CDR2equivalent residues, linking strands ␤CЈ and ␤CЉ (Fig. 4B). In murine and human TREM-1, NKp44, and pIgR, this hypervariable region extends from the end of ␤CЈ, through strand ␤CЉ to Arg 76 , located in the CЉ-D loop. In TLT-1, conserved residues surrounding CDR2 include Ser 65 (serine or threonine in related molecules) and Arg 76 . The hydrogen bond between a side-chain equivalent to Ser 65 and a backbone atom of strand ␤C is maintained in all known structures of TREM family members. Arg 76 , in turn, forms a hydrogen bond and a salt bridge with the side chains of Gln 95 and Asp 98 from the nearby loop D-E. These interactions likely stabilize the overall structure of the ␤-sheets while allowing sequence diversity within the hypervariable loops.
In hTLT-1, the third CDR-equivalent loop is very short and consists of only 4 residues that do not show hydrogen bonds characteristic of ␤-strands (Fig. 4C). This loop shows the greatest conformational variability between TREM-1, NKp44, and TLT-1 (the only TREM family structures currently known). In fact, significant difference is observed between the two hTREM-1 crystal structures. The first structure (1Q8M) is a domain-swapped dimer in which the loop CDR3 (F-G loop), the amino-terminal half of ␤G, and the amino terminus were in extended conformations, forming a large portion of the dimeric interface. In murine and human TREM-1, the F-G loop is three residues longer than hTLT-1. In mTREM-1 and monomeric hTREM-1 struc-  tures, this loop bends toward the groove formed by twisting of the GFCCЈ ␤-sheet. The flexibility observed near CDR3 in the domainswapped dimer of hTREM-1 results from a conformational change involved in dimerization and hints at structural changes occurring in these related molecules upon ligand binding.
Electrostatic Surfaces-Although the majority of backbone atoms of hTLT-1 and related molecules superimpose well, variations in loop conformations and the diversity of amino acid sequences result in significant differences between the molecular surfaces of these molecules (Fig.  5). As expected, the greatest differences are at the loop regions, particularly CDR-equivalent loops and the amino termini. However, even in structurally well conserved regions, the electrostatic potentials projected on molecular surfaces are significantly different, which may contribute to a ligand binding specificity.
The molecular electrostatic surface of hTLT-1 shows several charged patches separated by regions with low electrostatic surface potential (Fig. 5). The tips of CDR loops 1, 2, and 3, in hTLT-1, contribute positively charged residues, thus determining a predominantly positive charge near these areas. CDR1 contains Arg 41 , and Lys 46 extending away from the molecule center. CDR2 contains Asp 68 near the tip of the loop, forming an isolated negatively charged patch on the surface, but is followed in the sequence by Arg 69 and Arg 70 , which form a larger, more exposed, positively charged surface patch.
The electrostatic surface of NKp44 (Fig. 5) displays a distinctly positively charged groove created by the twist in the ␤-sheet containing strands CCЈFG and the loops connecting these strands. In hTLT-1 this groove is less charged and includes several negatively charged patches.
Although the deepest portions of this groove on the surface of TLT-1 are uncharged, the sides of the groove are lined with negative charges contributed by Asp 107 near CDR3 on ␤-strand F, and Glu 58 with Glu 60 located in the C-CЈ loop and the strand ␤CЈ. On the surface of monomeric hTREM-1, this groove is essentially covered by side chains of the C-CЈ and F-G loops, which are in closer proximity in TREM-1 compared with hTLT-1. Prominent on the surface of hTREM-1 (PDB code 1SMO), is a 5.5-Å approach of the carboxylate oxygen from Glu 121 and the methyl group of Met 63 . In hTLT-1, the F-G loop bends away from the C-CЈ loop and contacts a neighboring molecule in the crystal. A more exposed groove on the surfaces of hTLT-1 and NKp44 relative to murine or human TREM-1 results from the differences in loop conformations.
Comparison with Active mTREM-1 Peptide-Residues 94 -110 in hTLT-1 are structurally equivalent to mTREM-1 residues 103-119. A peptide composed of mTREM-1 residues 103-119 has been synthesized and shown to block pro-inflammatory cytokine production caused by intact, membrane-bound mTREM-1 in a mouse model of septic shock (12). In hTLT-1 and mTREM-1, this peptide encompasses roughly half of the E-F loop, the entire ␤-strand F, and part of the F-G loop (CDR3). Both ends of the active peptide from mTREM-1 form solvent-exposed loops, and much of the ␤-strand portion of the peptide contributes to the groove formed by twisting the GFCCЈ ␤-sheet and protrusion of the C-CЈ loop (Fig. 5). In hTLT-1, residues equivalent to the mTREM-1 peptide contribute negative charges to the electrostatic surface at the E-F loop, a property conserved in all known structures of TREM family members (NKp44, hTREM-1, mTREM-1, and hTLT-1). At the carboxyl end of the peptide, Arg 110 of hTLT-1 contributes a positive charge to the molecular surface at CDR3, whereas murine and human TREM-1 have proline in the equivalent location.
This 17-amino acid peptide is highly conserved between murine and human TREM-1, with only 3 amino acid substitutions in this region. The sequence of hTLT-1 shows additional variation. Only 6 of 17 amino acids are conserved, all near the amino terminus of the peptide, i.e. farthest from CDR3 (Fig. 6). The backbone atoms superimpose well between mTREM-1 and hTLT-1 over residues in the N-terminal half of the peptide corresponding to the E-F loop and most of strand F, but toward the C-terminal end of the peptide, the sequences and backbone traces diverge. The divergence between the two structures reflects the sequence diversity and conformational differences of CDR3. Thus, the 17-amino acid peptide comprises nearly equal portions of a conserved structural element, and one that is unique to each different protein.

DISCUSSION
Platelet activation is one of the first steps in a cascade of thrombolytic events, including clot formation and retraction (32). Here we show that the extracellular domain of murine and human TLT-1 is released from the surface of activated platelets. From murine platelets and serum we observed a single band, or tight doublet with an apparent molecular mass of 25 kDa, whereas the full-length mTLT-1 could be observed at an apparent mass of ϳ45 kDa. Soluble hTLT-1 isoforms appeared at apparent masses of 12 and 14 kDa, whereas the heaviest hTLT-1 isoform observed ran at an apparent mass of ϳ35 kDa. In both human and murine TLT-1, the mass difference between full-length and truncated forms is roughly 20 kDa (judged by SDS-PAGE). The molecular masses of mTLT-1 and hTLT-1, calculated from their amino acid sequences (omitting residues 1-20 as signal peptides) are 31.5 and 30.7 kDa, respectively. The difference between observed and calculated masses, therefore, likely results from different post-translational modification in mice relative to humans. FIGURE 5. The molecular surfaces of hTLT-1, hTREM-1, NKp44, and an antibody light chain variable domain colored by an electrostatic potential over a range of values from ؊5 kJ/mol/Å 2 (red) to 5 kJ/mol/Å 2 (blue). All models are superimposed relative to TLT-1 and displayed in the same orientation as in Fig. 3 (B-D), as well as Fig. 4. Structures of TREM family members have unique surface shape and electrostatic potential. An antibody light-chain variable domain represents an unrelated immunoglobulin variable domain that shares the overall structural fold but lacks the level of sequence conservation observed between the TREM gene products. The surfaces and electrostatic potentials were calculated with the program APBS (33) and displayed with Pymol (www.pymol.org).
The soluble species of mTLT-1 released from platelets and detected in serum likely contain modifications, as an unmodified 25-kDa form of mTLT-1 would retain its membrane-spanning region and most of the cytoplasmic domain. Intact hTLT-1, however, is observed at an apparent mass much closer to its predicted molecular mass than mTLT-1, and the soluble forms of hTLT-1 observed are significantly smaller. Apparent fragments of the hTLT-1 extracellular domain of masses 12 and 14 kDa correspond roughly to 100 and 110 amino acids of the extracellular domain, or approximately the complete immunoglobulin-like domain (105 residues total, 20 -125). Although the smaller forms of TLT-1 may result from alternative transcription, release of recombinant murine or human TLT-1 extracellular domain from HEK293 cells strongly suggests that the stalk region of TLT-1 is susceptible to hydrolysis. Immunoblots were probed with two different antibodies specific for the extracellular domain of mTLT-1, and a panel of control antibodies suggests that the observed forms are actually mTLT-1 (Fig. 1D), and not the result of nonspecific binding interactions observed with a single antibody. Our observation that TLT-1 can be detected in soluble forms released from platelets suggests the potential for biological activity of these apparent fragments. In the context of platelet function, surfaceassociated TLT-1 could enhance platelet aggregation or clot formation, and shedding of TLT-1 could participate in clot retraction by helping release platelets from the clot. Soluble TLT-1 species may bind the natural TLT-1 ligand and prevent productive engagement of surfaceassociated TLT-1. The crystal structure of this recombinant fragment of hTLT-1 (residues 20 -125) includes the most likely ligand binding portion of the extracellular domain and is similar to the soluble isoforms identified from natural sources.
The crystal structure of the immunoglobulin-like domain of hTLT-1 demonstrates a close relationship with other TREM receptors (5-8) but also with pIgR (22) and the mouse myeloid cell receptor Clm-1 (1ZOX). Structural similarities between hTLT-1 and antibody variable domains, T-cell receptors, and cell adhesion molecules suggest that TLT-1 can form a complex with one or more protein or peptide ligands. Comparing ␤-sheet residues of hTLT-1 with murine and human TREM-1, and NKp44, indicates a closer relationship with TREM-1 than NKp44, because the r.m.s.d. is nearly twice as high between hTLT-1 and NKp44, compared with hTLT-1 and either murine or human TREM-1. This was unexpected, because the structure confirmed that hTLT-1 and NKp44 share a second disulfide bond not common to hTLT-1 and TREM-1.
Antibodies use CDR loops to tightly bind protein or peptide antigens (30). CDR loops show the greatest variability in length and conformation with ␤-strands conserving the immunoglobulin V-type fold (Figs. 3 and 4), partly due to interactions between ␤-strands and ␤-sheets, which hold adjacent parts of the structure in a conserved relative orientation. Even within the conserved portions of the immunoglobulin-like domain, electrostatic surfaces demonstrate that hTLT-1 is unique, with sequence variability resulting in differences in surface shape and electrostatic properties (Fig. 5).
Recently, a recombinant mTREM-1 immunoglobulin-like domain was shown to block mTREM-1-mediated pro-inflammatory cytokine production (12). In addition to the intact immunoglobulin-like domain, a 17-amino acid fragment of this domain reduced pro-inflammatory cytokine production and death in a murine model of septic shock (11). The activity of this peptide in apparently blocking mTREM-1 activity is interesting, because the N-terminal portion of this peptide is highly conserved in sequence and structure within the TREM family, and the carboxyl portion of the peptide is unique between members of the TREM family. The location of this peptide-forming part of CDR3, its prominent exposure on the molecular surface, and its anti-mTREM-1 activity suggest that the peptide includes an important binding site for the natural ligand of TREM-1. The peptide may block productive binding between the ligand and surface-associated TREM-1 (11). We hypothesize that an equivalent fragment formed from hTLT-1 residues functions analogously to block hTLT-1 binding to its natural partner, thereby inhibiting hTLT-1 signaling. Once an hTLT-1 activity is established, we can test whether the intact immunoglobulin-like domain or smaller fragments of that domain can block the activity.
The overall structural conservation within immunoglobulin-like domains and the unique surface and electrostatic properties demonstrate how nature uses the immunoglobulin-like domain as a scaffold for building biological molecules with extreme ligand binding specificity. The presence of a naturally occurring soluble form of murine and human TLT-1 released from platelets suggests a TLT-1 function not directly related to its unique cytoplasmic domain. The hTLT-1 structure suggests experiments to uncover the mechanism of TLT-1 action, once such an activity is established. Together, these details extend the molecular understanding of TLT-1 and provide a foundation to explain continuing biochemical analysis of TLT-1 function.
Acknowledgment-We acknowledge the use of beamline 22-ID of the Southeast Regional Collaborative Access Team, located at the Advanced Photon Source, Argonne National Laboratory.  (1U9K). B, structural alignment of the 17-amino acid segments of mTREM-1 (residues 103-119, shown in magenta) and hTLT-1 (residues 94 -110, blue). C, sequence alignment of equivalent residues between TLT-1 and mTREM in the region of the active mTREM peptide.