T Cell Ig and Mucin Domain–Containing Protein 3 Is Recruited to the Immune Synapse, Disrupts Stable Synapse Formation, and Associates with Receptor Phosphatases Free

The immune system has developed multiple mechanisms to limit T cell responses to self-proteins to prevent autoimmunity. However, these mechanisms also limit T cell responses to cancer Ags and chronic infection (1, 2). During chronic infections, such as HIV type 1 and hepatitis C virus, CD8⁺ T cells become exhausted, causing CD8⁺ T cells to lose their effector function. Loss of production of IL-2 and TNF-α characterizes early exhaustion, whereas the production of IFN-γ is usually maintained until late-stage exhaustion (3–5). Increased expression of multiple negative, coinhibitory checkpoints, including programmed death receptor 1 (PD-1), has been associated with the exhausted phenotype. In addition, antagonizing these receptors with Abs or soluble fusion proteins results in partial rescue of effector function (6). We have previously shown that T cell Ig and mucin domain–containing protein 3 (Tim-3) is highly expressed on exhausted HIV-specific CD8⁺ T cells (4). Similar to the other coinhibitory receptors, Tim-3 blockade partially rescues the proliferation, cytokine production, and cytotoxicity of virus-specific CD8⁺ T cells, suggesting that Tim-3 plays a functional role in T cell exhaustion (4, 7). However, unlike PD-1, Tim-3 is relatively uncharacterized in terms of how it manipulates the cell to dampen T cell responses.

Human Tim-3 is a type I transmembrane protein with extracellular Ig V-like and mucin domains with two N- and one O-linked glycosylation sites (8). Tim-3 is expressed at low levels on naive CD8⁺ T cells, Th1 and Th17 cells, and regulatory CD4 T cells, increased on activated CD8⁺ T cells, and constitutively expressed on NK cells, dendritic cells, monocytes, and macrophages (4, 9–15). Known ligands for murine Tim-3 include phosphatidylserine (16) and galectin-9 (17). The interaction between galectin-9 and Tim-3 is carbohydrate dependent (17), and as such, galectin-9’s carbohydrate-binding, lectin properties suggest that it may also interact or cointeract with other surface glycoproteins including CD44 (18) and integrins (19), allowing association with Tim-3.

Currently, the galectin-9–induced Tim-3 signaling cascade is unknown. We and others have shown that exhausted Tim-3^hi CD8⁺ T cells respond more efficiently to TCR stimulation when the Tim-3 pathway is blocked, suggesting that Tim-3 engagement antagonizes TCR signaling pathways (4, 5, 7, 20–23). Previous studies have investigated Tim-3 signaling in artificial systems such as cell lines and transfection systems. Tyrosine phosphorylation of the Tim-3 cytoplasmic tail and enhancement of this phosphorylation with addition of galectin-9 have been reported in epithelial cell lines (24, 25). In addition, Tim-3 was shown to bind to Fyn, p85 (the PI3K adaptor), and Lck (26, 27), further suggesting a role for Tim-3 in TCR proximal signaling. Finally, Tim-3 expression was shown to suppress NFAT dephosphorylation and AP-1 transcription (28). However, these reports did not study Tim-3 in the context of intact TCR signaling with concurrent galectin-9 engagement. In addition, the pathway still remains to be studied in primary human CD8⁺ T cells. In this study, we characterized the interaction between Tim-3 on primary human CD8⁺ T cells and early signaling events, which may lead to the exhausted phenotype of CD8⁺ T cells found in association with chronic viral infection or cancer.

Healthy HIV-seronegative human volunteers were recruited for blood specimens obtained via leukophoresis. PBMCs were isolated using Ficoll-Paque PLUS (GE Healthcare Biosciences, Uppsala, Sweden). Informed consent was obtained in accordance with the guidelines for conduction of clinical research at the University of Toronto and St. Michael’s Hospital institutional ethics boards.

CD8⁺ T cells were isolated using EasySep Human CD8⁺ T Cell Negative Enrichment Kit (StemCell Technologies, Vancouver, BC, Canada). Cells achieved a purity of at least 95%, assessed via flow cytometry. Isolated CD8⁺ T cells were cultured in R-10 medium (RPMI 1640, 10% FBS, 1 U/ml penicillin, 100 μg/ml streptomycin, 2 mM glutamine; Wisent, Saint-Bruno, QC, Canada) and stimulated with 1 μg/ml LEAF-purified anti-CD3 mAb (clone OKT3; BioLegend, San Diego, CA) and 1 μg/ml LEAF-purified anti-CD28 mAb (clone 28.8; BioLegend) with 50 U/ml rIL-2 (National Institutes of Health, Bethesda, MD) for 5–7 d to upregulate Tim-3 (29). Cells were then washed three times in R-10 and rested overnight in R-10 with no additional stimulation. For Tim-3 blocking experiments, cells were washed in R-10 medium and treated with 10 μg/ml LEAF-purified anti–Tim-3 antagonistic mAb (clone 2E2; BioLegend) or 10 μg/ml LEAF-purified isotype control Ab (BioLegend) overnight without additional stimulation. EBV-transformed patient B cells were cultured in R-10 for up to 20 passages.

A total of 5 × 10⁶ of rested Tim-3^hi CD8⁺ T cells were harvested and resuspended in 20 μl TBS. The cells were lysed in 250 μl 0.5% Brij-58 TKM Lysis Buffer (50 mM Tris-HCl [pH 7.4], 25 mM KCl, 5 mM MgCl₂, 1 mM EDTA, 1 mM NaVO₄, and 5 mM NaF with 1× Roche Complete Mini EDTA-free Protease Inhibitor) for 30 min at 4°C and mixed with equal parts of 80% sucrose in TKM Buffer. The 36% sucrose in TKM Buffer (4.3 ml) was layered on top of the 40% sucrose cell lysate solution and subsequently layered with 5% sucrose in TKM Buffer (0.2 ml). The samples were ultracentrifuged at 50,000 rpm for 16 h at 4°C in a SW Ti-55 rotor (Beckman Coulter, Brea, CA) to isolate lipid rafts, which float to the top of the gradient (fraction 1), whereas nonlipid raft membrane should remain at the bottom of the gradient (fraction 10). The 0.5-ml fractions were collected from each 5-ml gradient. To assess the presence of lipid rafts, 100 μl each fraction was blotted onto nitrocellulose using a dot blot apparatus and probed with a 1:3000 dilution of cholera toxin–HRP (Sigma-Aldrich, Germany) to detect GM-1, a lipid raft–resident ganglioside.

Soluble Tim-3 (sTim-3) was produced and purified as previously described (4) and biotinylated using the EZ-Link Sulfo-NHS-LC-Biotinylation Kit (Thermo Scientific, Rockford, IL). B cells were washed twice in 2% FBS/PBS and incubated with varying amounts of biotinylated sTim-3 per 3 × 10⁵ cells for 30 min at 4°C with or without α-lactose followed by staining with streptavidin-APC (Invitrogen, Carlsbad, CA). B cells were also stained with Galectin-9–PE (BioLegend). Expression of Tim-3 on activated CD8⁺ T cells was assessed via staining with anti-CD8 and anti–Tim-3–PE (R&D Systems, Minneapolis, MN; catalog number FAB2365P). Data were acquired on a BD FACSCalibur (BD Biosciences) and analyzed on FlowJo software (Tree Star).

B cell–T cell conjugates were prepared as described in (30) with some modifications. Briefly, B cells were labeled with Cell Tracker Blue CMAC (Invitrogen) and incubated with 2 μg/ml Staphylococcal enterotoxin B (SEB; Sigma-Aldrich) for 30 min at 37°C. Rested or Tim-3–blocked Tim-3^hi CD8⁺ T cells were mixed 2:1 with the B cells and spun at 200 × g for 5 min, gently resuspended, and further incubated for 10 min at 37°C. The cells were placed onto poly-l-lysine–coated coverslips and allowed to adhere for 15 min. The coverslips were gently rinsed with PBS and fixed with 4% methanol-free formaldehyde for 15 min at 37°C. Cells were washed with PBS and then permeabilized using 0.2% Triton X-100 for 5 min at room temperature then washed with PBS. Blocking was performed using 10% BSA for 1 h at room temperature followed by primary Ab incubation overnight at 4°C using combinations of anti–Tim-3 (R&D Systems; catalog number AF2365), anti-CD3 (Abcam, Cambridge, U.K.; catalog number ab5690), anti–phospho-ERK1/2 (R&D Systems; catalog number MAB1018), phospho-Lck Y394 (R&D Systems; catalog number MAB7500), anti-CD45 (Abcam; catalog number ab33533), and anti-CD148 (Abcam; catalog number ab140236) Abs. Following subsequent washes in PBS, the coverslips were then incubated with appropriate secondary Ab mixes for 45 min at 37°C. Coverslips were mounted on slides using N-propyl gallate mounting buffer and sealed.

The slides were analyzed using a Zeiss LSM700 Confocal Microscope (Carl Zeiss) with a 63× oil immersion objective. Images of T cells, which formed an immunological synapse, were analyzed using LSM Zen LE 2011 Acquisition Software. For semiquantification of stable synapses, slides were analyzed using an epifluorescent microscope with 60× oil immersion objective. Representative pictures were taken and analysis was performed using ImageJ software (National Institutes of Health). Pictures were taken of multiple fields of view, and 25 pairs of T cell–B cell conjugates were counted per slide. Stable synapses were defined as a T cell–B cell conjugate with concentrated CD3 at the interface between the two cells, and the T cell was phospho-ERK⁺. Unstable synapses were defined as a T cell–B cell conjugate in which CD3 was not concentrated at the interface between the two cells or the T cell was not phospho-ERK⁺. Only conjugates with a Tim-3⁺ T cell were included in the counts. The number of stable synapses was blindly counted per 25 pairs of T cell–B cell conjugates.

A total of 1.5 × 10⁷ rested Tim-3^hi CD8⁺ T cells were washed in PBS, resuspended in 1 ml serum-free media, and then incubated at 37°C for 10 min. Cells were treated with or without 1 mM pervanadate for 5 min at room temperature and then lysed in 0.5 ml Nonidet P-40 Lysis Buffer (Invitrogen) at 4°C for 30 min. Cleared lysate was incubated with isotype or Tim-3 Ab (R&D Systems; catalog number AF2365) coupled Protein-G Dynabeads (Invitrogen) for 3 h at 4°C. Beads were washed five times with 1 ml ice-cold lysis buffer, and bound protein was eluted in LDS sample buffer with 50 mM DTT (Invitrogen). For anti-CD3 immunoprecipitations, 2 × 10⁶ rested Tim-3^hi CD8⁺ T cells were pretreated with anti-CD3 Ab-coupled Dynabeads (Invitrogen) for 30 min followed by addition of buffer or in-house produced recombinant galectin-9 for 1 h. The cells were lysed in 0.5% Brij-58 TKM lysis buffer, and anti-CD3 beads were collected. CD3-immunoprecipitated protein was eluted in LDS sample buffer with 50 mM DTT (Invitrogen). Samples were prepared in LDS sample buffer and 50 mM DTT as per the manufacturer’s instructions and analyzed via 4–12% Bis-Tris SDS-PAGE with subsequent Western blotting. Membranes were probed using anti-phosphotyrosine (EMD Millipore, Billerica, MA; catalog number 05-321), anti–Tim-3 (R&D Systems; catalog number AF2365), anti-Lck (Abcam; catalog number ab3885), anti-CD45 (Abcam; catalog number ab33533), anti-CD148 (R&D Systems; catalog number AF1934), and anti-CD3 (Abcam; catalog number ab5690) Abs followed by the appropriate L-chain–specific secondary Abs (Jackson ImmunoResearch Laboratories, West Grove, PA). Detection was performed using chemiluminescence (Thermo Scientific).

The medium isoform of human galectin-9 (UNIPROT ID: O00182-2) with a C-terminal 3XFLAG tag was codon optimized and subcloned into a pET15b vector. Following expression in ArcticExpress (DE3) competent cells (Agilent Technologies, Palo Alto, CA), the cells were induced with 0.5 mM isopropyl β-d-thiogalactoside and incubated at 12°C for 24 h. Cells were harvested and lysed using B-PER protein extraction reagent (Thermo Scientific) as per the manufacturer’s instructions. The lysate was cleared via centrifugation at 25,000 × g for 30 min at 4°C. Functional galectin-9 was isolated using α-lactose agarose (Sigma-Aldrich) and eluted using 200 mM α-lactose. Polishing purification was performed using a Superdex 75 (GE Healthcare, Buckinghamshire, U.K.) via AKTA HPLC. A purity >90% was obtained (assessed via SDS-PAGE and Coomassie and silver staining analysis), and endotoxin was <1 EU/mg protein. Identity of the protein was confirmed via galectin-9 Western blot (R&D Systems; catalog number AF2045).

The binding of recombinant human galectin-9 to recombinant human sTim-3 was tested using a Biacore X instrument (GE Healthcare). A CM5 chip was coupled with 1.82 ng sTim-3 at pH 4.5 (equivalent to 1820 response units). A 2-fold serial dilution of recombinant human galectin-9, from 563 to 18 nM, was injected over the sTim-3 surface, and the change in response units was measured.

A total of 10 μg in-house–produced recombinant galectin-9 was immobilized onto magnetic anti-FLAG beads (Sigma-Aldrich) and incubated with RIPA (Cell Signaling Technology, Beverly, MA) lysate prepared from 5 × 10⁶ rested Tim-3^hi CD8⁺ T cells for 3 h at 4°C. Following collection of the flow-through and subsequent washing with lysis buffer, bound proteins were first eluted with 200 mM α-lactose in lysis buffer followed by a subsequent denaturing elution in LDS sample buffer with 50 mM DTT. Lysate, flow-throughs, and elutions were analyzed via SDS-PAGE and Western blotting or in-gel silver staining (Sigma-Aldrich). Silver-stained bands were excised from the gel and sent for in-gel tryptic digestion and identification via liquid chromatography-tandem mass spectrometry. Analysis of peptide hits was performed using Scaffold 4 software.

TCR signaling predominantly occurs within distinct membrane microdomains called lipid rafts (31, 32). Productive TCR signaling requires shuttling of essential proximal signaling molecules in and out of the rafts during initial interaction with the target cell (33). We tested whether Tim-3 could be found within lipid rafts of CD8⁺ T cells. To do this, primary human CD8⁺ T cells were obtained from peripheral blood of normal human volunteers, and stimulated in culture for 5–7 d with anti-CD3, anti-CD28, and IL-2 to upregulate Tim-3 as previously described in Mujib et al. (29). Following in vitro stimulation, the cells were washed three times to remove residual Abs and IL-2, and the cells were rested overnight in plain medium to dampen any residual proximal signaling. The representative flow cytometry plots in Fig. 1A confirm Tim-3 expression on the CD8⁺ T cells used in all assays. Assessment of Tim-3′s membrane location in these primary rested Tim-3^hi CD8⁺ T cells using sucrose density gradient purification revealed the presence of a large amount of Tim-3 in lipid rafts (Fig. 1B). Approximately 40% of total cellular Tim-3 was found in the first fraction collected from the sucrose gradient, which contained ∼87% of GM-1, a lipid raft ganglioside. Lipid raft clustering is required for immunological synapse assembly (34), which provided us with a rationale to further investigate the role of Tim-3 at the immunological synapse.

Formation of an immunological synapse requires a T cell in conjugate with a target cell to provide a TCR stimulus and surface ligands/receptors on both cells to coordinate multiple costimulatory/inhibitory and immune accessory molecules (35, 36). In order to assess for immunological synapse formation, we used and modified a previous system (30) in which effector CD8⁺ T cells are coincubated with B cell lines as APCs that are pulsed with SEB as the Ag. To confirm the presence of endogenous Tim-3 ligand expressed on APCs, we employed a biotinylated recombinant sTim-3 with labeled streptavidin secondary to stain our target cells (human B cell line). The B cells bound labeled sTim-3, indicating the presence of the Tim-3 ligand on their surface. In addition, the cells stained for an Ab to galectin-9, a Tim-3 ligand (Supplemental Fig. 1A, 1B), indicating that surface associated galectin-9 was present on these cells. To confirm that galectin-9 was acting as the ligand, we treated the cells with sTim-3 in the presence of α-lactose, a galectin-9 substrate. This significantly inhibited the binding of the sTim-3 (Supplemental Fig. 1B), suggesting that galectin-9 was a major ligand for Tim-3 on these cells. However, α-lactose treatment of B cells did not completely abrogate sTim-3 binding, indicating that either there were noncarbohydrate-dependent ligands for Tim-3 on these B cells, as previously described (37), or the affinity of galectin-9 for α-lactose was less than that for Tim-3. Thus, we have confirmed that in the CTL/B cell line system that we will be using to model TCR stimulation, the requisite receptors and ligands are present to allow for simultaneous Tim-3 engagement.

Following conjugate formation with rested Tim-3^hi CD8⁺ T cells and the target (i.e., SEB-loaded B cells), the cells were stained for CD3 (as a marker for the central supramolecular activation cluster) and Tim-3 and then analyzed via confocal microscopy. As shown in Fig. 1C, Tim-3 was recruited to the immunological synapse, suggesting a role for Tim-3 in proximal TCR signaling. Similarly, PD-1 has also been found at the immunological synapse and negatively effects synapse stability, as defined by concentrated and centered CD3 at the T cell synapse (38). Using our conjugate-forming assay, we assessed whether blocking Tim-3 had an effect on synapse stability. Using epifluorescence microscopy, we defined stable synapses as conjugates with concentrated CD3 at the interface between the T cell and APC. In addition, to increase the stringency of our approach, we included phospho-ERK signaling, which results from an integration of signals from both the TCR and activated LFA-1 (39), as an additional parameter to define a stable synapse, which is capable of eliciting downstream signaling. Representative pictures of a stable and unstable synapse are shown in Fig. 2A. Blocking Tim-3 overnight on resting Tim-3^hi CD8⁺ T cells using the antagonistic Tim-3 mAb (10) significantly increased the number of stable synapses formed between T cells and their SEB-pulsed targets to 54(± 3.8)% compared with 32(± 2.8)% for the isotype control (Fig. 2B). This suggests that Tim-3 signaling plays a role in stable synapse formation.

Given that Tim-3 is localized in lipid rafts, which are crucial to immunological synapse formation, and that Tim-3 is located at the immunological synapse, we postulated that Tim-3 plays a role in proximal signaling events. Proximal signaling is regulated, in part, by tyrosine phosphorylation. Although previous studies have shown that Tim-3 can be phosphorylated in epithelial cell lines and is enhanced by addition of galectin-9 (24, 25), it is still undetermined if this is the case in primary CD8⁺ T cells. Although attempts were made to determine Tim-3's phosphorylation status with ligand treatment, we found that the addition of galectin-9 rendered Tim-3 insoluble even in harsh RIPA buffer, which hindered subsequent Tim-3 immunoprecipitation (data not shown). In addition, we could not detect Tim-3 phosphorylation during TCR stimulation, with or without galectin-9 (data not shown). However, it is possible that only a very small fraction of Tim-3 is being phosphorylated during TCR stimulation. To assess the potential for Tim-3 phosphorylation, we immunoprecipitated Tim-3 from rested Tim-3^hi CD8⁺ T cells treated with or without pervanadate, a potent phosphatase inhibitor, and analyzed the elutions via Western blot. Under these conditions, Tim-3 was found to contain phosphorylated tyrosine motifs (Fig. 3A), further suggesting Tim-3′s involvement in proximal signaling. Probing for additional proximal signaling molecules revealed that Tim-3 coimmunoprecipitated with Lck, confirming previous findings (27), but this binding, however, was not dependent upon Tim-3 phosphorylation (lane 1, Fig. 3A). In addition, assessment of conjugates between Tim-3^hi CD8⁺ T cells and SEB-loaded B cells revealed that Tim-3 colocalized minimally, if not at all, with the active form of Lck (p-Y394) at the immunological synapse (Fig. 3B, top panel). Interestingly, large, concentrated areas of active phospho-Lck appeared to exclude Tim-3 (Fig. 3B, bottom panel). This suggests the possibility that Tim-3 may be recruiting a phosphatase to dampen Lck phosphorylation and subsequent activity.

A previous study has demonstrated that mouse Tim-3 binds to mouse galectin-9 (17); however, the data are unresolved on this interaction in humans. To assess the binding of human Tim-3 to human galectin-9, we first produced and purified a FLAG-tagged recombinant human galectin-9 (Supplemental Fig. 2A). Next, we performed two independent surface plasmon resonance experiments to show that immobilized recombinant human Tim-3 was able to bind to recombinant human galectin-9 (Supplemental Fig. 2B). Results were comparable using our in-house produced recombinant human galectin-9 and human soluble Tim-3 and recombinant proteins obtained from R&D Systems (data not shown). However, because the interaction between galectin-9 and Tim-3 is carbohydrate dependent, endogenous, surface-expressed Tim-3 might have a different glycosylation state, which may prevent it from binding to galectin-9. We performed a galectin-9 affinity precipitation to determine if our recombinant human galectin-9 was able to precipitate endogenous Tim-3 from Tim-3^hi CD8⁺ T cells. Galectin-9 bound to endogenous Tim-3 in a carbohydrate-dependent manner, as shown by the presence of Tim-3 in the α-lactose elution (Fig. 4A, lane 4). Following the α-lactose elution, any remaining proteins were removed via boiling/denaturing elution (Fig. 4A, lane 5). A large portion of Tim-3 was found in the denaturing elution, indicating that either the 200 mM α-lactose elution could not completely disrupt the interaction between Tim-3 and galectin-9 or that Tim-3 is also able to bind to galectin-9 in a noncarbohydrate-dependent manner. In order to determine other binding partners of galectin-9, besides Tim-3, the α-lactose elutions were analyzed via SDS-PAGE and silver staining. As shown in Fig. 4B, multiple proteins bound to galectin-9 in a carbohydrate-specific manner. Mass spectrometry analysis identified most of these proteins as immune accessory molecules, including the phosphatases CD45 and CD148 (Fig. 4C). This suggests that although galectin-9 is the ligand for Tim-3, it is able to coordinate multiple surface receptors, including receptor phosphatases.

Because galectin-9 has two carbohydrate recognition domains joined by a flexible linker, it is possible that galectin-9 may colocalize Tim-3 with CD45 or CD148, allowing Tim-3 to associate with a phosphatase in a ligand-dependent manner. Assessment of conjugates between Tim-3^hi CD8⁺ T cells and SEB-loaded B cells showed that Tim-3 colocalized with CD45 and CD148 at the immunological synapse formed between Tim-3^hi CD8⁺ T cells and Tim-3 ligand–expressing APCs (Fig. 5A). Interestingly, addition of recombinant galectin-9 to Tim-3^hi CD8⁺ T cells during CD3 engagement resulted in accumulation of Tim-3, CD45, and CD148 within CD3 signaling complexes, as shown by CD3 immunoprecipitation and subsequent Western blotting (Fig. 5B). These data suggest that Tim-3 may mediate its inhibitory function through the recruitment of the receptor phosphatases to the synapse via its ligand, galectin-9.

Our study provides novel insight into the Tim-3 receptor coordination during immunological synapse formation and involvement in proximal TCR signaling. In this study, we show that human Tim-3 is found in lipid rafts and recruited to the immunological synapse. Further, Tim-3 may be involved in the process of stable synapse formation, suggesting a role for Tim-3 in proximal signaling. Tim-3 was found to have the potential for tyrosine phosphorylation in primary CD8⁺ T cells and bind to Lck, further confirming the finding by Rangachari et al. (27) that Tim-3 associates with this essential proximal signaling molecule. However, Tim-3 was not found to associate with the active form of Lck, phospho-Y394, at the synapse, suggesting that Tim-3 may be recruiting a phosphatase. Interestingly, in addition to confirming galectin-9 as the Tim-3 ligand, we described novel interactions between galectin-9 and receptor phosphatases CD45 and CD148. Finally, galectin-9 was able to enhance the interaction between these phosphatases and Tim-3 within CD3 signaling complexes. These data suggest the possibility that Tim-3 may regulate Lck function via interaction with CD45 or CD148 at the synapse, an interaction mediated by its ligand galectin-9.

Proximal TCR signaling is a complex process regulated not only by posttranslational modifications, including tyrosine phosphorylation, but also by lipid raft dynamics and compartmentalization of proteins. For T cells to initiate signaling, first the peptide-loaded MHC interacts with the TCR. This results in recruitment to the lipid rafts and exclusion of CD45 (33). Fyn and/or non-CD8–bound Lck become associated with the TCR–CD3 signaling complex, resulting in phosphorylation of CD3 ITAMs and recruitment of ZAP70. Following ZAP70 phosphorylation, the Src homology 2 (SH2) domain of CD8-bound Lck binds to phospho-ZAP70, recruiting CD8 to the MHC–TCR complex. CD8 binds to the MHC, increasing the stability of the MHC–TCR complex (40), sustaining proximal signaling and initiation of downstream signaling.

The finding that >40% of Tim-3 resides in lipid rafts and is recruited to the immunological synapse provides evidence that Tim-3 is involved in proximal signaling. Although this study assessed membrane location of Tim-3 on resting cells, further work is warranted to determine if Tim-3 accumulates within or is excluded from lipid rafts during TCR signaling, similar to Lck and CD45, respectively. However, because TCR triggering results in the accumulation of lipid rafts at the synapse (34), it is likely that Tim-3 would accumulate. In addition, we found that blocking Tim-3 using the antagonistic Ab 2E2 resulted in significantly more conjugates that had stable synapses. Initial proximal TCR signaling, such as Lck phosphorylation and subsequent downstream signaling to linker for activation of T cells, is required for stable synapse formation (41), further providing evidence that Tim-3 is involved in TCR proximal signaling. Although the exact mechanism of how Tim-3 effects proximal signaling is still under investigation, our work provides preliminary evidence that Tim-3 may be involved in Lck regulation.

Our finding that Tim-3 binds to Lck is not novel (27); however, we did show that this interaction occurred in primary human CD8⁺ T cells. In addition, we showed that Tim-3 did not localize with the active form of Lck, phospho-Y394. Lck activity is regulated via phosphorylation of the positive regulatory tyrosine Y394 in the kinase domain and the negative regulatory tyrosine Y505 in the C-terminal end (33). When Y505 is phosphorylated, it forms a bond with its own SH2 domain, preventing substrate access to the kinase domain. However, when Y394 is phosphorylated, the protein achieves full kinase activity. It would be prudent to assess Tim-3′s location in relation to the inactive form of Lck (pY505) versus the active form of Lck (pY394); however, the inactive form pY505 is bound to its own SH2 domain, preventing access to the phosphorylated motif by any Ab used for immunofluorescence. Thus, we are unable to assess, given currently available reagents, whether Tim-3 preferentially binds pY505 Lck. However, directly assessing Lck activity in Tim-3 immunoprecipitates using immune complex kinase assays would be more accurate rather than inferring activity based on the phosphorylation state of Tim-3–precipitated Lck. Thus, further studies to characterize the activity of Tim-3–associated Lck are warranted. Phosphorylation of Y505 and Y394 are achieved by the kinase Csk and via trans/autophosphorylation, respectively (33). Dephosphorylation of either tyrosine is mediated by the receptor phosphatase CD45. Interestingly, CD45 plays a dual role in positively and negatively regulating Lck, depending on the expression level of CD45 (42). Low to intermediate concentrations of CD45 result in dephosphorylation of Lck at Y505, resulting in an increase in Lck activity. However, high levels of CD45 result in dephosphorylation of Y394, effectively decreasing Lck kinase activity.

Despite a recent study that claimed human Tim-3 was not the receptor for human galectin-9 (43), we confirmed that galectin-9 was the physiological ligand for Tim-3 using two different biochemical assays. In addition, we discovered that galectin-9 bound to receptor phosphatases CD45 and CD148, suggesting a possible mechanism of how Tim-3 could regulate Lck activity. Inducing high local concentrations of CD45 in proximity to Tim-3 and associated Lck could result in dephosphorylation of Lck Y394 and Y505. In addition, we showed that galectin-9 was able to increase levels of Tim-3, CD45, and CD148 within CD3 signaling rafts. Thus, although CD45 and CD148 are normally excluded from the synapse (33, 44), concurrent ligation of these receptor phosphatases with Tim-3 via galectin-9 could enhance the presence of these phosphatases at the synapse, resulting in negative regulation of Lck as well as dampening of TCR signaling.

Importantly, all of this work was performed using primary CD8⁺ T cells with minimal manipulation of the T cells or target cells. Although Tim-3 expression was induced via in vitro stimulation, no proteins were ever overexpressed, which can lead to artificial results and false positives, such as chaperones binding to overexpressed proteins (45). In addition, this work studied Tim-3 under near physiological conditions. Use of primary CD8⁺ T cells provides an intact TCR signaling platform, including presence of proximal signaling molecules, regulation via lipid raft dynamics, and presence of receptor phosphatases that are essential for the initiation and final dampening of TCR signaling (44, 46, 47). This is absent in many of the cell lines used to study Tim-3 thus far (24, 25, 27). In addition, our system uses SEB loaded on target cells to provide a TCR stimulus, which does not bypass proximal signaling, unlike the PMA/ionomycin stimulus used to study Tim-3 (28). Finally, we provide a target cell, which expresses the physiological Tim-3 ligand as shown by sTim-3 binding to the B cell in a carbohydrate-specific manner. This is essential because previous studies have engaged Tim-3 using putative agonistic Abs (13, 27). Galectin-9 is a soluble protein, not a transmembrane protein, like programmed death ligand-1 and programmed death ligand-2. It can associate with the membrane via interaction with other glycans; however, its behavior during T cell engagement and synapse formation is currently unknown, stressing the importance to understand how galectin-9 effects Tim-3 before using Abs to engage Tim-3. Indeed, using a galectin-9 affinity precipitation, we did show that galectin-9 was able to bind to various immune accessory molecules. Thus, the Tim-3–galectin-9 interaction is not a closed system but a complex interplay with various other surface glycosylated molecules. Although tempted to use agonistic Abs to study the effects of Tim-3, the reality is that when Tim-3 is engaged under physiological conditions, so are other receptors, including receptor phosphatases like CD45.

Using the data we have collected, we wish to propose a working model. In the case of Tim-3 expressed on the T cell and galectin-9 expressed on the surface of the APCs, both CD45 and Tim-3 get recruited to the TCR signaling clusters at the immunological synapse. This results in CD45-mediated dephosphorylation of the active tyrosine (Y394) of Lck, and accumulation of inactive Lck within TCR signaling clusters. This increases the ratio of inactive Lck to active Lck and eventual dampening of TCR signaling. These results, thus far, are observational, and future functional investigations focusing on the Tim-3 intracellular pathway and galectin-9–mediated effects using our physiological T cell/APC system are warranted.

Efforts need to be made to understand the signaling pathways associated with coinhibitory receptors to allow for better development of therapeutic interventions. The signaling pathways for coinhibitory receptors CTLA-4 and PD-1 are currently known. In addition, these proteins have antagonistic Abs currently used in clinic or in phase III trials, respectively, for cancer treatment (48). Coinhibitory molecules are not redundant, as shown by the synergistic effects of dual blockade. This stresses the importance of understanding what specific effects each inhibitory receptor has on T cell function to better formulate blockades that are most effective against different diseases such as cancer and chronic viral infections.

We thank Dr. Michael Julius for help and expertise with the lipid raft isolation, Dr. Andrew Wilde for the use of microscope equipment, Dr. Walid Houry for help and expertise with the protein purifications and use of equipment, and Dr. R. Brad Jones for helpful comments.

The authors have no financial conflicts of interest.