A Disintegrin and Metalloprotease (ADAM) 10 and ADAM17 Are Major Sheddases of T Cell Immunoglobulin and Mucin Domain 3 (Tim-3)*

Background: Membrane-bound Tim-3 regulates immune responses. Results: Soluble Tim-3 is generated by A disintegrin and metalloprotease (ADAM) 10 and ADAM17. Conclusion: ADAM proteases are involved in Tim-3 processing. Significance: Shedding of Tim-3 influences immune cell activation. T cell immunoglobulin and mucin domain 3 (Tim-3) dampens the response of CD4+ and CD8+ effector T cells via induction of cell death and/or T cell exhaustion and enhances the ability of macrophages to clear pathogens via binding to galectin 9. Here we provide evidence that human Tim-3 is a target of A disintegrin and metalloprotease (ADAM)-mediated ectodomain shedding resulting in a soluble form of Tim-3. We identified ADAM10 and ADAM17 as major sheddases of Tim-3 as shown by ADAM-specific inhibitors and the ADAM10 pro-domain in HEK293 cells and ADAM10/ADAM17-deficient murine embryonic fibroblasts. PMA-induced shedding of Tim-3 was abrogated by deletion of amino acids Glu181–Asp190 of the stalk region and Tim-3 lacking the intracellular domain was not efficiently cleaved after PMA stimulation. Surprisingly, a single lysine residue within the intracellular domain rescues shedding of Tim-3. Shedding of endogenous Tim-3 was found in primary human CD14+ monocytes after PMA and ionomycin stimulation. Importantly, the recently described down-regulation of Tim-3 from Toll-like receptor-activated CD14+ monocytes was caused by ADAM10- and ADAM17-mediated shedding. Inhibition of Tim-3 shedding from lipopolysaccharide-induced monocytes did not influence lipopolysaccharide-induced TNFα and IL-6 but increases IL-12 expression. In summary, we describe Tim-3 as novel target for ADAM-mediated ectodomain shedding and suggest a role of Tim-3 shedding in TLR-mediated immune responses of CD14+ monocytes.

The T cell immunoglobulin and mucin domain (Tim) 2 family consists of 3 human (Tim-1, -3, and -4) and 4 murine (Tim-1 to -4) Tim proteins (1). Tim-3 is a type I cell-surface glycoprotein with a N-terminal immunoglobulin (Ig)-like domain, a mucin domain with O-linked and N-linked glycosylation, a single transmembrane domain, and a cytoplasmic region with a tyrosine phosphorylation motif (2). Proteins of the Tim family are critically involved in the regulation of immune responses, including allergy, asthma, transplant tolerance, autoimmunity, the response to viral infections, and cancer (1). Galectin-9 is one ligand of Tim-3. Binding of galectin-9 to Tim-3 on T-helper 1 (Th1) and Tc1 cells (a subpopulation of CD8 ϩ cytotoxic T cells) induces cellular apoptosis and Tim-3 deficiency reduces galectin-9-mediated Th1 cell death in vivo (3). In line, downregulation of Tim-3 allows autoreactive T cells to escape negative regulation in multiple sclerosis (4).
Tim-3 is also highly expressed or up-regulated in exhausted CD8 ϩ T cells in various chronic viral infections (5)(6)(7)(8) and in tumor bearing hosts (9 -12). Exhausted T cells did not proliferate and fail to exert functions such as cytotoxicity and cytokine secretion in response to antigen stimulation. Interestingly, exhaustion of T cells can be partially overcome by blockade of the interaction of Tim-3 and its ligand. Another hallmark of exhausted T cells is the expression of the inhibitory molecule programmed cell death 1. Like Tim-3, blockade of programmed cell death 1 and programmed cell death 1 ligand (PD-L1) interactions can partially reverse T cell exhaustion (13,14). Furthermore, blocking of Tim-3 and programmed cell death 1 synergistically restores T cell proliferation, enhances T cell cytokine production, and is effective in restoring anti-tumor immunity in vivo (9). It is, however, not clear whether T cell exhaustion is correlated with interaction of Tim-3 to galectin-9 or another ligand (15).
Tim-3 is also expressed on cells of the innate immune system and can synergize with Toll-like receptors to promote TNF␣ secretion (16). On CD14 ϩ monocytes, Tim-3 synergizes with TLR signaling to dampen IL-12 secretion with almost no influence on TNF␣ secretion (17).
Alternative splicing of Tim-3 has been described to result in a hypothetical soluble Tim-3 protein (18). Administration of a human soluble Tim-3-Fc fusion protein caused hyper-proliferation of Th1 cells and Th1 cytokine release and may serve as a inhibitor of endogenous Tim-3 function (18). Furthermore, recombinant mouse soluble Tim-3 inhibited T cell responses to antigen-specific stimulation (19). A naturally occurring sTim-3 protein might have antagonistic properties. However, the occurrence of a sTim-3 protein in vivo remains to be shown.
Ectodomain shedding or limited proteolysis of membranebound proteins results in protein down-regulation on the cell surface and the production of soluble protein ectodomains with agonistic or antagonistic properties. Members of the A Disintegrin and mtalloprotease (ADAM) gene family have emerged as major ectodomain shedding proteinases. With more than 100 described substrates so far, ADAM17 and its close relative ADAM10 are the major sheddases of this family (20). There is, however, extensive overlap and compensation between ADAM proteases for several substrates (21,22). Different stimuli including phorbol ester (phorbol 12-myristate 13-acetate (PMA)), ionomycin, ligands of G protein-coupled receptors, ATP, bacterial toxins, bacterial metalloproteinases, and apoptosis activate ADAM10-and/or ADAM17-mediated shedding of transmembrane proteins (20). For some ADAM target proteins such as Notch, induction of intracellular signaling by the remaining intracellular domain cleavage product has been described (23).
Here, we discovered Tim-3 as a novel substrate of ADAM10 and ADAM17, resulting in the release of a soluble Tim-3 protein. A 10-amino acid deletion from Glu 181 -Asp 190 completely abrogated ADAM17-mediated shedding of Tim-3. Whereas deletion of the intracellular domain of Tim-3 largely abrogates PMA-induced shedding, a single lysine residue of the intracellular domain rescued PMA-induced shedding. Finally, down-regulation of Tim-3 from LPS-activated primary human CD14 ϩ monocytes was mediated by ADAM10-mediated shedding.
Antibody Production-The pcDNA3.1-hTim-3 plasmid was used as a template to amplify the IgV-domain of the human Tim-3 with primers 5ЈIgVhTim-3 (5Ј-GATCCATATGTCAG-AAGTGGAATACAGAG-3Ј) and 3ЈIgVhTim-3 (5Ј-GACTG-CGGCCGCTCCCTTGGCTGGTTTGA-3Ј). The resulting PCR product was subcloned into the bacterial expression vector pET23a (Novagen, Darmstadt, Germany) using NdeI and NotI restriction sites. The pET23a-hTim-3-IgV was transformed into Escherichia coli BL21. Transformed cells were grown at 37°C Lysogeny Broth (LB) medium supplemented with ampicillin (100 g/ml), and protein expression was induced by 1 mM isopropyl 1-thio-␤-D-galactopyranoside at A 600 0.6 -0.8 at 37°C. After overnight culture, the bacteria were harvested by centrifugation and lysed using sonification. The resulting inclusion bodies containing the hTim-3-IgV protein were denaturated using 6 M guanidine HCl in 50 mM Tris buffer, pH 8. For renaturation the refolding buffer contained 0.1 M Tris-HCl, pH 8.5, 400 mM L-arginine, 2 mM EDTA, 4 mM reduced glutathione and 0.4 mM oxidized glutathione was used. The refolded hTim-3-IgV was purified via a NAP-25 column (GE Healthcare). Fractions, which contained the hTim-3-IgV, were purified via Superdex75 16/60 column (GE Healthcare) equilibrated in PBS. The monomeric protein was collected and concentrated to 100 g/ml. Wistar rats were initially immunized intraperitoneally with 100 g of purified hTim-3-IgV domain in 60 l of phosphate-buffered saline (PBS) emulsified with 40 l of Gerbu adjuvant MM (Gerbu, Heidelberg, Germany). The rats were boosted intraperitoneally on days 14 and 21 with 50 g of purified protein emulsified with 20% of the adjuvant. The last two doses (50 g of hTim-3-IgV in PBS) were performed on days 28 and 29 without adjuvant, whereas fusion was done on day 30. Spleen cells from immunized animals were collected and fused with Ag8.653 myeloma cells using polyethylene glycol 1500 (Roche Diagnostics). The fused cells were cultured in selection medium (hypoxanthine/aminopterin/thymidine medium, Sigma) for 10 days and screened by ELISA for anti-hTim-3-IgV antibodies.
Purification of mAbs-The hybridomas were cultivated in serum-free medium (RPMI 1640 hybridoma express plus; PAA) supplemented with ␤-mercaptoethanol and antibiotics (penicillin/streptomycin; PAA) and incubated with 5% CO 2 at 37°C. After 2 weeks, the supernatants were pooled and filtered on a 0.22-m filter and concentrated 10 times. The purification was performed using HiTrap Protein G HP (GE Healthcare) according to the manufacturer's procedure. Briefly, the column was equilibrated with binding buffer (20 mM phosphate, pH 7.4) followed by loading of the sample. After washing with binding buffer, the mAbs were eluted with 0.1 M glycine, pH 2.7, and directly neutralized with 1 M Tris-HCl, pH 9.0. The buffer was changed by NAP-10 column (GE Healthcare) to PBS. The purity of the mAbs was verified by SDS-PAGE (data not shown).
Western Blotting-For immunochemical detection of Tim-3 proteins, cells were washed three times with sterile PBS. Inhibitors were added 30 min before stimulation with PMA, ionomycin, or W7. Stimulation was performed in serum-free DMEM. Subsequently, cells were centrifuged and the pellet was directly frozen in liquid nitrogen. Cells were lysed in lysis buffer (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.5 mM EDTA, and 0.5% IGEPAL (Nonidet P-40), supplemented with complete protease inhibitor mixture tablets (Roche Diagnostics). The conditioned media were concentrated to final volume of 300 l by Speed-Vac Plus SC110A (ThermoScientific, Dreieich, Germany). Proteins were separated by SDS-PAGE and transferred to a polyvinylidene difluoride membrane (Whatman-Fisher Scientific GmbH, Schwerte, Germany). The membrane was blocked with 6% skimmed milk in Tris-buffered saline with Tween 20 (TBS-T; 10 mM Tris-HCl, pH 7.6, 150 mM NaCl, and 0.5% Tween 20) and probed with primary antibodies as indicated at 4°C overnight. After washing with TBS-T, the membranes were incubated with the appropriate secondary antibodies conjugated to horseradish peroxidase (HRP; Thermo Scientific/ Pierce, Perbio, Bonn, Germany), and protein bands were visualized with the Pierce ECL Western blotting substrates (Thermo Scientific Inc., Rockford, IL) according to the manufacturer's instructions.
Phosphatase Assay-For detection of AP-Tim-3 proteins, cells were treated as described for Western blotting. The conditioned media were collected and directly frozen in liquid nitrogen. For alkaline phosphatase activity measurements, 50 l of reaction solution (0.1 M glycine, 1 mM MgCl 2 , 1 mM ZnCl 2 containing 1 mg/ml of 4-nitrophenyl phosphate disodium salt hexahydrate, Sigma) were added to 50 l of the conditioned medium. The alkaline phosphatase removes the phosphate group from the p-nitrophenyl phosphate, which is then deprotonated under alkaline conditions to produce p-nitrophenolate that has a strong absorption at 405 nm. So the absorbance was read at 405 nm and refers to the alkaline phosphatase activity. The absorbance of the AP-hTim-3 proteins in the supernatant was normalized to the transfection efficiency of each sample, by division of the total absorbance of the transfected cells (sum of the absorbance of the supernatant and the lysate).
Isolation and Cultivation of Human CD14 ϩ Monocytes-Human peripheral blood mononuclear cells (PBMC) were isolated from peripheral blood using lymphocyte separation medium LSM 1077 (PAA) density gradient centrifugation. CD14 ϩ monocytes were further purified from PBMCs by negative selection using monocyte Isolation KitII with column purification following the manufacturer's instructions (Miltenyi Biotec, Bergisch-Gladbach, Germany). The cells were cultured in RPMI 1640 containing 10% FCS, 100 g/ml of penicillin-streptomycin, and 2 mM L-glutamine (PAA). All patients underwent a written, informed-consent process approved by the ethics commission of the Medical Faculty of Kiel University (AD 404/12).
Flow Cytometry Staining and Analysis-PBMCs from 3 to 5 healthy subjects were stimulated with 5 g/ml of LPS (Invitrogen) for various time points as indicated. Cell-surface staining was carried out using anti-Tim-3-PE clone F38 -2E2 (BioLegend, London, Great Britain) or isotype-matched control antibodies (BD Biosciences) to determine the levels of background staining. The cells were analyzed by flow cytometry (BD Biosciences; FACSCanto and FACS DIVA software).
To detect the surface expression of hTim-3 and mutants thereof, transiently transfected HEK293 cells were washed with FACS buffer (PBS, 0.25% bovine serum albumin (BSA) and incubated at 5 ϫ 10 5 cells/100 l of FACS buffer containing 1:20 diluted PE-anti-human Tim-3, clone F38-2E2 (Biolegend) or PE-mouse IgG1, isotype control in FACS buffer for 60 min at 4°C. Cells were washed once with FACS buffer, re-suspended, and analyzed by flow cytometry (BD Biosciences; FACSCanto and FACS DIVA software). For the combination of different experiments, the amount of Tim-3 expressing cells was set to 100% and loss of Tim-3 was calculated.
ELISA-Cytokine levels were measured via ELISA using DuoSet ELISA kits (R&D Systems).

ADAM10 and ADAM17 Are the Major Sheddases of Tim-3-
Tim-3 is a type I membrane protein and consists of a N-termi-nal Ig-like domain, a highly glycosylated mucin stalk, a single transmembrane domain, and a cytoplasmic region. HEK293 cells were transfected with an expression plasmid coding for AP-tagged Tim-3. AP-Tim-3 cell-surface expression was verified by flow cytometry using an anti-Tim-3 mAb in HEK293 cells (Fig. 1A). To test whether TIM-3 is a substrate for ADAM proteases, AP-Tim-3 expressing HEK293 cells were stimulated with PMA or ionomycin in the absence or presence of the ADAM10 inhibitor GI254023X (GI) and ADAM10/ADAM17 inhibitor GW280264X (GW) or left untreated. Ionomycin and benzoyl-ATP are activators of ADAM10-mediated ectodomain shedding and PMA is a general activator of ADAM17-mediated shedding (23). PMA and ionomycin treatment of AP-Tim-3 expressing HEK293 cells led to the release of soluble AP-Tim-3 into the cell culture supernatant as quantified by an alkaline phosphatase assay (Fig. 1B). The combined ADAM10/ADAM17 inhibitor GW but not the ADAM10 inhibitor GI completely inhibited the PMA-induced and constitutive release of AP-Tim-3 into the cell culture supernatant. Ionomycin-induced shedding was inhibited by both, the ADAM10 inhibitor GI and the ADAM10/ADAM17 inhibitor GW (Fig. 1B). The presence of a soluble AP-Tim-3 protein with the expected size was detected by Western blotting using anti-Tim-3 mAb directed against the Ig-like domain. Again PMA-induced shedding was only inhibited by GW, whereas ionomycin-induced shedding was inhibited by GI and GW (Fig. 1C). Next, we inhibited ADAM10 with the recombinant pro-domain of ADAM10 (proA10) (28) in PMA-and ionomycin-treated HEK293 cells, transiently transfected with AP-Tim-3 cDNA. As expected, proA10 inhibits ionomycin-induced AP-Tim-3 shedding. However, it appears also that PMA-induced shedding of AP-Tim-3 was partly inhibited by proA10 (Fig. 1, D and E), which might be due to partial blocking of ADAM17 (29). We have, however, not observed cross-inhibition with proADAM10 of ADAM17-mediated shedding of IL-6R (21). MEFs deficient for ADAM10, ADAM17, and ADAM10/ADAM17 were transiently transfected with the cDNA coding for AP-Tim-3. As depicted in Fig.  1, F and G, shedding of AP-Tim-3 was induced by PMA and ionomycin treatment in wild-type MEFs. PMA-induced shedding was, however, completely abrogated in ADAM17-deficient MEFs but only minimally affected in ADAM10-deficient MEFs. Ionomycin-induced shedding was almost completely inhibited in ADAM10-deficient MEFs but not in ADAM17deficient MEFs. Shedding of Tim-3 was completely inhibited in ADAM10/ADAM17 double-deficient MEFs after PMA and ionomycin treatment.
To exclude that the shedding results obtained with AP-tagged Tim-3 were influenced by the AP tag, we removed the AP tag from the AP-Tim-3 cDNA. N terminally FLAG-tagged human Tim-3 was stably transfected in HEK293 cells, and cell-surface expression of Tim-3 was verified by flow cytometry (Fig. 1H) Tim-3 in cell lysates and supernatants. As AP-tagged Tim-3, Tim-3 was shed after PMA and ionomycin treatment from HEK293 cells (Fig. 1I). Again, PMA-induced shedding of Tim-3 was inhibited by GW but to a much lower extend by GI or proA10 (Fig. 1, I and J). Ionomycin-induced shedding of Tim-3 was inhibited by GI, GW, and proA10 (Fig. 1, I and J). Taken together, these results indicated that Tim-3 is a substrate of ADAM proteases. Whereas ionomycin-induced ADAM10-mediated Tim-3 shedding, PMA-induced ADAM17-mediated shedding of Tim-3. Our inhibitor studies but not the data obtained from gene-deficient MEFs indicate that ADAM10 might also play a minor role in PMA-induced shedding of Tim-3.
PMA-induced Shedding of Tim-3 Was Abrogated by Deletion of 10 Amino Acids from the Stalk Region-Shedding of cellsurface proteins by ADAM proteases is described to occur in close juxta-position to the plasma membrane (23). For instance, the ADAM17-cleavage site for the interleukin-6 receptor, a substrate for ADAM10 and ADAM17, is located between Gln 357 and Asp 358 about 8 amino acids away from the plasma membrane, and deletion of 10 amino acids surrounding the cleavage site (Ser 353 -Val 362 ) completely inhibited PMA-induced shedding by ADAM17 (30) but not of ADAM10 (31). The prediction of ADAM cleavage sites is complicated by the fact that no clear consensus cleavage sequence exists in substrates of ADAM10 and ADAM17. A tendency, however, for smaller aliphatic residues at P1Ј (Val, Ala) for ADAM17 and ADAM10 is eminent (32). This illustrates why it is not possible to predict ADAM10 and ADAM17 substrates using sequence analysis. The stalk of Tim-3 compromises 70 amino acids from Ala 132 to Ile 201 . Here, we generated 3 different variants of Tim-3 (⌬S1, ⌬191-201; ⌬S2, ⌬181-190; ⌬S3, ⌬171-180) each lacking 10 amino acids prior to the transmembrane domain ( Fig. 2A). As shown in Fig. 2B, all three Tim-3 deletion variants were expressed on the cell surface of transiently transfected HEK293 cells. PMA-induced shedding was, however, completely abrogated for AP-Tim-3⌬S2 but not affected in AP-Tim-3⌬S1 and AP-Tim-3⌬S3 (Fig. 2, C and D). Again, AP fusion did not influence the shedding behavior of Tim-3 (Fig. 2E). Ionomycin-induced shedding was not changed for any of the three Tim-3 deletion variants (Fig. 2, C-E). Although, we have not determined the cleavage site(s) for ADAM17 directly, we conclude from these experiments that the cleavage site for PMA-induced proteolysis of Tim-3 (mediated by ADAM17) is located between amino acid positions 181 and 190. Alternatively, recognition of Tim-3 by ADAM17 might be regulated by this sequence. Our results also illustrate that the cleavage sites for ADAM10 and ADAM17 in Tim-3 are different. Our results open up the possibility to generate ADAM17-uncleavable knock-in mice for Tim-3.
Calmodulin Contributes to Tim-3 Cell Surface Expression-The proteolytic processing of Tim-3 shares some common feature with the shedding of L-selectin. The expression of L-selectin is rapidly down-regulated upon cell activation through proteolysis at a membrane-proximal site (35). The calcium regulatory protein calmodulin (CaM) is directly associated with the cytoplasmic and transmembrane domains of L-selectin. Calmodulin antagonists including N-(6-aminohexyl)-5-chloro-1-naphthalenesulfonamide (W7) induce the proteolytic release FIGURE 2. PMA-induced shedding of Tim-3 was abrogated by deletion of 10 amino acids from the stalk region. A, schematic overview of the AP-Tim-3 constructs used in this study. Three constructs with three different deletions of 10 amino acids each from the mucin stalk region (AP-Tim-3⌬S1 to AP-Tim-3⌬S3) were created as described. All constructs contain the alkaline phosphatase at the N terminus. B, HEK293 cells were transfected with the expression plasmids for AP-Tim-3 deletion mutants (AP-Tim-3⌬S1 to AP-Tim-3⌬S3) and immunostained for Tim-3 using anti-Tim-3-PE clone F38 -2E2. C, HEK293 cells were transfected with expression plasmids encoding AP-Tim-3 or the deletion mutants (AP-Tim-3⌬S1 to AP-Tim-3⌬S3) and treated for 1 h with PMA (100 nM) or 30 min with ionomycin (5 M). The alkaline phosphatase activity was measured in the lysate and conditioned medium. The AP activity is calculated as the ratio of the alkaline phosphatase activity in the lysate and supernatant. D, membrane-bound Tim-3 in the cell lysates and soluble Tim-3 in the conditioned medium were subsequently visualized by Western blotting using anti-hTim-3 IgV-domain mAbs (description under "Experimental Procedures"). The values are from one representative experiment (n ϭ 5). E, HEK293 cells were transfected with an expression plasmid encoding Tim-3 deletion mutants (Tim-3⌬S1 to Tim-3⌬S3) without alkaline phosphatase at the N terminus and treated for 1 h with PMA (100 nM) or 30 min with ionomycin (5 M). Membrane-bound Tim-3 in the cell lysates and soluble Tim-3 in the conditioned medium were subsequently visualized by Western blotting using anti-hTim-3 IgV-domain mAb (description under "Experimental Procedures"). The values are from one representative experiment (n ϭ 5). n.s., not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001 (unpaired Student's t tests).
of L-selectin from the cell surface. These effects can be prevented by co-treatment with hydroxamic acid-based metalloprotease inhibitors. The binding of CaM also to other proteins such as collagen receptors or amyloid precursor protein prevents shedding and disruption of these interactions result in the activation of ADAM-mediated ectodomain shedding (36).
HEK293 cells expressing AP-Tim-3 were stimulated with W7 in the absence or presence of GI and GW or left untreated. AP-Tim-3 protein released into the supernatant was quantified by AP assay. W7 treatment led to the release of soluble AP-Tim-3 into the cell culture supernatant as shown by the AP assay and Western blotting (Fig. 4, A and B). The ADAM10 inhibitor GI and combined ADAM10/ADAM17 inhibitor GW prevented the W7-induced release of AP-Tim-3 into the cell culture supernatant (Fig. 4, A and B), with GW having a stronger effect.
Next, we investigated if the PMA-resistant Tim-3 variant AP-Tim-3⌬S2 is proteolytically cleaved after W7 treatment. As expected, the W7-induced shedding was also nearly completely abrogated for AP-Tim-3⌬S2 as shown before for PMA-induced cleavage (Fig. 4, C and D). Taken together, these data indicated that W7 induced ADAM10-and ADAM17-mediated shedding of Tim-3.
CaM binds to the membrane-proximal region of the cytoplasmic domain plus a small region of the membrane-spanning domain of L-selectin (37). To analyze if the CaM-binding site of Tim-3 is also located in the intracellular domain of Tim-3, we used the AP-Tim-3 cDNA lacking the coding sequence of the complete intracellular domain from Lys 225 to Gln 301 (AP-Tim-3⌬ICD) and the deletion variant AP-Tim-3⌬W226-Gln 301 (⌬ICDϩ1), which differs from AP-Tim-3⌬ICD by the addition of the first intracellular residue Lys 225 . Shedding of AP-Tim-3⌬ICD was strongly diminished in comparison to the AP-Tim-3, whereas the deletion variant AP-Tim-3⌬W226-Gln 301 (⌬ICDϩ1) was shed upon CaM inhibition, which was reduced by GI and GW (Fig. 4, E-G).
To exclude that results obtained with AP-tagged Tim-3 were influenced by the AP tag, we used HEK293 cells transiently transfected with Tim-3. Shedding of Tim-3 was verified by Western blotting against Tim-3 in cell lysates and supernatants. As AP-tagged Tim-3, Tim-3 was shed after W7 treatment from HEK293 cells (Fig. 4H). Again, W7-induced shedding of Tim-3 was inhibited by GW and GI, albeit to a lower extent (Fig. 4, A  and F). Shedding of Tim-3⌬S2 and Tim-3⌬ICD was also comparable with the AP-tagged variants in transfected HEK293 cells (Fig. 4, A and F). Finally, we showed that CaM was precipitated in a complex with Tim-3, indicating that CaM interacts with Tim-3 but also with all deletion mutants (Fig. 4I). The finding that the intracellular domain of Tim-3 was negligible for calmodulin-prevented shedding was surprising. It has been recently shown that calmodulin not only binds to residues within the intracellular domain but also to areas within the transmembrane domain (34). Therefore calmodulin might also interact with the intracellular domain of Tim-3. We cannot, however, exclude at the moment, that calmodulin is interacting with another protein involved in regulation of ADAM-medi-ated shedding of Tim-3, which might be part of a higher order Tim-3 protein complex. Taken together, these results show that ADAM-mediated shedding of Tim-3 is CaM-dependent and mediated by ADAM10 and -17.

PMA-and Ionomycin-induced Shedding of Endogenous Tim-3 in Primary Human CD14 ϩ Monocytes-To analyze if endogenous
Tim-3 is also a substrate of ADAM-mediated shedding, we used CD14 ϩ monocytes isolated from PBMCs, which naturally express Tim-3 (17). CD14 ϩ monocytes also express ADAM10 and ADAM17 on the cell surface (data not shown). At least 49% (varies between 49 and 77% from experiment to experiment) of the isolated CD14 ϩ monocytes expressed Tim-3 on the cell surface as judged by flow cytometry (Fig. 5). Ionomycin treatment led to a drastic reduction of Tim-3 cell-surface expression and only about 7.4% of the cells showed cell-surface expression of Tim-3. Importantly, ionomycin-induced loss of cell surface Tim-3 was prevented by co-incubation with GI or GW (Fig. 5, A and B), indicating that the loss of cell surface Tim-3 was due to ADAM10-mediated shedding of Tim-3. Albeit to a lesser extent, PMA treatment led to a reduction of Tim-3 from the cell surface of CD14 ϩ monocytes and only about 30.9% (untreated 64%) of the cells showed cellsurface expression of Tim-3. Loss of cell surface Tim-3 was rescued by co-incubation with the combined ADAM10 and ADAM17 inhibitor GW (50.6%) and to a lesser degree by the ADAM10 inhibitor GI (42.7%) (Fig. 5, C and D). Taken together, endogenous Tim-3 is a substrate of ADAM10 after ionomycin treatment and preferentially a substrate of ADAM17 after PMA treatment. However, our data again indicate that ADAM10 might also be involved in PMA-induced shedding of Tim-3.

Shedding of Tim-3 from Toll-like Receptor (TLR)-activated CD14 ϩ Monocytes Was Caused by ADAM10-mediated Shedding-
LPS leads to TLR-mediated activation of CD14 ϩ monocytes, and expression of cytokines including IL-6, TNF␣, and IL-12. Recently, rapid down-regulation of Tim-3 from the cell surface of LPS-activated CD14 ϩ monocytes was linked to up-regulation of IL-6 and IL-12 (17). Blockade of Tim-3 signaling or silencing of Tim-3 led to a significant acceleration of LPS-mediated IL-6 and IL-12 expression (17), indicating that Tim-3 regulates innate immune responses in CD14 ϩ monocytes. Here, we analyzed if down-regulation of Tim-3 on CD14 ϩ monocytes after LPS stimulation was caused by shedding and whether Tim-3 shedding was functionally linked to cytokine  (n ϭ 3). n.s., not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001 (unpaired Student's t tests). NOVEMBER 29, 2013 • VOLUME 288 • NUMBER 48 expression. Primary human CD14 ϩ monocytes were stimulated with LPS and cell-surface expressed Tim-3 was quantified by flow cytometry. Again, CD14 ϩ monocytes expressed high levels of Tim-3, with about 77.4% of the cells expressing Tim-3 on the cell surface. After 2 h of LPS stimulation, only about 35.4% of the monocytes expressed Tim-3 on the cell surface (Fig. 5, E and F). Importantly, Tim-3 down-regulation was prevented by treatment with GI (57.8%), GW (71.6%), and proA10 (66.4%), indicating that ADAM10 and ADAM17 were activated after LPS stimulation to shed Tim-3 from the cell surface in a time-dependent manner (Fig. 5, E-G). Next, we tested if shedding of Tim-3 negatively regulates TNF␣, IL-6, and IL-12 expression. Therefore, CD14 ϩ monocytes were stimulated with LPS for 24 h. Shedding of Tim-3 was prevented by addition of GI and GW. Subsequently, TNF␣, IL-6, and IL-12 were quantified after 2, 4, 6, 12, 18, and 24 h in the cell culture supernatants. As expected, LPS stimulation increases the concentrations of IL-6, TNF␣, and IL-12 (Fig. 6, A-C). If shedding of Tim-3 would negatively regulate IL-6 and IL-12 cytokine expression, than we would expect that inhibition of shedding would lead to a decrease of IL-6 and IL-12 cytokine production. However, abrogation of LPS-induced shedding by GI and GW has no influence on the accumulation of TNF␣ and IL-6 ( Fig. 6,  A-C). GW inhibits the generation of TNF␣, but this effect is caused by blockade of TNF␣ shedding by ADAM17 (Fig. 6A). Interestingly, GI and GW led to an increase of IL-12 after LPS stimulation (Fig. 6C). We have to admit that the up-regulation of IL-12 was not necessarily functionally connected to inhibition of Tim-3 shedding, because the inhibitors generally inhibit ADAM10 and ADAM17. Final proof, however, cannot be achieved with inhibition of ADAM proteases but with a selective inhibition of Tim-3 shedding, as we have shown for Tim-3⌬S2 and Tim-3⌬ICD. Therefore, the generation of Tim-3 uncleavable knock-in or transgenic mice would be desirable to explore the function of Tim-3 shedding during immune responses.

DISCUSSION
Here, we identified Tim-3 as a novel substrate of ADAM10and ADAM17-mediated ectodomain shedding. Shedding of Tim-3 from HEK293 cells was induced after stimulation with ionomycin or PMA. Ionomycin-induced shedding was inhibited by ADAM10-and ADAM10/ADAM17-selective inhibitors GI and GW and the ADAM10 pro-domain. PMA-induced shedding was inhibited by the ADAM10/ADAM17-selective inhibitor GW and to a lesser extent by GI and the ADAM10 pro-domain. These results indicate that PMA-induced shedding is mainly mediated by ADAM17 and in part by ADAM10, whereas ionomycin-induced shedding is solely mediated by ADAM10. These results were verified using ADAM10-and/or ADAM17-deficient murine embryonic fibroblasts. Here, we used a self-made anti-Tim-3 mAb to detect Tim-3 by Western blotting. To verify Tim-3 shedding, we also used the detection of AP-tagged Tim-3, which was described previously (38).
ADAM10 or ADAM17 knock-out mice are embryonically lethal (25,39), which underlines the physiological importance of these two proteases. Conditional knock-out mice have highlighted the importance of ADAM10 and ADAM17 in various physiological conditions and inhibition of one or both proteases is still considered a therapeutic option (40 -42). ADAM10 and ADAM17 have, however, more than 100 substrates (23), some are exclusively shed by ADAM10 or ADAM17 and some are shared substrates. Importantly, no defined consensus cleavage sequences exists for ADAM10 and ADAM17 (32). Moreover, substrate recognition and regulation of selective substrate cleavage by ADAM proteases is largely unknown. However, recently, iRhom2 was reported to facilitate ADAM17 trafficking and maturation. Consequently, iRhom2deficient mice fail to shed TNF (43,44). Moreover, iRhom1 and iRhom2 were shown to be involved in substrate recognition by ADAM17 (45).
For some substrates including TNF␣, EGF, and Notch, ADAM-mediated cleavage has been validated in vivo and the relevance of the cleavage of most substrates is still unknown. Genetic alteration of the ADAM cleavage site of a single substrate seems to be a promising strategy for in vivo analysis of defective shedding of a single substrate. Genetically modified mice encoding an ADAM17 shedding-resistant L-selectin variant showed a drastically reduced serum level of soluble L-selectin and increased cell-surface expression (46,47). Using deletion variants of the stalk domain, we identified an ADAM17 un-cleavable Tim-3 variant (Tim3⌬S2), which after translation into murine Tim-3 enables the generation of an un-cleavable murine Tim-3 knock-in mice to study the in vivo relevance of Tim-3 shedding in mice. Moreover, we speculate that the cleavage site of ADAM17 in Tim-3 is located between Glu 181 and Asp 190 (Tim3⌬S2).
The intracellular domain of Tim-3 also contributes to ADAM17-but not to ADAM10-mediated shedding of Tim-3  NOVEMBER 29, 2013 • VOLUME 288 • NUMBER 48  's t tests). E and F, decreased Tim-3 expression on positively selected CD14 ϩ M/M ⌽ upon TLR stimulation. This effect was restored by specific inhibition of ADAM10 and ADAM17. CD14 ϩ monocytes isolated from healthy subjects were treated for 2 h with LPS (5 g/ml). GI (9 M) or GW (9 M) were added 30 min prior to stimulation. Inhibitory recombinant ADAM10 pro-domain (A10pro, 10 M) was also added 30 min prior to stimulation or without 5 g/ml of LPS for 2 h, followed by immunostaining and flow cytometric analysis for the expression of CD14 and Tim-3. Cells were first gated on the monocytic population and then analyzed for the percentage of Tim-3 positive cells in the CD14 ϩ cell population. F, summary data (mean Ϯ S.D. from four healthy subjects) of percentages of Tim-3 positive cells are shown after 2 h of TLR stimulation. G, kinetics of Tim-3 expression in CD14 ϩ M/M ⌽ following TLR stimulation. CD14 ϩ monocytes from three healthy subjects were incubated with or without TLR ligand LPS for various time points, followed by immunostaining and flow cytometry analysis for the expression of CD14 and Tim-3. The percentage of Tim-3 CD14 ϩ M/M ⌽ at different time points is calculated, and the mean Ϯ S.D. of the double positive cell from these healthy subjects is shown. **, p Ͻ 0.01; ***, p Ͻ 0.001 (unpaired Student's t tests).

Shedding of Tim-3
because deletion of the complete intracellular domain abrogates PMA-but not ionomycin-induced shedding (Tim-3⌬ICD). To identify the region within the intracellular domain that contributes to PMA-induced shedding, we generated and tested additional intracellular deletion variants. Truncation 1 (⌬I1) lacks all but the three tyrosines, whereas truncation 2 (⌬I2) comprises all but one tyrosine, which resides close to the membrane. Tyrosines 256 and 263, which are deleted in Tim-3⌬I2, were shown to be phosphorylated and to enhance NFAT/ AP-1 activation (34). Shedding of Tim-3⌬I1 and Tim-3⌬I2 was, however, indistinguishable from wild-type Tim-3. Also the three deletion variants Tim-3⌬I3, -⌬I4, and -⌬I5, comprising deletions of 9 -12 amino acids from Lys 225 to Ser 260 , were shed like wild-type Tim-3 after PMA stimulation. These results were unexpected and we decided to generate the Tim-3 deletion variant Tim-3⌬ICDϩ1, which covers a deletion of the intracellular domain from Trp 226 to Pro 301 . Surprisingly, the single intracellular lysine residue was sufficient to rescue PMA-induced shedding. At the moment, we have not investigated the mechanism for this phenomenon. To the best of our knowledge, this is the first example that ADAM17-mediated shedding is regulated by the presence or absence of a single intracellular amino acid. For most substrates, involvement of the intracellular domain for ADAM-mediated shedding was not investigated. In the case of the IL-6R, deletion of the intracellular domain did not influence PMA-induced shedding (30). For L-selectin, the intracellular domain participates in regulation of shedding. Here, CaM is constitutively associated with the intracellular membraneproximal region of L-selectin and abrogation of calmodulin binding increases the proteolytic turnover of L-selectin (48). Here, we show that CaM is also interacting with Tim-3 and inhibition of CaM binding to Tim-3 by the inhibitor W7 induces ADAM17-mediated shedding of Tim-3. Shedding also requires anchoring of L-selectin to the cytoskeleton via members of the ezrin-radixin-moesin family, however, this interaction needs more than a single intracellular amino acid (33,35). Lysine 225 was, however, also not mandatory for shedding, because lysine 225 is deleted in Tim-3⌬I1 without affecting PMA-induced shedding. Using flow cytometry, we excluded the possibility that Tim-3⌬ICD is not shed by ADAM17, because of defective transport to the plasma membrane. We therefore conclude from these experiments that the intracellular domain is not needed to facilitate efficient ADAM10 and ADAM17 shedding of Tim-3, with the exception of the deletion of the complete intracellular domain. We hypothesize, that cell-surface localization or integration of Tim-3⌬ICD has changed in comparison to wild-type Tim-3. This might influence accessibility of Tim-3 or the cleavage site for ADAM17, respectively. It has been proposed that inhibition of L-selectin shedding by calmodulin is mediated by such a process, in which calmodulin binds to amino acids of the intracellular and transmembrane domain, thereby changing the conformation of the stalk region and reducing the accessibility for ADAM proteases (37). Whether such a mechanism might also contribute to Tim-3 shedding has to be investigated in future experiments.
Next, we showed that endogenously expressed Tim-3 is also shed by ADAM10 and ADAM17 after PMA or ionomycin stimulation, respectively. Tim-3 is expressed on human peripheral blood CD14 ϩ monocytes (17). Moreover, LPS was recently shown to down-regulate Tim-3 from PBMCs by an unknown mechanism (17). Here, we demonstrate that mainly ADAM10 and to a lesser extend also ADAM17 are responsible for LPSinduced down-regulation of Tim-3, indicating that shedding of Tim-3 might contribute to the immune status of CD14 ϩ monocytes. Our data did, however, not exclude the possibility that other ADAM-or MMP-proteases are involved in Tim-3 shedding. MMP-3 and MT1-MMP were shown to efficiently shed Tim-1 (49,50) and ADAM8 has been shown to release L-selectin from activated neutrophils (51). ADAM8 is also expressed by macrophages (52). Moreover, we also did not analyze the expression of other catalytically active ADAM-proteases such as ADAM9, -12, and -15 (53), which also might contribute to Tim-3 shedding.
Tim-3 was suggested to negatively regulate the expression of TNF␣ and IL-12 after LPS stimulation because blockade of Tim-3 increased cytokine expression. It was speculated that down-regulation of Tim-3 during LPS stimulation positively contributes to cytokine expression of PBMCs. We speculate that inhibition of Tim-3 shedding might interfere with upregulation of cytokine expression. However, inhibition of ADAM10, which is the main protease responsible for LPS-induced shedding of Tim-3 from CD14 ϩ monocytes did not influence the release of TNF␣, IL-12, and IL-6. However, blocking Tim-3 shedding by ADAM10 and ADAM17 even increased the secretion of IL-12 but completely blocks TNF␣ secretion, the latter being caused by the need of ADAM17 to release membrane-bound TNF␣ from the cell surface.
Tim-3 is expressed on many cell types, including activated Th1 cells, APCs, T regs , and NK cells (15). It remains, however, to be seen whether shedding of Tim-3 is also found on these cell types and how it contributes to immune regulation in vivo.