Tissue Inhibitor of Metalloproteinases-3 Inhibits Shedding of L-selectin from Leukocytes*

Although the enzyme or enzymes mediating shedding of L-selectin have not yet been identified, this activity can be blocked by synthetic hydroxamic acid-based inhibitors of metalloproteinases such as Ro 31-9790. However, the endogenous matrix metalloproteinase inhibitor tissue inhibitor of metalloproteinases (TIMP)-1 does not block L-selectin shedding. Here, we report that TIMP-3, but not TIMP-2, inhibits L-selectin shedding from mouse and human lymphocytes, Jurkat T cells, and human monocytes. TIMP-3 has an IC50 of 0.3–0.4 μm on these cell types compared with 0.7–4.8 μm for Ro 31-9790. A metalloproteinase (tumor necrosis factor-α (TNF-α)-converting enzyme; ADAM17) has recently been identified which cleaves the pro-form of TNF-α to produce soluble cytokine. We compared inhibition of L-selectin shedding by TIMPs and Ro 31-9790 with inhibition of TNF-α shedding from human monocytes. TIMP-3 inhibited TNF-α shedding (IC50 of 0.1 μm), as did Ro 31-9790 (IC50 of 0.4 μm). TIMP-2 had a partial effect, and TIMP-1 did not inhibit. This study confirms that L-selectin sheddase is a metalloproteinase, but not a matrix metalloproteinase, and investigates the relationship between shedding of L-selectin and TNF-α.

L-selectin is one of three members of the selectin family and is expressed on leukocytes (for review, see Ref. 1). Through binding to mucin-like carbohydrate ligands on vascular endothelium, L-selectin mediates the tethering and rolling of leukocytes, which precedes their migration into lymphoid organs and sites of inflammation. L-selectin is rapidly down-regulated from the leukocyte surface within minutes by a number of stimuli including chemoattractants (2), phorbol ester (3,4), and cross-linking of cell surface L-selectin (5), or more slowly by cross-linking of the T cell CD3 complex (6). L-selectin is also down-regulated on leukocytes that have migrated across cultured endothelial cells (7,8) or into sites of inflammation (9). The latter observation led to the suggestion that L-selectin down-regulation from the leukocyte surface may be required for transendothelial migration, although it is not necessary for migration of neutrophils across human umbilical vein endothelial cells (10).
Rapid down-regulation of L-selectin has recently been shown to be the result of proteolytic cleavage in the membrane proximal extracellular domain (11). The cleavage site has been mapped in human L-selectin to Lys 283 -Ser 284 . The activity mediating this proteolytic cleavage (L-selectin sheddase) is in-hibited by the hydroxamic acid-based inhibitors of zinc-dependent metalloproteinases Ro 31-9790 (4), KD-IX-73-4 (12), and TAPI (13), but not by inhibitors of other classes of proteases (4). L-selectin sheddase is also not inhibited by tissue inhibitor of metalloproteinases (TIMP)-1 1 (4), an endogenous inhibitor of matrix metalloproteinases (MMPs).
To characterize L-selectin sheddase further, we have tested the ability of two additional members of the TIMP family, TIMP-2 and TIMP-3 (for review, see Ref. 14), to inhibit Lselectin shedding from leukocytes. These inhibitors are structurally homologous to TIMP-1, which we have already shown does not inhibit shedding of L-selectin (4). However, the expression patterns and inhibitory specificities of the TIMPs differ (15,16), suggesting that although all three proteins inhibit the MMPs, they have some functional differences that may be useful in the characterization of L-selectin sheddase.
Proteolytic cleavage is a commonly used mechanism for down-regulation of cell surface proteins (17). Like L-selectin, the shedding of a number of these proteins is inhibited by the hydroxamic acid-based inhibitors at concentrations much greater than those required to inhibit the MMPs (18,19). Recently, a cell surface metalloproteinase has been identified which mediates cleavage of one of these cell surface molecules, the pro-form of tumor necrosis factor (TNF)-␣ (20,21). This enzyme belongs to the ADAM (a disintegrin and metalloproteinase) family of zinc-dependent metalloproteinases. To determine the relationship between these enzymes and L-selectin sheddase, we compared the sensitivities of TNF-␣ shedding and L-selectin shedding from human peripheral blood monocytes to inhibition by TIMPs and synthetic hydroxamic acid-based inhibitors.

MATERIALS AND METHODS
Reagents and Antibodies-Phorbol myristate acetate (PMA; Sigma) was dissolved in dimethyl sulfoxide to a concentration of 1 mM and stored at Ϫ20°C in aliquots. Ro 31-9790 (Roche Products, Herts.) was dissolved in dimethyl sulfoxide to a concentration of 30 mM and stored at Ϫ20°C in aliquots. Lipopolysaccharide (LPS; Sigma) was dissolved in RPMI 1640 to a concentration of 1 mg/ml and stored at Ϫ20°C in aliquots. Recombinant TIMPs were produced as described in (22)(23)(24). Antibodies used are described in Table I.
Cells-Mouse lymphocytes were isolated from pooled axillary and cervical and mesenteric lymph nodes of 6 -8-week-old BALB/c mice bred in the SPF unit at the NIMR. Tissues were collected into calciumand magnesium-free phosphate-buffered saline (PBS-CMF) at 4°C and cell suspensions prepared. Human lymphocytes and monocytes were prepared from buffy coat preparations of peripheral blood obtained from * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed. L-selectin Shedding Assay-Mouse lymphocytes or Jurkat T cells (2.5 ϫ 10 6 in 50 l) were incubated with 100 nM PMA or an equivalent volume of dimethyl sulfoxide for 60 min. Cells were pretreated with inhibitor for 20 min and throughout the assay. Inhibitors and PMA were removed by washing in PBS-CMF containing 0.2% bovine serum albumin, and cell surface L-selectin expression was determined by flow cytometry. This assay was also carried out using PBMC (10 6 in 0.5 ml) within 90 min of isolation from buffy coat.
TNF-␣ Shedding Assay-PBMC (10 6 in 0.5 ml) were incubated in the presence or absence of inhibitor for 20 min followed by the addition of 1 M LPS for 3 h. Inhibitors and LPS were removed by washing in PBS-CMF containing 5% FCS, and cell surface TNF-␣ expression was determined by flow cytometry. The washing and staining procedures for flow cytometric analysis were carried out on ice, using chilled buffer to prevent proteolytic cleavage of cell surface TNF-␣ after washing out of inhibitor.
Flow Cytometry-L-selectin was measured on mouse lymphocytes and Jurkat T cells by direct immunofluorescence using fluorescein isothiocyanate-conjugated MEL-14 and fluorescein isothiocyanate-conjugated FMC46, respectively. L-selectin and cell surface TNF-␣ on human PBMC were measured using two-color indirect immunofluorescence. LAM1.3 was used to detect L-selectin and 101/4 to detect TNF-␣, followed by phycoerythrin-conjugated anti-mouse IgG1; fluorescein isothiocyanate-labeled anti-human CD14 was used to identify monocytes in PBMC. Cells were incubated with appropriate antibodies at each stage of staining for 20 min at 4°C; mouse lymphocytes and Jurkat T cells were washed in PBS-CMF with 0.2% bovine serum albumin (Sigma), and PBMC were washed in PBS-CMF with 5% FCS. Stained cells were analyzed on a Becton Dickinson FACStar Plus or FACS Vantage. Data for 10,000 events were collected for mouse lymphocytes or Jurkat T cells and 30,000 -50,000 events (depending on the proportion of monocytes in the samples) for human mononuclear cells. Data were analyzed using FACSplot software developed by John Green, Computing Lab, NIMR, or using WinMDI (Joseph Trotter, Scripps Institute, CA).
Measurement of Soluble L-selectin by ELISA-These assays were carried out as described in (4). 96-well ELISA plates (Nunc, Maxisorb) were coated with MEL-14 (10 g/ml) overnight at 4°C, blocked for 60 min with Dulbecco's PBS with 1% bovine serum albumin, and washed three times with Dulbecco's PBS with 0.05% Tween 20. Doubling dilutions (from 1/4) of supernatants from mouse lymphocytes (treated with PMA to induce L-selectin shedding in the presence or absence of inhibitor) were added to the plate and incubated for 90 min at room temperature. Captured L-selectin was detected using biotinylated T28.45 (5 g/ml) and streptavidin-horseradish peroxidase conjugate (DAKO). The reaction was stopped with 2 M H 2 SO 4 , and absorbance was read at 450 nm. A standard curve was constructed using a mouse L-selectin-Ig fusion protein (LEC-Ig, generously provided by Dr. Susan Watson, Genentech) over the range of 0.75-100 ng/ml; absorbance values for an irrelevant fusion protein (mouse CTLA4-Ig; generously provided by Dr. Peter Lane, Basel Institute for Immunology, Basel) were subtracted from the standards and test samples.
where x ϭ value for sample under test, y ϭ value for sample set as 0%, z ϭ value for sample set at 100%. Results are presented as percent inhibition Ϯ S.E. for at least three replicates for each point.
Calculation of IC 50 Values-The IC 50 value for inhibitor in each system was defined as the concentration of inhibitor which gave 50% inhibition, where 100% inhibition was set as the percentage of cells positive for L-selectin or cell surface TNF-␣ in the presence of PMA ϩ 50 M Ro 31-9790 or LPS ϩ 50 M Ro 31-9790, respectively (doses of Ro 31-9790 which gave maximal inhibition) and 0% inhibition was set as the percentage of cells positive for L-selectin or TNF-␣ in the presence of PMA or LPS alone, respectively.

Inhibition of Mouse Lymphocyte L-selectin Shedding by TIMP-3-
We have shown previously that PMA-induced shedding of L-selectin from mouse lymphocytes is inhibited by the synthetic metalloproteinase inhibitor Ro 31-9790 but not by the endogenous MMP inhibitor TIMP-1 (4). These studies have been extended by including TIMP-2 and TIMP-3 in the same in vitro L-selectin shedding assay. Lymphocytes isolated from mouse lymph nodes were treated with PMA in the presence or absence of varying concentrations of each of the inhibitors, and L-selectin remaining on the cell surface was detected by flow cytometry. As reported previously, TIMP-1 had no effect on PMA-induced shedding of cell surface L-selectin at concentrations up to 1.1 M (30 g/ml) (Fig. 1, A and B), whereas the hydroxamic acid-based inhibitor Ro 31-9790 completely inhibited PMA-induced shedding at a concentration of 10 M (Fig.  1B). TIMP-2 was also unable to inhibit L-selectin shedding (Fig. 1, A and B), but TIMP-3 inhibited PMA-induced L-selectin shedding in a dose-dependent manner with complete inhibition at 0.8 M (20 g/ml; Fig. 1B). To confirm that PMA was inducing shedding of L-selectin, rather than internalization, soluble L-selectin was quantitated by ELISA in supernatants from untreated, PMA-treated, and PMA ϩ inhibitor-treated mouse lymphocytes. Although there was a background level of soluble L-selectin present in cell supernatants in the absence of PMA, an increase of approximately 100% was seen in response to phorbol ester (Fig. 1C); this PMA-induced increase was inhibited completely by TIMP-3 and Ro 31-9790 but not by TIMP-1 or TIMP-2. This background level of L-selectin shedding, detectable as soluble L-selectin in the supernatants of untreated mouse lymphocytes, explains the Ͼ100% inhibition of PMA-induced shedding by TIMP-3 and Ro 31-9790 (Fig. 1B); these inhibitors prevent both the PMA-induced and the background shedding, allowing cell surface L-selectin to increase to a level greater than that seen on untreated lymphocytes. However, even in the presence of TIMP-3 or Ro 31-9790 some residual soluble L-selectin can be detected in cell supernatants by ELISA. TIMP-3 Inhibits L-selectin Shedding from Human Lymphocytes-The TIMPs used in the experiments described above were recombinant human TIMPs. It was possible that the absence of inhibition of L-selectin shedding by TIMP-1 and TIMP-2 was caused by the inability of the human TIMPs to function in a mouse cell system. Therefore, the ability of the TIMPs to inhibit shedding of L-selectin from human cells was investigated. Jurkat T cells were treated with PMA in the presence and absence of inhibitors, and cell surface L-selectin was detected using direct immunofluorescence. As found for mouse lymphocytes, TIMP-3 and Ro 31-9790 inhibited PMA-induced shedding of L-selectin, whereas TIMP-1 and TIMP-2 were without effect ( Fig. 2A). Similarly, TIMP-3 and Ro 31-9790 (but not TIMP-1 or TIMP-2) inhibited L-selectin shedding from human peripheral blood lymphocytes (Fig. 2B). Therefore, similar results were obtained with each of the TIMPs using both mouse and human cells.
TIMP-3 Inhibits Both L-selectin Shedding and TNF-␣ Shedding on Human Monocytes-The cytokine TNF-␣ is a well characterized example of a cell surface molecule that is shed by proteolytic cleavage. This molecule is produced as a 26-kDa cell surface protein and is processed by zinc-dependent metalloproteinases to generate the 17-kDa soluble cytokine. We compared the abilities of TIMP-1, TIMP-2, TIMP-3, and Ro 31-9790 to inhibit shedding of L-selectin and TNF-␣ from human monocytes in order to investigate the relationship between these enzyme activities on the same cell. Human peripheral blood monocytes express L-selectin, therefore PMA was used to induce shedding, and cell surface L-selectin was measured as described for lymphocytes. Cell surface TNF-␣ is not normally detectable on monocytes. To measure the effect of metalloproteinase inhibitors on TNF-␣ shedding, monocytes were stimulated with LPS, to up-regulate TNF-␣ synthesis, and the level of TNF-␣ which accumulated at the cell surface, and which was released into the cell supernatant, in the presence and absence of inhibitors was determined. As found for lymphocytes, TIMP-3 and Ro 31-9790 (0.83 and 30 M, respectively) blocked PMA-induced L-selectin shedding from human monocytes ( Fig.  3 and Table II). TIMP-3 and Ro 31-9790 also mediated accumulation of TNF-␣ on the surface of human monocytes (Fig. 3 and Table II), indicating a block in the TNF-␣ shedding pathway. This accumulation of TNF-␣ on the cell surface was mirrored by decreased soluble TNF-␣ in supernatants of cells treated with LPS and TIMP-3 (data not shown). In contrast, TIMP-1 and TIMP-2 had no effect on shedding of L-selectin from monocytes (Table II). TIMP-1 was also without effect on cell surface TNF-␣ on monocytes, but inclusion of TIMP-2 at  1.44 M resulted in an increase in the percentage of cells positive for TNF-␣, from 39.3% to 55.7% (31% inhibition), although this was less than that obtained with TIMP-3 (Table II).

DISCUSSION
The enzyme or enzymes mediating L-selectin shedding from leukocytes have yet to be identified. However, using a new category of MMP inhibitors based on hydroxamic acid, L-selectin shedding has been shown to be blocked in a variety of cell types in response to both physiological and nonphysiological stimuli (4,10,12,13), suggesting that L-selectin shedding in these different circumstances is mediated by similar metalloproteinase dependent mechanisms. In this report, we extend our previous work and demonstrate that the endogenous metalloproteinase inhibitor TIMP-3 inhibits L-selectin shedding from mouse and human lymphocytes and human monocytes. We have shown previously that TIMP-1 did not inhibit Lselectin shedding (4), and we demonstrate here that the closely related inhibitor TIMP-2 is also ineffective.
Some MMPs, and collagenase in particular, mediate limited L-selectin proteolysis (4), suggesting that the L-selectin sheddase activity may be an MMP or an MMP-related enzyme. However, the work reported here and a previous study (4) demonstrate that phorbol ester-induced L-selectin shedding is not inhibited by two members of the TIMP family, TIMP-1 and TIMP-2. Between them, these two proteins inhibit the catalytic activity of all known MMPs, and the inability of TIMP-1 and TIMP-2 to prevent shedding of L-selectin therefore suggests that L-selectin sheddase is not a known MMP. This is supported by the evidence that L-selectin sheddase is a cell-associated, rather than a soluble, activity (4,25), as the MMPs are secreted enzymes. Recently, a subgroup of transmembrane MMPs (membrane-type MMPs or MT-MMPs) has been identified (26). Like L-selectin sheddase, MT1-MMP is not inhibited by TIMP-1 (27). However, this cell surface enzyme can be excluded as a potential candidate for the L-selectin sheddase because it is inhibited by both TIMP-2 and TIMP-3 (27), and the data presented here show that L-selectin shedding is not inhibited by TIMP-2.
The hydroxamic acid-based inhibitors were originally designed as inhibitors of MMPs, and Ro 31-9790 inhibits recombinant human fibroblast collagenase (MMP1) with an IC 50 of 4.9 nM. As described here and elsewhere, hydroxamic acidbased inhibitors prevent phorbol ester-induced shedding of cell surface L-selectin with IC 50 values in the low micromolar range (4,12,13) i.e. they are about 1,000-fold less potent in inhibiting L-selectin shedding in cell-based assays than in inhibiting recombinant MMPs in cell-free assays. When considered together with the inability of TIMP-1 and TIMP-2 to inhibit L-selectin shedding from all of the cell types tested, it is unlikely that the L-selectin-shedding enzyme(s) is an MMP. Proteolytic cleavage of a number of other cell surface molecules, such as angiotensin-converting enzyme (ACE) (28), amyloid precursor protein (29), transforming growth factor-␣ (18), and the pro-form of TNF-␣ ( (20,21) and this study), is also inhibited by synthetic hydroxamate inhibitors at micromolar concentrations, suggesting that the activities mediating shedding of these molecules and L-selectin shedding may be related.
An enzyme has recently been cloned, called TNF-␣ converting enzyme (TACE; ADAM17) which correctly cleaves both a pro-TNF-␣ cleavage site peptide and full-length pro-TNF-␣ (20,21). In addition, leukocytes from TACE-deficient mice have a much reduced ability to secrete soluble TNF-␣, suggesting an important role for TACE in mediating TNF-␣ shedding in vivo. TACE is a member of a recently characterized family of zincdependent metalloproteinases known as the ADAM family (for review, see Ref. 30). An association between TACE and Lselectin shedding has recently been identified: thymocytes from TACE-deficient mice do not shed L-selectin in response to PMA stimulation. 2 In addition, high concentrations of recombinant catalytic domain of TACE have been shown to cleave a peptide representing the cleavage site of L-selectin. 2 This suggests that TACE may be a physiological L-selectin sheddase. A second member of the ADAM family (ADAM10) also cleaves a TNF-␣ cleavage site peptide but has yet to be shown to cleave fulllength pro-TNF-␣ correctly (31,32). We therefore investigated the possible relationship between enzymes mediating shedding of L-selectin and TNF-␣ by comparing the inhibitory profiles of the TIMPs and Ro 31-9790 on L-selectin and TNF-␣ shedding from human monocytes. We show here that TIMP-3 mediates accumulation of TNF-␣ on the surface of human monocytes in vitro. TIMP-1 has no effect, and TIMP-2 is less effective than TIMP-3. Solomon et al. (33) have shown that although the level of cell surface TNF-␣ on human peripheral blood monocytes may not quantitatively reflect alterations in TNF-␣ shedding (because of degradation of unprocessed pro-TNF-␣), this measurement can be used as a marker of inhibition of shedding at the time point utilized in our TNF-␣-shedding assays (3 h). Measurement of soluble TNF-␣ in supernatants from monocytes stimulated with LPS in the presence or absence of TIMP-3 gave an IC 50 very similar to that obtained from cell surface expression. This indicates that both types of measurement can be used in our assay. Therefore differences can be found in the inhibitory profiles of the TIMPs and Ro 31-9790 on shedding of L-selectin and TNF-␣ from human monocytes and of Ro 31-9790 on shedding of L-selectin from different cell types, suggesting the potential involvement of more than one enzyme in these proteolytic events. However, it is also possible that the differences obtained result from differential accessibility of enzymes(s) to the inhibitors, particularly since shedding of the two proteins could not be investigated under identical experimental conditions. Interestingly, TIMP-3 has recently been reported to inhibit processing of a peptide encoding the cleavage site of pro-TNF-␣ by recombinant catalytic domain of TACE (34). As in our cellbased assays, TIMP-1 had no effect on peptide proteolysis by TACE, and TIMP-2 was less effective than TIMP-3. Shedding of cell surface interleukin 6 receptor, which is sensitive to the hydroxamic acid-based inhibitors, has also recently been found to be blocked by TIMP-3 (35). The effect of TIMPs on the activity of other non-MMP zinc-dependent metalloproteinases is not known. Although the TIMPs are structurally homologous and can all inhibit the MMPs (except MT1-MMP) equally well, they exhibit different expression patterns, with TIMP-3 alone being restricted to the extracellular matrix (16). The ability of TIMP-3 to inhibit L-selectin shedding and TNF-␣ processing suggests either that TIMP-3 is functionally less specific than TIMP-1 and TIMP-2 or that it can more easily gain access to the L-selectin and TNF-␣ shedding enzymes and so inhibit their activity. TIMP-3 is the only TIMP that binds heparan sulfate proteoglycans, 3 which are expressed at the cell surface, and binding to these glycosaminoglycans may localize TIMP-3 and enable interaction with cell surface metalloproteinases. There may be additional motifs within TIMP-3 which contribute to the ability of this protein, but not TIMP-1 or TIMP-2, to inhibit L-selectin shedding.
Recent reports have shown that overexpression of TIMP-3 in some cell types results in increased apoptosis of the cells (36,37). These studies used non-hemopoietic cells and time courses of several days, with little apoptosis occurring before 12-24 h of cell culture, whereas the assays reported here used much shorter incubations with TIMP-3 (1-3 h). However, to exclude the possibility that leukocytes are more sensitive to TIMP-3induced apoptosis, we used an annexin V binding apoptosis assay to compare levels of apoptosis in TIMP-3-treated and control cell samples. No differences were found using either Jurkat T cells or human peripheral blood monocytes. 4 Thus, TIMP-3 does not appear to induce apoptosis of these cells within the times used in our assays.
Shedding of many different cell surface proteins, in addition to L-selectin and TNF-␣, is inhibited by hydroxamic acid-based inhibitors (18,19). The close relationship between the shedding mechanisms involved in proteolytic cleavage of these molecules has been revealed through the use of a mutant Chinese hamster ovary cell line, defective in the shedding of transfected L-selectin, transforming growth factor-␣, interleukin-6 receptor, ␤-amyloid precursor protein, and a number of unidentified endogenous proteins present in the supernatants of parental, but not mutant cells (18). There is evidence that this cell line differs from the parental line by alteration in only a single gene, suggesting that a common mechanism underlies shedding of many cell surface proteins, possibly a "pan-sheddase." In support of this theory, TACE-deficient cells are also defective in the shedding of a number of other cell surface molecules, including L-selectin. 2 Further studies identifying the sheddases for other metalloproteinase-cleaved cell surface molecules, including L-selectin sheddase, should clarify the relationship between the sheddases. The results presented here address this relationship and show differences in the susceptibility of L-selectin and TNF-␣ shedding to inhibition by the TIMPs and also in the IC 50 values obtained with TIMP-3 and Ro 31-9790. This raises the possibility that shedding of L-selectin and TNF-␣ from human monocytes may be mediated by different enzymes, although other explanations exist.