Natural, Proteolytic Release of a Soluble Form of Human IL-15 Receptor α-Chain That Behaves as a Specific, High Affinity IL-15 Antagonist¹

Many soluble cytokine receptors have been described that correspond to the extracellular portions of their full-length transmembrane receptor counterparts. These soluble receptors are found in cell culture supernatants and in body fluids of humans and mice. They have been shown to keep their ligand binding capacities and therefore to be potential regulators of the biological functions of cytokines. They can either function as natural antagonists (e.g., IL-2Rα, IL-4R, LIF-R), biological agonists (e.g., IL-6R, CNTF-R), carrier molecules (e.g., IL-4R), or chaperones to protect their ligands from proteolytic degradation (1). Two major mechanisms responsible for the generation of these soluble receptors have been identified. The first is related to alternative gene splicing leading to the synthesis of mRNAs that encode soluble, truncated receptors lacking the intracellular and transmembrane domains. The second is the proteolytic cleavage of the extracellular domain of membrane-anchored receptors, mostly by metalloproteinases (2, 3).

IL-15 and IL-2 are two structurally related cytokines that belong to the four α helix bundle family (4). They elicit similar biological effects in vitro, a redundancy that is explained by the common usage within their functional high affinity receptors of the IL-2Rβγ signaling complex (5). Cytokine specificity is conferred by additional private chains, IL-2R α-chain and IL-15R α-chain (IL-15Rα),³ which are structurally related (6). These two chains contain structural domains (called sushi domains) previously found in some complement and adhesion molecules (7). IL-2R α-chain contains two such domains, whereas IL-15Rα contains only one. One noticeable difference is that IL-2 binds to IL-2R α-chain with an affinity (K_d = 10 nM) far lower than IL-15 to IL-15Rα (K_d = 0.05 nM). Due to the sharing of the βγ complex, both cytokines trigger similar downstream signaling pathways including activation of Jak-1/Jak-3 tyrosine kinases and subsequent nuclear translocation of the phosphorylated Stat-3 and Stat-5, activation of Lck and Syk tyrosine kinases, activation of the MAPK pathway, and induction of Bcl-2 (8, 9).

Soluble forms of the IL-2R components have been described. The soluble IL-2R α-chain is naturally released by proteolytic shedding from activated T cells and this shedding may represent a mechanism that participate to the down-regulation of lymphocyte activation, by decreasing the level of cell surface IL-2R α-chain and subsequent lymphocyte sensitivity to the action of IL-2, and by acting as a soluble antagonist of IL-2 action (10). The soluble IL-2R α-chain is however, like its membrane counterpart, a low affinity receptor for IL-2 and high concentrations are required to exert a neutralizing effect in vitro (11). Proteolytic cleavage of IL-2R α-chain is also a mechanism used to bias the immune response. For example, it has been shown that the cysteine protease Der p1, a major mite allergen, can cleave IL-2R α-chain from the surface of human peripheral blood T cells, thereby inhibiting the propagation of Th1 cells and favoring the development of a Th2 associated allergic environment (12). Another example is the recent demonstration that tumor-derived metalloproteinases can induce the proteolytic cleavage of IL-2R α-chain on activated T cells and suppress the proliferative capacity of cancer-encountered T cells (13). Soluble forms of the IL-2R β-chain and IL-2R γ-chain have also been described in vivo (14).

No report concerning a naturally produced soluble form of IL-15R α-chain (sIL-15Rα) has been published so far. In this study, we demonstrate for the first time that the IL-15Rα can be shedded by a metalloproteinase-dependent proteolytic mechanism, that this sIL-15Rα retains high affinity for IL-15 binding and is able to inhibit IL-15 biological activity at very low levels. It might therefore act as an important regulator of IL-15 responses.

Human IL-1β, murine IL-3, and human GM-CSF were purchased from R&D Systems (Abington, U.K.). Human rIL-15 and human TNF-α were purchased from PeproTech (Rocky Hill, NJ) and human rIL-2 from Chiron (Emeryville, CA). Polyclonal goat anti-human IL-15Rα Ab AF247 was purchased from R&D Systems and the monoclonal mouse anti-human IL-15Rα Abs M161 and M162 (IgG1 isotype) were kindly provided by GenMab (Copenhagen, Denmark). A control isotype IgG1 mouse mAb was purchased from BD Bioscience (Le Pont de Claix, France). PE-conjugated goat anti-mouse F(ab′)₂ were purchased from Beckman Coulter (Marseille, France). The proline endopeptidase inhibitor, S17092-1 (15), was kindly provided by Dr. F. Checler (Centre National de la Recherche Scientifique, Unité Propre de Recherche 411, Valbonne, France). The hydroxamic acid inhibitor of matrix metalloproteinases (MMPs), GM6001, and its negative control (GM6001-neg), were purchased from Calbiochem (Nottingham, U.K.). PMA and ionomycin were purchased from Sigma-Aldrich (Saint Quentin Fallavier, France).

The nonadherent U-937 human histiocytic sarcoma cell line from the American Type Culture Collection (ATCC, Manassas, VA; CRL-1593.2) and the adherent COS-7 monkey cell line (ATCC CRL-1651) were purchased. The nonadherent Kit 225 human T lymphoma cell line (16) was obtained from Dr. D. Cantrell (University of Dundee, Scotland, U.K.). The adherent A172 human gliosblastoma cell line was kindly provided by Dr. H. Gascan (Institut National de la Santé et de la Recherche Médicale Unitè 564, Angers, France). The nonadherent TF-1 human erythroleukemia human cell line (17) and transfected TF1-β cells (18) were kindly provided by Dr. B. Azzarone (Institut Gustave-Roussy, Villejuif, France). 32DβR cells (J. Bernard, unpublished observations) were obtained by electroporating 32Dβ cells (19) (kindly given to us by Dr. W. Leonard, National Institutes of Health, Bethesda, MD) to express human IL-15Rα.

All cells were grown in 5% CO₂ at 37°C in a water-saturated atmosphere. TF-1 were cultured in RPMI 1640 medium containing 10% heat-inactivated FCS, 2 mM glutamine, and 1 ng/ml GM-CSF. TF-1β cells were cultured in the same medium supplemented with 250 μg/ml geneticin and 32DβR cells in the same medium containing 0.4 ng/ml IL-3, 10 μM 2-ME, 250 μg/ml geneticin, and 0.5 μg/ml puromycin. U-937 and A172 cells were cultured in RPMI 1640 medium containing 10% FCS and 2 mM glutamine; Kit 225 cells were cultured in RPMI 1640 medium containing 6% FCS, 2 mM glutamine, and 10 ng/ml human rIL-2; and COS-7 cells were cultured in DMEM containing 10% FCS, 2 mM glutamine, and 1 mg/ml glucose.

Plasmids pcDNA-15Rmh, pcDNA-Δ215Rmh, and pcDNA-Δ315Rmh corresponding to full-length IL-15Rα, and IL-15Rα deleted of exon 2 or exon 3, respectively, have been previously described (20). The plasmid corresponding to the extracellular form of IL-15Rα (pcDNA-s15Rmh) was generated by PCR using the antisense primer 5′-ATCTGAGAGTGGTGTCGCTGTGGCCCT-3′ (nested Xho restriction site underlined) and the sense primer 5′-ATCTGAGAGTCCAGCGGTGTCCTGTGG-3′. After amplification, the sequences were ligated at the Xho site of the pcDNA3.1/myc-His mammalian expression vector (Invitrogen, San Diego, CA). For generation of the two IL-15Rα variants deleted of exon 4 (pcDNA-Δ415Rmh) and exon 5 (pcDNA-Δ515Rmh), the 5′ and 3′ sequences framing each exon were amplified separately and ligated in pcDNA3.1/myc-His mammalian expression vector (Invitrogen). For exon 4 deletion, the antisense primer 5′-CGGGATCCTTTTCCAGAAGGGGAGAGGC-3′ (nested BamHI restriction site underlined) and the sense primer 5′-CGGGATCCGGTGTGTATCCACAGGGC-3′ were used. For exon 5 deletion, the antisense primer 5′-CGGGATCCTGGCGGCTGGTGGGAGG-3′ and the sense primer 5′-CGGGATCCGCTATCTCCACGTCCACTG-3′ were used. In the final deleted genes, the BamHI site (coding for the dipeptide Gly-Ser) behaved as a linker between the 5′ and the 3′ sequences. For transfection, 10⁶ COS-7 cells were cultured overnight and adherent cells were transfected with 20 μg of plasmid per 10 cm plate following a standard calcium phosphate protocol. After 6 h, the medium was replaced with fresh medium and supernatants were harvested 48 h later.

Adherent A172 cells and transfected COS-7 cells were brought in suspension with PBS plus 200 μg/ml EDTA, washed three times, seeded in six-well plates (2 × 10⁶ cells/well), and cultured with or without 100 ng/ml PMA for 1 h. Cells were suspended again with PBS plus 200 μg/ml EDTA, washed three times with PBS-BSA 0.1%, incubated 20 min in the dark at 4°C with 10 μg/ml anti-IL-15Rα mAb (M161 or M162) or isotype-matched control mAb. They were then washed twice, incubated with 5 μg/ml goat anti-mouse IgG1-PE, and analyzed on a FACScan fluorocytometer (BD Bioscience). Data were acquired and analyzed with the use of CellQuest software.

Culture supernatants from COS-IL-15Rα, COS-sIL-15Rα, A172, TF-1, TF-1β, or Kit 225 cells (200 ml; 10⁶ cells/ml) were collected and filtrated (0.22 μm filter). Soluble IL-15Rα from each supernatant was then purified on an affinity column prepared by grafting the anti-IL-15Rα M161 (1 mg) onto Affi-Gel 10 (1 ml) (Bio-Rad, Marnes la Coquette, France) following the protocol of the manufacturer. The column was washed exhaustively with 10 mM Tris, pH 7.5 and eluted with 200 mM glycine/HCl, pH 3. Fractions collected were neutralized using 1 M Tris. The concentration of sIL-15Rα in each fraction was determined using the radioimmunoassay (RIA) described below. Active fractions were pooled, dialyzed against PBS, and sterilized by filtration (0.22 μm filter).

For quantification of sIL-15Rα, a sandwich RIA was set up in which the polyclonal goat anti-human IL-15Rα Ab AF247 was used as capture Ab and radio-iodinated monoclonal anti-human IL-15Rα Ab M161 used as tracer. A purified recombinant sIL-15Rα-IL-2 fusion protein (21) was used as standard. AF247 was coated (5 μg/ml; 50 μl/well) to high-adsorption wells (breakable strips). Wells were saturated with PBS-BSA 0.5% for 15 min and sIL-15Rα containing samples (50 μl/well) were incubated for 1 h at 4°C. The M161 mAb iodinated using the iodogen method (22) was then added (1 nM, 50 μl/well) for 1 h at 4°C. Supernatants of each well were collected and the wells washed twice with PBS. The radioactivity associated to the wells (bound M161 fraction) and contained in the supernatants and washings (unbound M161 fraction) were determined.

Human rIL-15 was radiolabeled with ¹²⁵I-labeled iodine (specific radioactivity of around 2000 cpm/fmol) using a chloramine T method (23). To measure IL-15 binding to sIL-15Rα, polyclonal goat anti-human IL-15Rα AF247 (5 μg/ml in PBS; 50 μl/well) was coated on high adsorption wells (breakable strip) overnight at 4°C. Wells were saturated with PBS-BSA 0.5% for 15 min and sIL-15Rα was incubated for 1 h at 4°C. Increasing concentrations of iodinated human rIL-15 were then incubated for 1 h at 4°C. Supernatants of each well were harvested and radioactivity of wells and supernatants were determined. The nonspecific binding was determined in the presence of a 100-fold excess of unlabeled human rIL-15 and subtracted from total binding to generate the specific binding component. Regression analysis of the binding data was accomplished using a one-site equilibrium binding equation (Graphit; Erithacus Software, Staines, U.K.) and data plotted in the Scatchard coordinate system.

To measure inhibition of human rIL-15 binding, TF-1 cells were incubated with a fixed concentration of iodinated human rIL-15 and increasing concentrations of sIL-15Rα. Cell bound and unbound fractions were determined, the nonspecific binding component subtracted, and regression analysis of the data performed using a 4-parameter logistic equation (Graphit; Erithacus Software).

Affinity cross-linking of radio-iodinated human rIL-15 to cell surface IL-15Rα (TF-1 cells) or sIL-15Rα purified from supernatants of various cells was conducted using the homobifunctional cross-linker ethylene-glycol-bis-succinimidylsuccinate (EGS; Sigma-Aldrich) following the protocol of the manufacturer. The cross-link complexes were separated on a 10% SDS polyacrylamide gel, and visualized by autoradiography (PhosphoImager 445 SI; Molecular Dynamics, Sunnyvale, CA).

The proliferative responses of Kit 225 and 32DβR cells to IL-15 and the inhibitory activity of sIL-15Rα on these responses were measured by [³H]thymidine incorporation. Cells were maintained in culture medium for 3 days, washed twice, and starved for 2 h in medium without cytokine. They were then seeded in multiwell plates at 10⁴ cells/well in 100 μl and cultured for 48 h in medium supplemented with a fixed concentration of human rIL-15 or human rIL-2 and increasing concentrations of sIL-15Rα, or in medium supplemented with a fixed concentration of sIL-15Rα and increasing concentrations of human rIL-15. Cells were pulsed for 16 h with 0.5 μCi/well of [³H]thymidine, harvested onto glass fiber filters, and cell-associated radioactivity was measured.

The phorbol ester PMA has been shown to induce the shedding of different cytokine receptors like the IL-6R (24). As a first mean to analyze whether IL-15Rα could be similarly regulated, COS-7 cells transfected with a cDNA encoding full-length human IL-15Rα (COS-IL-15Rα) were submitted to PMA treatment and their IL-15Rα cell surface expression analyzed by flow cytometry using the anti-human IL-15Rα mAb M161. As shown in Fig. 1, PMA treatment induced a reduction of this labeling. COS-7 cells transfected with empty vector, treated or not with PMA, were negative. A172 is a human glioblastoma cell line that expresses membrane IL-15Rα constitutively. PMA treatment of A172 also resulted in a decrease of its cell surface labeling with M161 (Fig. 1). This effect was time-dependent, with a maximal effect achieved at about 1 h PMA treatment (data not shown).

We next analyzed whether this decrease in IL-15Rα cell surface expression was accompanied by the appearance in the culture supernatants of a sIL-15Rα (Fig. 2). For that purpose, a “sandwich” RIA was set up that used the polyclonal anti-IL15Rα Ab AF247 as capture, and radio-iodinated M161 mAb as tracer. When calibrated with a purified recombinant form of human IL-15Rα (IL-15Rα-IL-2 fusion protein), this RIA gave a linear response up to an IL-15Rα-IL-2 concentration of 500 pM and with a sensitivity of ∼1 pM (Fig. 2,A). In contrast to supernatants of COS-7 transfected with empty vector that were always negative in the RIA (data not shown), supernatants of COS-IL-15Rα cells became reactive in the RIA with time of culture after transfection (Fig. 2,B), and treatment with PMA increased the kinetic and level of reactivity. The calcium ionophore ionomycin also increased the release of sIL-15Rα by transfected COS-7 cells. The effects of both PMA and ionomycin were concentration dependent (Fig. 2,C), with half-maximal effects observed at 70 ng/ml and 100 ng/ml, respectively. In addition to PMA and ionomycin, IL-1β and TNF-α were also found to dose dependently increase the level of sIL-15Rα in A172 cells supernatants (Fig. 2 D) and transfected COS-7 cells supernatants (data not shown).

Crude supernatants collected from unstimulated A172 cells contained sIL-15Rα levels (1 pM) that were at the limit of detection by the RIA, and lower than those measured (35 pM) in supernatants from unstimulated transfected COS-7 cells. This could be linked to the fact that A172 cells express lower cell surface IL-15Rα than do transfected COS-7 cells (Fig. 1). To increase the sensitivity of detection, cell supernatants (200 ml) were purified and concentrated by immunoaffinity on a column coupled with the M161 mAb. No RIA positivity was found in the fractions eluted from the supernatant of the U-937 cell line that is IL-15Rα negative by flow cytometry with the M161 mAb, as shown in Table I (25, 26, 27). By contrast, supernatants from the four IL-15Rα positive cell lines A172, TF-1, Kit 225, and TF-1β, all contained RIA-positive material (Table I). Thus, sIL-15Rα release was correlated with IL-15Rα cell surface expression, whether or not the cell lines examined expressed IL-2Rβ and/or IL-2Rγ.

To determine the molecular mass of the sIL-15Rα shed from the different cell lines, the immunopurified fractions were incubated with iodinated IL-15 and the mixture subjected to chemical cross-linking (Fig. 3). No cross-linking was seen with supernatants from COS-7 cells transfected with empty vector (data not shown). With material purified from COS-IL-15Rα cells, a broad cross-linked band at ∼55 kDa was observed that corresponded to the cross-linking of one molecule of IL-15 (13 kDa) with a receptor molecule of 42 kDa. A cross-linked band at about the same mass was obtained with supernatants of COS-7 cells transfected with a cDNA encoding the extracellular domain of IL-15Rα (COS-sIL-15Rα), suggesting that the shedding from IL-15Rα transfected COS-7 cells occurred via cleavage of the membrane receptor at a site proximal to the cell membrane. IL-15 cross-linked bands of similar masses (∼55 kDa) were also found with supernatants of the two human cell lines TF-1 and A172. Control IL-15 cross-linking to cell surface IL-15Rα expressed by TF-1 cells showed a band at 68 kDa, corresponding to the cross-linking of one molecule of IL-15 with a receptor of 55 kDa. Thus, the shedding of IL-15Rα on this cell line resulted in the removal of 13 kDa (from 55 to 42 kDa).

To further locate the site of cleavage, IL-15Rα cDNA lacking exons encoding different parts of the extracellular domain were transfected in COS-7 cells and the supernatants analyzed in the IL-15Rα-specific RIA (Fig. 4). No signal was obtained with COS-7 cells transfected with IL-15RαΔ2 (COS-IL-15RαΔ2). This result was expected because one of the mAb in the RIA (M161) reacts with an epitope belonging to the domain (so called sushi domain) encoded by exon 2 (J. Bernard, unpublished observations). RIA-positive sIL-15Rα material was observed in supernatants of COS-7 cells transfected with IL-15Rα lacking either the exon 3, exon 4, or exon 5 encoded domain. In each case, the release of sIL-15Rα was increased by PMA (Fig. 4). By IL-15 cross-linking, it was observed in each case that the sIL-15Rα had a molecular mass lower than the one released from COS-7 cells transfected with wild-type IL-15Rα, in agreement with the deletion of part of the extracellular domain. This is illustrated in Fig. 3 for the soluble receptor immunopurified from the supernatants of IL-15RαΔ3 transfected COS-7 cells (COS-IL-15RαΔ3); a cross-linked band is observed with a mass (46 kDa) ∼7 kDa lower than the one derived from COS-7 cells transfected with wild-type IL-15Rα.

Several protease inhibitors were used to define the molecular mechanisms and enzyme(s) responsible for the cleavage of membrane IL-15R. Common cystein-, serine- and acidic-proteases inhibitors, aprotinin (0.1 μM), leupeptin (1 μM), EDTA (1 mM), and PMSF (1 mM), were unable to inhibit the IL-15Rα shedding by COS-IL-15Rα cells stimulated or not with PMA (data not shown). The proline endopeptidase inhibitor, S17092-1 (1 μM) (15), was also inactive. By contrast (Fig. 5), the GM6001 (4 μg/ml) (28), a potent broad-spectrum hydroxamic acid inhibitor of MMPs, completely prevented the stimulatory effect of PMA on membrane-bound IL-15Rα shedding. GM6001-neg, a control synthetic compound that is not inhibitory of MMPs, was without effect. However, the basal IL-15Rα shedding observed in the absence of PMA was not significantly affected by GM6001 (Fig. 5), even when used at 3-fold higher concentrations (12 μg/ml) (data not shown).

The affinity for IL-15 of the soluble IL-15Rα prepared from full-length IL-15Rα transfected COS-7 cells was evaluated in a sandwich assay using AF247 to capture the receptor and increasing concentrations of iodinated human rIL-15 as tracer. A saturation-binding curve was obtained (Fig. 6 A), which after Scatchard transformation, allowed the estimate of the affinity of IL-15 for the soluble receptor. An equilibrium dissociation constant (K_D = 166 pM) was measured, a value that is comparable to the one describing high affinity binding of IL-15 to cell surface IL-15Rα on TF-1 cells (K_D = 79 pM, data not shown).

We next analyzed whether the soluble receptor could compete with IL-15 binding to cell surface receptors. TF-1 cells express cell surface IL-15Rα and IL-2R γ-chain but not IL-2R β-chain, and IL-15 binding reflects binding to the isolated IL-15Rα. As shown in Fig. 6 B, soluble IL-15Rα material immunopurified from different cell supernatants were able to efficiently inhibit IL-15 binding to TF-1 cells. Based on the concentrations of sIL-15Rα present in these immunopurified fractions (as determined by the RIA), the IC₅₀ reflecting half-maximal inhibitory effects were calculated. Soluble IL-15Rα from COS-IL-15Rα and A172 displayed very similar, high affinity, IC₅₀ (20 and 22 pM, respectively). A sIL-15Rα purified from the supernatants of COS-7 cells transfected with a cDNA encoding the extracellular domain of human IL-15Rα (COS-sIL-15Rα) also fully inhibited IL-15 binding to TF-1 cells with a similar efficiency (IC₅₀ = 35 pM).

The Kit 225 human cell line expresses the IL-2R β-chain and IL-2R γ-chain as well as the IL-2R α-chain and IL-15Rα, and proliferates in response to IL-2 or IL-15. The sIL-15Rα purified from the supernatants of different cell lines (COS-IL-15Rα, COS-sIL-15Rα, A172) were all found to inhibit the IL-15-dependent proliferation of Kit 225 cells (Fig. 7,A), while being without effect on the IL-2-dependent proliferation (Fig. 7 B). The inhibitory effects were complete with high affinity IC₅₀ in the range from 3 to 10 pM.

The murine 32DβR cells express endogenous mouse IL-2R γ-chain and transfected human IL-2R β-chain and IL-15Rα. The different sIL-15Rα also inhibited IL-15-induced proliferation of 32DβR (Fig. 7,C) with inhibitory effects comparable to those found with Kit 225 cells. As shown in Fig. 7 D, 32DβR proliferate dose dependently in response to human IL-15 (EC₅₀ = 6 pM). In the presence of a constant low concentration (15 pM) of soluble IL-15Rα purified from COS-IL-15Rα, this dose-response curve was found to be markedly shifted toward higher IL-15 concentrations (EC₅₀ = 80 pM). Material similarly purified from COS-7 cells transfected with empty vector did not affect the dose-response curve.

In this study, we have shown the existence of a soluble form of the human IL-15Rα that is naturally produced in the supernatants of cells expressing the membrane-anchored cognate receptor chain. This sIL-15Rα is constitutively released and this process can be increased by several stimuli. It binds IL-15 with high affinity and is able at low, picomolar concentrations to specifically block IL-15 interaction with its cell surface receptor and inhibit cell proliferation induced by IL-15.

The identification of sIL-15Rα was made feasible by the development of a highly sensitive RIA (detection limit of ∼1 pM). One of the Abs used in this RIA, M161, reacts with an IL-15Rα epitope involved in the binding of IL-15 and blocks this binding (J. Bernard, unpublished observations). We therefore anticipate that this RIA only detects sIL-15Rα that is not bound to endogenous IL-15 that might be produced by the cell lines investigated.

Two physiological mechanisms have been described for the generation of soluble receptors. The first one is based on the de novo synthesis of the soluble receptor from alternatively spliced mRNA, the second involves liberation of the soluble receptor following proteolytic cleavage of the membrane-anchored receptor counterpart (1). Two observations initially suggested that the sIL-15Rα was likely generated by proteolytic cleavage and release. First, whereas a number of alternatively spliced transcripts coding for different isoforms of membrane-anchored IL-15Rα have been identified (20, 26, 29), no transcript coding for sIL-15Rα has been described so far. Second, stimulation of soluble receptor production in culture supernatants was accompanied by a decrease in IL-15Rα expressed at the cell surface. Finally, direct proof of the involvement of receptor shedding was provided by transfection experiments in COS-7 cells that do not express endogenous human IL-15Rα. After transfection with a cDNA coding for the full-length IL-15Rα, these cells released in their supernatants a sIL-15Rα whose molecular mass (42 kDa) was 13 kDa lower than that of membrane bound IL-15Rα (55 kDa), and was indistinguishable from that of a recombinant form of the extracellular domain of IL-15Rα expressed in the same COS-7 cells. Thus, at least in these transfectants, alternative splicing could be excluded as being involved in sIL-15Rα generation. Soluble IL-15Rα species produced by the different cell lines examined had molecular masses also indistinguishable from that of recombinant soluble IL-15Rα, indicating the occurrence in these cells a shedding process similar as in COS-7 cells.

A number of protease inhibitors were tested to define the enzyme(s) responsible for the cleavage of the membrane bound IL-15Rα. In contrast to the protease inhibitors aprotinin, EDTA, leupeptin, PMSF, and S17092-1, the synthetic zinc-metalloproteinases inhibitor GM6001 blocked PMA-induced sIL-15Rα release in culture supernatants, while leaving basal unstimulated sIL-15Rα release unaffected. At least two different proteolytic mechanisms appear to participate in the cleavage of IL-15Rα, one of which specifically involving metalloproteinases, as already shown for other cytokine receptors (30), and in particular the structurally related IL-2R α-chain (13). The family of metalloproteases includes a disintegrin and metalloproteinase, ADAM, whose first identified member is the TNF-α converting enzyme (ADAM17), and the MMPs. As GM6001 is broad spectrum MMP inhibitor (28), additional experiments using purified enzymes are required to define the MMPs involved.

For the majority of proteins released by shedding, the proteases responsible have not been identified yet. In addition, it appears that for one protein, different proteases can be involved, and conversely, one protease can cleave several proteins. The size of the sIL-15Rα indicated that it was generated by cleavage at a site or sites proximal to the plasma membrane, as previously shown for other soluble cytokine receptors (1). This result suggests that at least one of the potential cleavage sites is located in a region of the extracellular domain of the receptor that is proximal to the transmembrane domain. Deletion of the exon 4 and exon 5 encoded domains that are in proximity to the plasma membrane did not abrogate the shedding processes, either basal or stimulated with PMA. Two hypotheses can be raised that are not mutually exclusive. The first is that several cleavage sites can be used that are located at several positions in the extracellular domain of the receptor, the second is that a cleavage site is located at the N-terminal end of the exon 6 encoded transmembrane domain and accessible to metalloproteinases. Additional deletion and mutational experiments are under way to define more precisely the site(s) of cleavage.

All cell lines tested that express membrane IL-15Rα (A172, TF-1, TF1-β, and Kit 225) released sIL-15Rα in their supernatants, and it can be anticipated that such IL-15Rα-positive cells may represent a potential source for this soluble receptor in vivo. The phorbol ester PMA, previously shown to enhance the proteolytic shedding of several cytokine receptors (24, 31), is similarly shown in this study to strongly enhance the shedding of sIL-15Rα, indicating that protein kinase C-dependent pathways are involved in the induction of receptor release (32). Ca²⁺ mobilization with ionomycin also stimulated IL-15Rα shedding, as previously shown for IL-6R (33). IL-1β and TNF-α also increased sIL-15Rα release, a process that might also be related to the reported induction of protein kinase C and MMP-9 by these cytokines (34).

In most cases, soluble receptor ectodomains generated by proteolysis appear to retain the ligand binding characteristics of their membrane-anchored counterparts. Therefore, in cases in which the receptor (IL-2Rα, IL-5Rα, IL-7Rα, GM-CSFRα) is a low affinity receptor that associates with other receptor subunits to build the high affinity membrane receptor complex, its soluble counterpart has an affinity that is much lower than that of the high affinity membrane receptor. In other cases (IL-4R, IL-1RI and IL-1RII; TNFRI and TNFRII) in which the receptor contributes to most of the affinity of the multimeric membrane receptor, the soluble counterpart has a high affinity similar to that of the membrane receptor. Our results indicate that the second case applies for the sIL-15Rα. It binds IL-15 with a high affinity similar to that measured for the cell-associated IL-15Rα expressed alone or in complex with the IL-2Rβ and IL-2Rγ subunit. Accordingly, the sIL-15Rα acted as a potent competitor of IL-15 binding to the membrane receptors, being able at low concentrations (IC₅₀ = 20 pM) to completely abrogate this binding. The sIL-15Rα behaved also as a potent and specific antagonist of the biological activity of IL-15. It completely inhibited IL-15-induced cell proliferation at concentrations (IC₅₀ = 3 to 10 pM) similar to those required to block IL-15 binding, without affecting IL-2 driven cell proliferation. Similar results have been reported for a soluble form of the IL-4R. In that case however, the amounts of sIL-4R needed to block the biological activity of IL-4 were substantially (∼10-fold) higher than those required for inhibition of IL-4 binding (35). This difference was due to a faster off-rate for the soluble form, a result consistent with a potential role as carrier protein for IL-4 (36). The fact that the sIL-15Rα is as or even more efficient in inhibiting biological activity than in binding argues against such a role. Whether it could act, like the sIL-4R (36) as a protective ligand against IL-15 degradation remains to be evaluated.

The existence of soluble receptors suggests that they have an important role in normal physiology, in the response to diseases and in the development of pathological processes (37). The IL-15 system is tightly regulated under physiological conditions and a number of disorders have been reported that are associated with activation or defects of the IL-15 system. Among these disorders are autoimmune diseases, inflammatory diseases, infectious diseases, transplant rejection, cancer, and immunodeficiencies (38). Inasmuch as IL-15Rα transcripts are expressed by many cell types and tissues, and as sIL-15Rα displays high affinity and IL-15 blocking efficiency in vitro, this soluble receptor can be anticipated to exert an important role in physiology and pathology. IL-15 has been detected in the biological fluids of patients with chronic inflammatory diseases such as rheumatoid arthritis (39), inflammatory bowel diseases (40), pulmonary diseases (41), and chronic active hepatitis (42). The availability of specific RIAs able to measure the levels of sIL-15Rα (this study) and IL-15-sIL-15Rα complex (E. Mortier, unpublished observations) will enable a better analysis of the regulation of the IL-15 system in such diseases.

Based on their cytokine neutralizing properties, some soluble receptors have been introduced in clinical trials, one well known example being the soluble TNFR used for the treatment of rheumatoid arthritis (43). Due to its efficient antagonist activity, sIL-15Rα is an attractive candidate for therapeutic applications. Indeed, preclinical studies in the mouse (44, 45) have already highlighted the potential interest of sIL-15Rα.