A Disintegrin and Metalloproteinase 10-Mediated Cleavage and Shedding Regulates the Cell Surface Expression of CXC Chemokine Ligand 16

The scavenger receptor for phosphatidylserine and oxidized low-density lipoprotein (SR-PSOX)³ and CXC chemokine ligand (CXCL)16 were independently identified as a scavenger receptor and chemokine, respectively, but have since been shown to be identical (1, 2, 3). SR-PSOX was identified through an expression cloning strategy designed to identify receptors that could mediate cell adhesion to phosphatidylserine-coated surfaces (1). It was subsequently shown to bind and internalize oxidized low-density lipoprotein (OxLDL) (1), making it a member of the structurally diverse scavenger receptor family of cell surface receptors that is defined by the ability to recognize modified low-density lipoprotein (4). Further analysis showed that SR-PSOX is expressed by macrophages in vitro and in atherosclerotic lesions in vivo (1, 5), suggesting that SR-PSOX activity may be involved in the massive accumulation of cellular cholesterol during the generation of macrophage foam cells associated with atherosclerotic lesion development (6). Similar to other members of the scavenger receptor family, SR-PSOX has recently been shown to mediate the uptake of Gram-positive and -negative bacteria when expressed by macrophages and dendritic cells, indicating that this receptor may play a role in innate immunity and initiation of the acquired immune response (7, 8).

CXCL16 was independently identified by two groups as the ligand for the orphan G-protein-coupled chemokine receptor Bonzo/CXCR6 (2, 3). CXCL16 is the second transmembrane chemokine identified to date, and bears significant structural homology to fractalkine/CX₃C chemokine ligand (CX₃CL)1 (9, 10). A combination of immunohistochemistry and FACS analysis showed that CXCL16 is selectively expressed by APCs, including DCs, macrophages, and B cells (2, 3). When expressed by macrophages, soluble CXCL16 is released into the medium and has chemoattractant activity that is mediated solely through the CXCR6 receptor. CXCR6 is expressed by many cell types including naive CD8⁺ T cells, NK T cells, and a subset of memory CD4⁺ T cells, although only activated CD4⁺ and CD8⁺ T cells appear to migrate strongly to the soluble chemokine (2, 11). Furthermore, CXCL16-positive cells in the spleen were seen in close opposition to CD8⁺ T cells, suggesting that, similar to CX₃CL1, CXCL16 may act as an intercellular adhesion molecule when expressed on the cell surface (2, 12, 13).

Given the potential distinct functional activities of membrane-bound CXCL16 as a scavenger receptor, and soluble CXCL16 as a chemokine, the mechanisms that regulate the conversion between these two forms would appear to be important for determining the role played by this molecule in vivo. We have previously shown that membrane-bound CX₃CL1 can be proteolytically cleaved from the cell surface by at least two distinct metalloproteinases (14). We identified a disintegrin and metalloproteinase (ADAM) family member ADAM17 as the protease responsible for stimulated shedding of CX₃CL1 (14, 15), whereas constitutive release of CX₃CL1 has subsequently been shown to be mediated by ADAM10 (16). Given the structural similarity between CX₃CL1 and CXCL16, we have examined whether similar proteolytic mechanisms are responsible for generating soluble CXCL16. In this study, we demonstrate that CXCL16 is synthesized as an intracellular precursor that is rapidly transported to the cell surface where it undergoes metalloproteinase-dependent cleavage. By manipulating the functional expression of ADAM10, we show that it is responsible for the constitutive shedding of CXCL16, and that this proteolytic activity is a key regulator of CXCL16 cell surface expression. This identification of ADAM10 as a major protease responsible for the conversion of CXCL16 from a membrane-bound scavenger receptor to a soluble chemoattractant should help elucidate the physiological function of this molecule.

The murine macrophage cell line RAW-264.7 was maintained in RPMI 1640 supplemented with 10% FCS. Thioglycolate-elicited peritoneal macrophages (ThioMφ) were isolated by peritoneal lavage with PBS from C57BL/6 mice injected 4 days previously with 1 ml of 3% thioglycolate i.p., and were cultured in RPMI 1640 plus 10% FCS. Bone marrow-derived macrophages (BMDMs) were generated by culturing bone marrow cells in RMPI 1640 containing 10% FCS and 2000 U/ml recombinant human M-CSF (R&D Systems, Minneapolis, MN) as previously described (17). Dermal fibroblasts expressing an IFN-γ-inducible, temperature-sensitive SV40 large T Ag transgene, and lacking functional expression of ADAM9 or ADAM17, were isolated as previously described (18). These cells were maintained at 32°C in DMEM plus 10% FCS plus 5 U/ml recombinant murine (rm)IFN-γ (R&D Systems). For experiments, dermal fibroblasts were plated in the absence of IFN-γ and grown at 37°C in DMEM plus 10% FCS for 24 h to arrest immortalization by the large T Ag. The following Abs were used: polyclonal rabbit anti-hemagglutinin (HA) epitope tag (Zymed Laboratories, San Francisco, CA), rat monoclonal and biotinylated goat polyclonal anti-mCXCL16, rat monoclonal anti-mADAM10 (R&D Systems), and PE- and peroxidase-conjugated streptavidin, anti-rabbit and anti-rat IgG (Jackson ImmunoResearch, West Grove, PA). PMA and all other chemicals not specified were from Sigma-Aldrich (St. Louis, MO). GM6001 was purchased from Elastin Products (Owensville, MO).

All constructs were generated using standard molecular biology techniques, and were verified by DNA sequencing. A cDNA-encoding murine CXCL16 with a cytoplasmic, C-terminal HA epitope tag was amplified by RT-PCR using primers containing a 5′ BamHI site and a 3′ HA epitope tag and a NotI site, and cloned into the retroviral expression vectors pBM-IRES-EGFP (19) and pBM-IRES-PURO (20). An expression construct for murine ADAM10 was generated by cloning an RT-PCR-amplified cDNA containing 5′ BamHI and 3′ NotI sites into the pBM-IRES-PURO retroviral expression vector. PCR mutagenesis was subsequently used to mutate Glu³⁸⁵>Ala, generating a catalytically inactive ADAM10 mutant (ADAM10 E>A). A 315-bp fragment of the mouse U6 gene promoter was PCR amplified from genomic DNA using primers to incorporate 5′ XhoI and BamHI sites, a PmeI site at the transcriptional start site and 3′ EcoRI and HindIII sites. The fragment was cloned into the vector LZRS-SIN-CD68L-HA-EGFP (21), which had been digested with XhoI and HindIII to remove the CD68-HA-EGFP expression cassette, generating the plasmid SIN-U6. Complementary oligonucleotides encoding an siRNA hairpin with a 9-bp loop against nt 609–627 of mouse ADAM10 cDNA were annealed and cloned via the PmeI and EcoRI sites of SIN-U6 to generate the plasmid SIN-U6 ADAM10. Further details of the sequences of the oligonucleotides and plasmids used for cloning are available on request. High-titer retroviral supernatants were prepared by calcium phosphate-mediated transfection of Phoenix amphotropic packaging cells (generously provided by G. Nolan (Stanford University, Stanford, CA)) as previously described (19). For transduction, 2 × 10⁵ cells were plated per well of a six-well plate 24 h before a 10-h incubation with retroviral supernatants containing 4 μg/ml Polybrene (Sigma-Aldrich). Transduction efficiency was enhanced by centrifuging plates at 1700 × g for 2 h at 37°C at the beginning of the 10-h incubation period. After transduction, retroviral supernatant was replaced with fresh medium, and cells were allowed to recover for at least 48 h before use in subsequent experiments. For puromycin selection, cells were cultured in the presence of 15 μg/ml puromycin for 48 h following recovery from transduction.

Macrophages (RAW-264, BMDM, and ThioMφ) were plated at a density of 2.5 × 10⁶ cells per 60-mm dish, and dermal fibroblasts at a density of 5 × 10⁵ cells per 60-mm dish in complete growth medium 24 h before stimulation. Cells were pretreated for 20 min by addition of GM6001 (50 μM final) or DMSO vehicle control directly to the culture medium. Cells were washed with serum-free medium and stimulated with 2 ml of serum-free medium with or without PMA (100 ng/ml), GM6001 (50 μM), or DMSO control, followed by incubation at 37°C for 30 min or the indicated time. Following stimulation, supernatants were removed, and cells were washed once with cold PBS and subsequently lysed with 1 ml of radioimmunoprecipitation assay (RIPA) buffer (PBS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 2 μg/ml aprotinin, 2 μg/ml leupeptin, 1 μg/ml pepstatin, and 100 μg/ml PMSF). Resulting cell supernatants and lysates were cleared by centrifugation at 15,000 × g and stored at −20°C until analysis. CXCL16 concentrations in cellular lysates and supernatants were determined by ELISA using a rat monoclonal capture Ab, a biotinylated goat polyclonal detection Ab, and an rCXCL16 standard (R&D Systems). The levels of CXCL16 present in medium or in detergent extracts were determined for triplicate dishes and are reported as the average ± SD.

Cells were plated and stimulated as described above. Poststimulation, cells were washed twice with PBS and lysed in 350 μl of RIPA buffer for 30 min on ice. Cell lysates were cleared by centrifugation at 15,000 × g for 10 min, and protein concentrations were determined using the bicinchoninic acid protein assay (Pierce, Rockford, IL). Lysates were separated by SDS-PAGE under reducing conditions, transferred to Immobilon PVDF membranes (Millipore, Bedford, MA), and subsequently immunoblotted with specific Abs, before visualization by ECL (Amersham Pharmacia Biotech, Piscataway, NJ). For metabolic labeling, cells were washed twice with PBS and then incubated with RPMI 1640 lacking cysteine and methionine, supplemented with 10% dialyzed calf serum for 1 h. Cells were subsequently labeled for 30 min with 500 μCi/ml Translabel [³⁵S]Met and [³⁵S]Cys (ICN, Irvine, CA), and chased for the indicated times in RPMI 1640 plus 10% FCS. Cells were washed twice with PBS and lysed as above, and CXCL16 was immunoprecipitated by overnight incubation with 5 μg of anti-HA Ab (Zymed Laboratories) and 30 μl of a 50% slurry of protein A-agarose (Santa Cruz Biotechnology, Santa Cruz, CA). Immunoprecipitates were separated by 12% SDS-PAGE under reducing conditions and visualized by autoradiography. For cell surface protein biotinylation, cells were washed in cold PBS and incubated where indicated with 1 mg/ml NHS-LC-biotin (Pierce) in PBS for 45 min on ice. Labeling reagent was quenched with 0.1 M glycine, and cells were lysed in RIPA buffer as above. Where indicated, cell extracts were incubated with 100 μl of a 50% slurry of agarose-streptavidin for 1 h at 4°C followed by centrifugation at 2500 × g to remove biotinylated proteins. Equal volumes of the resulting cell extracts were separated by SDS-PAGE and immunoblotted with an anti-HA Ab as above.

Cell surface levels of CXCL16 were determined by staining cells stimulated with PMA in the presence or absence of GM6001 as described above, with a rat monoclonal anti-CXCL16 Ab followed by PE-conjugated anti-rat IgG secondary Ab. CXCL16 expression was measured by flow cytometry using a FACScan (BD Biosciences, San Jose, CA) flow cytometer and data analysis with CellQuest software. Relative CXCL16 cell surface expression is reported as the average mean fluorescence intensity of triplicate samples ± SD.

Total RNA was isolated from macrophage populations (RAW-264, BMDM, and ThioMφ) using TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s recommended protocol. RNA was reverse transcribed using Moloney murine leukemia virus reverse transcriptase and an oligo(dT) primer (Invitrogen), with reactions set up in the presence and absence of reverse transcriptase to control for genomic DNA contamination. The cDNA obtained served as a template for PCR using the oligonucleotide pairs specific for murine ADAM2, -8, -9, -10, -12, -15, and -17 (sequence of primers are available on request). PCR products were separated by electrophoresis on a 1.2% agarose gel and visualized by ethidium bromide staining.

CXCL16 can exist as two functionally distinct forms: a cell-associated form and a soluble form that has been proposed to arise from processing of the membrane-bound protein (2, 3). To study the relationship between these two forms, we expressed murine CXCL16 containing a cytoplasmic, C-terminal HA epitope tag (Fig. 1,A) in the murine macrophage cell line RAW-264.7, which does not express endogenous CXCL16 (B). Western blot analysis of detergent cell extracts using Abs recognizing either the CXCL16 extracellular domain or the HA epitope tag revealed multiple CXCL16 protein species (Fig. 1,B). Both Abs recognized a ∼55-kDa form that has previously been suggested to represent mature CXCL16 protein (2). Similar Western blots performed after depletion of cell surface proteins through their biotinylation and incubation with agarose-streptavidin confirmed that this 55-kDa form was the only CXCL16 species present at the cell surface (Fig. 1,C). A band of ∼15 kDa that migrated close to the gel front was detected by the Ab against the cytoplasmic tail HA epitope tag, but not by the Ab against the CXCL16 extracellular domain, suggesting that it was generated following proteolytic cleavage of full-length CXCL16 (Fig. 1,B). Conditioned cell medium contained a single ∼40-kDa CXCL16 species that could be detected only with an Ab against the extracellular domain of CXCL16, again indicating that it was generated through the juxtamembrane cleavage of full-length CXCL16 (Fig. 1 B).

To characterize further the maturation and processing of CXCL16, we performed pulse chase analysis of RAW-264 cells expressing murine CXCL16 (Fig. 1,D). CXCL16 is synthesized as 28- and 35-kDa precursors that undergo rapid maturation, presumably by glycosylation, to yield 55-kDa mature CXCL16 (Fig. 1,D). These precursors can also be detected under steady-state conditions through Western blotting of cell lysates (Fig. 1,B). Mature CXCL16 disappears very rapidly from the cell, with a half-life of ∼45 min. The disappearance of CXCL16 is associated with the generation of a 15-kDa fragment that can be immunoprecipitated with an Ab against the C-terminal HA epitope tag, although this fragment does not accumulate within the cell (Fig. 1 D). These data suggest that CXCL16 is initially synthesized as an intracellular precursor that undergoes very rapid glycosylation and transport to the cell surface as a 55-kDa glycoprotein. This mature CXCL16 protein can then be released from the cell surface, yielding a soluble 40-kDa fragment that likely contains the majority of the glycosylated ectodomain and a 15-kDa transmembrane cytoplasmic tail fragment.

It has recently been shown that the constitutive release of CX₃CL1 is mediated by ADAM10, and this can be inhibited by broad-spectrum metalloproteinase inhibitors (14, 15, 16). Given the structural similarity of CX₃CL1 and CXCL16, we sought to determine whether the constitutive shedding of CXCL16 was also mediated by a metalloproteinase. When expressed by RAW-264 cells, CXCL16 is constitutively released at a constant rate as determined by ELISA measurements of CXCL16 present in cell supernatants (Fig. 2,A). Constitutive CXCL16 release could be efficiently inhibited by a broad-spectrum zinc-dependent metalloproteinase inhibitor, GM6001 (Fig. 2,B), and this led to a corresponding increase in the amount of cell-associated CXCL16 and a decrease in the ratio of soluble/cellular CXCL16 consistent with decreased CXCL16 shedding (Fig. 2 B).

In addition to ADAM10-mediated constitutive release, CX₃CL1 shedding can be markedly stimulated by phorbol esters such as PMA, a process that is mediated by another ADAM family protease, ADAM17 (14, 15). To examine whether shedding of CXCL16 could be similarly enhanced, we measured CXCL16 in lysates and supernatants of RAW-264 cells expressing CXCL16 following stimulation with PMA in the presence or absence of GM6001 (Fig. 2,B). PMA stimulation led to a modest increase in soluble CXCL16 levels, a concomitant slight decrease in cell-associated CXCL16, leading to an increase in the soluble/cellular CXCL16 ratio (Fig. 2 B). This enhanced shedding of CXCL16 in response to PMA stimulation could be effectively inhibited by GM6001 pretreatment. Although the enhanced shedding of CXCL16 induced by PMA was highly reproducible, the magnitude of this increase was somewhat variable, and was always significantly less than we have previously observed for CX₃CL1 (14).

To examine the consequences of both constitutive and PMA-induced shedding on the expression of CXCL16, we used flow cytometry to measure the levels of cell surface CXCL16 (Fig. 2,C). PMA stimulation caused a modest, but reproducible, decrease in the cell surface expression of CXCL16. In contrast, inhibition of constitutive CXCL16 shedding, through incubation of cells with GM6001, led to an almost doubling of CXCL16 expression on the cell surface. This increase was also seen in cells stimulated with PMA in the presence of GM6001. The significance of constitutive shedding on the expression of CXCL16 was confirmed by Western blot analysis of detergent cell extracts using Abs against extracellular and intracellular CXCL16 epitopes (Fig. 2 D). As seen with FACS analysis, PMA stimulation slightly decreased levels of mature, 55-kDa CXCL16, whereas pretreatment with GM6001 caused a significant increase in the amount of this form of the protein. Taken together, these results support a model in which full-length mature CXCL16 at the cell surface is constitutively cleaved by a metalloproteinase to release the majority of the CXCL16 ectodomain, leaving a cell-associated cytoplasmic tail fragment. Furthermore, this metalloproteinase-mediated constitutive cleavage appears to play an important role in determining the cell surface expression of CXCL16.

The data presented so far show that CXCL16 can be constitutively released by a metalloproteinase when overexpressed in RAW-264 cells, and this shedding can be moderately enhanced by stimulation with PMA. To determine whether this shedding was a consequence of high-level overexpression, we examined the release of endogenous CXCL16 in primary macrophages. ThioMφ and BMDM both shed CXCL16 constitutively at a rate similar to that seen in RAW-264 cells, and this release can be inhibited by GM6001 (Fig. 3). PMA stimulation caused an increase in the shedding of CXCL16 by ThioMφ, and a more modest increase in BMDMs, although this was largely attributable to the differences in the rate of constitutive CXCL16 shedding. In both cell types, the PMA-induced component could be inhibited by GM6001, indicating that it was also mediated by metalloproteinase activity. Hence, shedding of endogenous and overexpressed CXCL16 appears to be mediated by similar mechanisms.

Previous studies examining the shedding of CX₃CL1, in addition to the inhibition studies above, suggested that ADAM proteases were good candidates for the cleavage and shedding of CXCL16 (14, 15, 16). To examine which ADAM proteases were expressed by macrophages that have the ability to shed CXCL16, we performed RT-PCR analysis using primer pairs specific for multiple ADAM family members (Fig. 4). ThioMφ, BMDM, and RAW-264 cells show a similar profile of ADAM expression, including no or low-level expression of ADAM2 and -12, but significant expression of ADAM8, -9, -10, -15, and -17, making these good candidates for the shedding of CXCL16.

To begin to examine the role played by specific ADAM proteases, we analyzed CXCL16 shedding in cells derived from ADAM-deficient mice. For these studies, we used immortalized dermal fibroblasts isolated from mice obtained by crossing ADAM-deficient mice with transgenic mice expressing an IFN-γ-inducible and temperature-sensitive SV40 large T Ag allele (14, 18). Dermal fibroblasts derived from wild-type (DF-WT), or ADAM9 (DF-9KO)- and ADAM17 (DF-17KO)-deficient mice were transduced with a retrovirus encoding CXCL16 with a C-terminal HA epitope tag, and constitutive and PMA-inducible CXCL16 shedding was determined by ELISA. Wild-type dermal fibroblasts shed CXCL16 constitutively at a rate similar to that seen in RAW-264 cells, and this shedding could be modestly enhanced by stimulation with PMA (Fig. 5). Similar to macrophages, CXCL16 shedding by dermal fibroblasts was also metalloproteinase mediated, as determined by experiments using GM6001 (data not shown). Despite some variability between experiments, we never observed any significant differences in the constitutive or PMA-induced shedding of CXCL16 in dermal fibroblasts lacking functional expression of ADAM9 or ADAM17 (Fig. 5). This indicates that, unlike CX₃CL1, ADAM17 does not play a major role in the generation of soluble CXCL16.

Given that the constitutive release of CX₃CL1 is mediated by ADAM10 (16), we next examined whether this protease also plays a role in the shedding of CXCL16. To modulate the expression of ADAM10, we took advantage of the ability of siRNAs to efficiently reduce the expression of a target gene in mammalian cells (22). To drive the expression of siRNAs, we developed a novel retroviral vector, SIN-U6, containing the RNA polymerase III U6 gene promoter in a self-inactivating (SIN) vector backbone, such that only the U6 promoter has transcriptional activity in transduced cells (Fig. 6,A) (23). To generate an siRNA specific for ADAM10, we cloned in a DNA fragment encoding a dsRNA hairpin with a 9-bp loop targeting nt 609–627 of the ADAM10 coding sequence, which would be predicted to be processed by the RNase III enzyme Dicer to generate an active siRNA (24). To test the function of this retroviral vector, we used it to transduce dermal fibroblasts expressing CXCL16, and examined the expression of ADAM10 by Western blotting. Cells transduced with the empty SIN-U6 vector showed a similar level of ADAM10 expression to that of mock-transduced control cells (Fig. 6,B). In contrast, cells transduced with the SIN-U6 ADAM10 vector showed a >90% reduction in ADAM10 protein levels (Fig. 6 B), and this appeared specific for ADAM10, because the expression of ADAM17 remained unchanged (data not shown).

The constitutive and PMA-induced shedding of CXCL16 in cells with reduced expression of ADAM10 was determined by ELISA. Mock- and SIN-U6 vector-transduced cells showed a similar constitutive and PMA-induced shedding of CXCL16, but there was a significant reduction of 60% in the basal shedding of CXCL16 by cells expressing the ADAM10 siRNA (Fig. 6,C). Similar to other cell types, PMA-induced shedding of CXCL16 in cells transduced with the SIN-U6 ADAM10 vector was variable, but was still detected (Fig. 6,D). To characterize the CXCL16 shedding activity that remains in cells expressing the ADAM10 siRNA, we examined its sensitivity to GM6001. Both constitutive and PMA-inducible shedding of CXCL16 could be completely abolished by GM6001, suggesting that there was either incomplete inhibition of ADAM10 function, or that there is/are other metalloproteinase(s) mediating CXCL16 shedding (Fig. 6 D).

To confirm the data obtained through the expression of an siRNA against ADAM10, we examined CXCL16 shedding in dermal fibroblasts overexpressing wild-type ADAM10 or a catalytically inactive Glu³⁸⁵>Ala ADAM10 mutant. We have previously shown that a similar catalytically inactive ADAM17 mutant has dominant-negative properties, inhibiting the shedding of CX₃CL1 and VCAM-1, presumably through competition with endogenous enzyme for substrate, or key regulatory cytoplasmic factors (14, 18). Western blotting of transduced cells revealed significant overexpression of both ADAM10 proteins when expressed using a pBM-IRES-PURO retroviral vector (Fig. 7,A). Overexpression of wild-type ADAM10 led to a >2-fold increase in the constitutive release of CXCL16, but the relative increase in CXCL16 shedding in response to PMA stimulation remained unchanged (Fig. 7 B). In contrast, overexpression of the catalytically inactive ADAM10 mutant inhibited CXCL16 constitutive shedding by >50%, without affecting the relative magnitude of the PMA-induced increase in CXCL16 shedding. These results confirm the data from cells expressing an siRNA against ADAM10, namely that ADAM10 contributes significantly to the constitutive shedding of CXCL16.

Having already shown that metalloproteinase activity regulates the expression of CXCL16 on the cell surface, we wanted to explore whether ADAM10-mediated shedding was responsible for these observations. Flow-cytometric analysis showed that overexpression of ADAM10 resulted in a significant decrease, whereas catalytically inactive ADAM10 increased the cell surface levels of CXCL16 compared with cells transduced with the empty vector control (Fig. 7,C). These results were confirmed by Western blot analysis of detergent cell extracts using an Ab against the CXCL16 extracellular domain (Fig. 7 D). As seen with FACS analysis, overexpression of ADAM10 decreased levels of mature, 55-kDa CXCL16, whereas overexpression of the catalytically inactive ADAM10 mutant caused a significant increase in the amount of this form of the protein. Taken together, these results demonstrate that ADAM10 plays a significant role in the constitutive cleavage and shedding of CXCL16, and its activity is a key determinant of CXCL16 cell surface levels.

It has previously been shown that CXCL16 can exist as either a membrane-anchored cell surface protein or as a soluble chemokine (1, 2, 3). In this study, we have shown that CXCL16 is synthesized as an intracellular precursor, undergoes rapid maturation, presumably by glycosylation, followed by transport to the cell surface where it quickly becomes a substrate for metalloproteinase-mediated cleavage and shedding. This high rate of basal shedding can be modestly enhanced by the phorbol ester PMA through the activity of a metalloproteinase. Using a novel retroviral vector to drive the expression of siRNAs, and overexpression of wild-type and catalytically inactive enzyme, we identify ADAM10 as the predominant protease responsible for constitutive shedding of CXCL16. Furthermore, we show that ADAM10-mediated cleavage is a key regulator of CXCL16 cell surface expression.

Cleavage of CXCL16 by ADAM10 is consistent with its structural similarity to the other transmembrane chemokine identified to date, CX₃CL1, whose constitutive cleavage and shedding is also mediated by ADAM10 (16). However, unlike CX₃CL1 where PMA stimulation induces a very robust increase in shedding, incubation of cells expressing CXCL16 with PMA only produced a modest and variable enhancement of basal CXCL16 release (14). Indeed, we could not detect any differences in constitutive or PMA-induced shedding of CXCL16 in cells lacking functional ADAM17 expression, unlike our studies looking at CX₃CL1 shedding (14). Given the higher rate of basal, ADAM10-mediated shedding of CXCL16 when compared with CX₃CL1, it may be that any role for ADAM17 in CXCL16 shedding may be masked. However, PMA-stimulated shedding remained weak when ADAM10 function was inhibited through siRNA-mediated knockdown or the expression of dominant-negative ADAM10 (Figs. 6,C and 7,B). Indeed, the enhancement of CXCL16 cell surface expression seen in cells stimulated with PMA in the presence of GM6001 (Fig. 2 C) suggests that the enhanced shedding seen in response to PMA may reflect an increase in intracellular protein trafficking rather than stimulation of a specific enzymatic activity. These apparent differences in proteolytic specificity between CXCL16 and CX₃CL1 means that domain-swap experiments of their juxtamembrane regions may prove valuable in determining the sequence and structural epitopes required for generating ADAM10 and ADAM17 substrates.

Determining the physiological relevance of ADAM10-mediated shedding of CXCL16 will prove difficult given the current tools available. ADAM10-deficient mice die at day 9.5 of embryonic development with multiple defects of both the central nervous and cardiovascular systems (25), and this therefore precludes studies addressing the role of ADAM10-mediated shedding of CXCL16 in vivo. In an attempt to examine the significance of ADAM10-mediated CXCL16 shedding in macrophages in vitro, we used the various retroviral constructs described in this paper to enhance and inhibit ADAM10 function in the RAW-264 cell line. Despite the ability of these tools to modulate ADAM10 function in dermal fibroblasts, expression of either wild-type or catalytically inactive ADAM10 enzyme, or ADAM10 siRNA led to significant cell death, and only a 10–20% increase or decrease in ADAM10 expression in surviving cells (data not shown). These observations were identical independent of the expression of CXCL16. We are currently investigating the significance and mechanism of this observation, but it suggests that the precise level of ADAM10 activity is critical for survival in this macrophage cell line.

Recent data have suggested that constitutive ectodomain shedding observed in vitro may actually represent a form of induced shedding stimulated by the serum component lysophosphatidic acid (LPA) (26). Although definitive evidence linking LPA binding by its G-protein-coupled receptor (GPCR) to activation of ADAM10 proteolytic activity is lacking, LPA has been shown to activate the shedding of known ADAM10 substrates including epidermal growth factor (EGF) and heparin-binding EGF (27, 28, 29). Other GPCR agonists have been shown to induce activation of ADAM10 proteolytic activity, including bombesin and platelet-activating factor, mediating receptor transactivation through the shedding of EGF receptor ligands (30, 31). In addition to GPCR-mediated activation, ADAM10 proteolytic activity has also been shown to be regulated by cellular cholesterol levels. Depletion of plasma membrane cholesterol using the cholesterol acceptor methyl-β-cyclodextrin, has been shown to stimulate ADAM10-mediated shedding of the L1 adhesion molecule, the IL-6R, and the amyloid precursor protein (32, 33, 34). Furthermore, inhibition of endogenous cholesterol biosynthesis by the 3-hydroxy-3-methylglutaryl-CoA reductase inhibitor lovastatin, also stimulated ADAM10 proteolytic activity (33). We have also seen enhanced shedding of CXCL16 in response to treatment of RAW-264 cells with methyl-β-cyclodextrin (data not shown). Given the ability of CXCL16 to act as a scavenger receptor and internalize cholesterol from OxLDL, we are currently examining the intriguing possibility that CXCL16 cell surface levels may be up-regulated by a positive-feedback loop due to the inhibition of ADAM10-mediated shedding caused by increased cellular cholesterol during macrophage foam cell formation.

We have shown that ADAM10 proteolytic activity is a key determinant of CXCL16 cell surface expression, and therefore of its ability to act as a scavenger receptor for modified low-density lipoprotein and bacteria. To attempt to address this issue, we have examined the ability of RAW-264 cells expressing CXCL16 to internalize fluorescently labeled OxLDL in the presence or absence of GM6001. To our surprise, overexpression of CXCL16 did not augment the ability of these cells to endocytose OxLDL (data not shown). Similar experiments performed using dermal fibroblasts, which do not show detectable endogenous scavenger receptor activity, also failed to reveal any significant uptake of OxLDL by cells expressing CXCL16 (data not shown). In addition to the uptake of OxLDL, experiments looking at the phagocytosis of fluorescently labeled Escherichia coli by both RAW-264 cells and dermal fibroblasts failed to show any enhanced uptake by cells overexpressing CXCL16, even when assays were done in the presence of GM6001 to maximize the cell surface expression of CXCL16 (data not shown). We are currently examining the reasons for the differences between our results and published data (1, 8), but it could represent variations in experimental conditions, the cell types used, and the species of CXCL16 being used.

In addition to affecting scavenger receptor activity, ADAM10-mediated cleavage would be predicted to affect the function of CXCL16 as a chemokine at multiple levels. As a soluble form, CXCL16 can act as a chemoattractant for CD4⁺ and CD8⁺ T cells bearing the receptor CXCR6 (2, 3), and therefore ADAM10-mediated shedding of membrane-bound CXCL16 would be predicted to act as a regulator of CXCL16 chemoattractant bioactivity. Membrane-bound CX₃CL1 has been shown to act as an intercellular adhesion molecule for cells expressing the CX₃CR1 receptor (12, 13). During the course of the review of this manuscript, Shimaoka et al. (35) reported that CXCL16-CXCR6 interactions can similarly mediate intercellular adhesion. Furthermore, they show that these adhesive interactions can be strengthened by increasing the cell surface expression of CXCL16, suggesting that ADAM10-mediated CXCL16 shedding may regulate these adhesive properties as is seen for CX₃CL1 (16). Definitive answers regarding these in vivo functions of CXCL16 as a scavenger receptor and chemokine will be provided by analyses of CXCL16-deficient mice. Irrespective of the true in vivo role of CXCL16, ADAM10-mediated cleavage is likely to act as an important regulator of CXCL16 function.

We thank Julie Philalay and Cindy Chang for technical assistance, and Garry Nolan for providing retroviral expression plasmids.