The Low Affinity IgE Receptor (CD23) Is Cleaved by the Metalloproteinase ADAM10*

The low affinity IgE receptor, FcϵRII (CD23), is both a positive and negative regulator of IgE synthesis. The proteinase activity that converts the membrane-bound form of CD23 into a soluble species (sCD23) is an important regulator of the function of CD23 and may be an important therapeutic target for the control of allergy and inflammation. We have characterized the catalytic activity of ADAM (a disintegrin and metalloproteinase) 10 toward human CD23. We found that ADAM10 efficiently catalyzes the cleavage of peptides derived from two distinct cleavage sites in the CD23 backbone. Tissue inhibitors of metalloproteinases and a specific prodomain-based inhibitor of ADAM10 perturb the release of endogenously produced CD23 from human leukemia cell lines as well as primary cultures of human B-cells. Expression of a mutant metalloproteinase-deficient construct of ADAM10 partially inhibited the production of sCD23. Similarly, small inhibitory RNA knockdown of ADAM10 partially inhibited CD23 release and resulted in the accumulation of the membrane-bound form of CD23 on the cells. ADAM10 contributes to CD23 shedding and thus could be considered a potential therapeutic target for the treatment of allergic disease.

The low affinity IgE receptor Fc⑀RII (CD23) is a 46-kDa type II membrane protein that is expressed on B-cells and cells of the myeloid lineage (1). CD23 has multiple functions. It is both a positive and negative regulator of IgE synthesis (2). It facilitates IgE-dependent antigen presentation through its binding of IgEantigen complexes (3,4). Moreover, the release of proinflammmatory cytokines from macrophages is stimulated by CD23 binding to CD18/11b and /11c (complement receptors 3 and 4) (5-7).
In humans, two isoforms of CD23 that differ by only seven amino acids in the short N-terminal cytoplasmic domain are observed (8). CD23a is expressed only on B-cells. Stimulation of B-cells and cells of the myeloid lineage with interleukin-4 (IL-4) 3 induces the expression of CD23b. The C-terminal extracellular domain consists of a globular fold that has homology to the C-type lectin family (9). This globular domain has been shown to contain two distinct binding sites, one for IgE and a second that recognizes CD21 (complement receptor 2) (10,11). At the cell surface CD23 self-assembles to form homotrimers that have a higher affinity for IgE than the CD23 monomer (12,13). The self-association is driven by a leucine zipper-like domain (14) that connects the N-terminal cytoplasmic and transmembrane domains to the C-terminal globular domain.
Homotrimeric CD23 molecules exhibit a 15-nm ␣-helical coiled coil stalk that extends the globular C-terminal domains from the plasma membrane (15). Cleavage in the stalk region by a membrane-associated endoproteolytic activity generates soluble fragments of CD23 (sCD23) that possess apparent molecular masses of 37, 33, and 29 kDa (16). All three of these sCD23 fragments exist as homotrimers (15). Smaller fragments of CD23 (25 and 16 kDa) are known. However, these are thought to be formed from cleavage of the larger soluble fragments by a cysteine proteinase (16). In vitro sCD23 promotes the differentiation of germinal center B-cells into plasma cells (17), stimulates IgE synthesis in B-cells (18), and induces the secretion of proinflammatory cytokines in monocytes (6,19,20). In vivo, inhibition of proteolytic shedding of CD23 using small molecule metalloproteinase inhibitors has been shown to suppress IgE synthesis in Scid mice that have transplanted human peripheral blood lymphocytes (21,22).
The identities of the proteinase activities responsible for generating the 33-and 37-kDa forms of CD23 remain elusive. Release of the 33-kDa fragment from the human B-cell line RPMI 8866 has been shown to be mediated by a membraneassociated metalloproteinase activity (23). It has been demonstrated that members of the disintegrin/metalloproteinase fam-ily (ADAM8, -15, and -28) are capable of cleaving CD23 (24). However, the peptide sequence used to test proteolytic activity was not derived from the known 33-and 37-kDa cleavage sites of CD23 (16). In this paper, we provide evidence that ADAM10 is catalytically competent in cleaving human CD23 at the known physiologic sites and that it sheds CD23 from human cell lines and primary cultures of human B-cells.

EXPERIMENTAL PROCEDURES
Reagents-Commercial reagents were obtained from the following sources: Recombinant human IL-4 and recombinant human catalytic/disintegrin domains of ADAMs 8, 10, and 17 were from R&D Systems (Minneapolis, MN). Tissue inhibitors of metalloproteinase (TIMPs) were from R&D Systems. TAPI-2 was obtained from Calbiochem. Complete proteinase inhibitor mixture was from Roche Applied Science. All other chemicals unless specified otherwise were obtained from Sigma. 2,4-Dinitrophenyl-labeled (DNP) peptides (Table 1) were purchased either from the University of North Carolina, Chapel Hill, Peptide Chemistry Department or from SynPep (Dublin, CA). Recombinant prodomain constructs of ADAM10 (A10-(23-213) and A10-(23-181)) were expressed in Escherichia coli and purified by nickel-histidine affinity chromatography followed by gel filtration. 4 Antibodies-Rabbit polyclonal antisera directed against ADAM8, -10, and -17 were from Chemicon (Temecula, CA). Rabbit polyclonal anti-hemagglutinin (HA) and anti-FLAG tags were from Sigma. Monoclonal anti-FLAG (M2) was from Stratagene (La Jolla, CA) and anti-HA (HA.11) from Covance (Berkeley, CA). Rabbit polyclonal ant-CD23 was from Anaspec (San Jose, CA). Human CD23 ELISA kits were from BD Biosciences and Invitrogen.
Determination of Enzyme Concentration-The substrate dabcyl-LAQA(homoPhe)RSC(fluorescein)-NH 2 , 5-10 M, was added from a 5 mM Me 2 SO stock into buffer, 20 mM Tris, pH 7.5, and 0.0075% Brij 35. Enzyme (10 l of a 1/30 diluted stock solution; 200 g/ml) was added to a 96-well black-coated Costar plate. TIMP1, from a 7.6 M stock solution, was diluted to 1 M in the above buffer. It was then serially diluted in 3-fold increments from 1 M to 3 nM. The diluted TIMP1, 10 l, was added to the enzyme and allowed to preincubate for 10 -15 min after which substrate in buffer (80 l) was added to initiate the reaction. Reaction progress was monitored continuously in a Cytofluor plate reader by following the change in fluorescence using an excitation of 485 nm and an emission of 530 nm. Data were fit to the Morrison equation to calculate the enzyme concentration (44).
k cat /K m Determination for DNP-Peptide Substrates-The specificity constant k cat /K m (s Ϫ1 M Ϫ1 ) is calculated as k cat /K m ϭ M/(3600⅐A e ⅐C e ), where M (counts/h) is the slope of the net absorbance versus time curve in the early linear range, A e (counts) is the net increase in absorbance reading at reaction end, C e (M) is the concentration of enzyme in reaction, and 3600 is a conversion factor (s/h).
Cleavage of DNP-Peptide Substrates-Recombinant ADAM proteinase (ADAM8, -10, or -17), 5 M, was diluted into 50 l of buffer containing 20 -100 M DNP-substrate. The final concentration of enzyme varied from 50 to 500 nM. The reaction progressed at 37°C for 40 -120 min. The reactions were quenched with an equal volume of 1% heptafluorobutyric acid and spun to remove precipitated protein. Reaction products were run on a C18 Vyadac column using solvents containing 0.1% HFBA with an acetonitrile/water gradient. Peaks were detected with a wavelength of 350 nM. Cleavage sites were determined either by liquid chromatography/mass spectros-copy (LC/MS; Synpep and Duke University) or by co-chromatography with standard products using two HPLC systems.
Retroviral Transduction-Stable cell lines expressing dominant negative ADAM constructs and siRNAs were produced by retroviral transduction using vesicular stomatitis virus type G (VSV-G) enveloped viruses. The viruses were produced using a pantropic packaging system (Clontech). Retroviral expression vectors (pQCXIP or pSIREN-retro-Q) were co-transfected (FuGENE 6, Roche Applied Science) with a vector encoding the VSV-G envelope protein into GP2-293T packaging cells. 48 h later, U937 cells were suspended in the conditioned medium from the packaging cells that was supplemented with 4 g/ml Polybrene; 24 h after transduction, stable cells were selected with 2 g/ml puromycin for 5 days.
Co-immunoprecipitation-HEK293T cells were seeded into 10-cm plates and allowed to adhere overnight. The cells were then co-transfected (Transfectin, Bio-Rad) with expression vectors encoding ADAM10⌬MP-FLAG and CD23-HA or vector controls (pQCXIP or pIRES2eGFP). Forty-eight hours posttransfection the cells were washed twice with PBS (all procedures at 4°C) and then lysed in co-immunoprecipitation buffer (50 mM Tris-HCl, pH 7.4, 140 mM NaCl, 10 mM CHAPS, 10% glycerol, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, Complete proteinase inhibitor mixture). Lysates were clarified by centrifugation (16,500 ϫ g, 20 min). The total protein concentration of the supernatants was adjusted to 1 mg/ml with lysis buffer. Polyclonal sera (5 g, Sigma) against either the HA or FLAG epitopes were added to 1 ml of lysate, incubated for 2 h, and then precipitated using protein A-agarose (Invitrogen). The agarose beads were washed with co-immunoprecipitation buffer and then treated with loading buffer, separated by SDS-PAGE, and blotted onto PVDF membranes. The membranes were probed with mouse monoclonal antibodies against the FLAG epitope (M2, Stratagene) or against the HA epitope (HA.11, Covance).
The supernatants harvested following centrifugation (10 min at 1,000 ϫ g) were then assayed for CD23 content by ELISA.
Time Course for CD23 Shedding-U937 cells (both vector control and ADAM10 siRNA transduced cells) were cultured with IL-4 overnight. Cells were washed twice with PBS and resuspended at a density of 500,000/ml in fresh growth medium. Cells were incubated at 37°C for time periods of 1, 3, 5, and 7 h. After each time point, clarified supernatants were obtained by centrifugation (10 min at 1,000 ϫ g) and stored at Ϫ80°C until CD23 levels could be determined by ELISA.

RESULTS
In human serum, there are five observed molecular weights of sCD23: 37, 33, 29, 25 and 16 kDa. The cleavage sites that give rise to the 37-and 33-kDa forms of the receptor are known and were used to design N-terminal DNP-labeled peptide substrates. Cleavage of the peptides by recombinant human ADAM10 (rhADAM10) was monitored by HPLC separation of the cleavage reactions followed by colorimetric detection of the DNP-chromophore. The peptides used and the resulting specificity constants (k cat /K m ) that were determined for each cleavage reaction are reported in Table 1. The specificity constant for a DNP-labeled peptide spanning the TNF-␣ cleavage site was determined to be 1200 M Ϫ1 s Ϫ1 and is comparable with the literature value of 2000 M Ϫ1 s Ϫ1 reported for the N-terminalacetylated peptide (25). The substrate DNP-HGDQ-MAQKSQST-NH 2 is based on the cleavage site that gives rise to the 37-kDa form of sCD23. This substrate was cleaved by rhADAM10 with a specificity constant of 250 M Ϫ1 s Ϫ1 . The specificity constant of 250 M Ϫ1 s Ϫ1 indicated that the efficiency of the cleavage reaction was comparable to that of DNP-EVH-HGKLVTFFAE (350 M Ϫ1 s Ϫ1 ), a peptide derived from the ␣-secretase cleavage site of amyloid precursor protein (APP), which is a known ADAM10 substrate (25). The peptide corresponding to the 33-kDa cleavage site sequence of CD23, DNP-RAEQQRLKSQDL, was cleaved in two places by rhADAM10 having specificity constants of 190 and 90 M Ϫ1 s Ϫ1 . The proteolysis by rhADAM10 appeared to be selective, as other DNPlabeled peptides derived from the known cleavage sites of TNF receptor I (Table 1) and epidermal growth factor, heparin-binding epidermal growth factor, IL-6 receptor, TNF receptor II, and L-selectin (data not shown) were not processed by this enzyme.
We also determined the cleavage site specificity that rhADAM10 exhibited toward the 37-and 33-kDa CD23 peptides and compared it with that of rhADAM8 and rhADAM17 ( Table 2). The 37-kDa peptide was cleaved by both rhADAM10 and rhADAM8 at the physiological Ala-Gln site. The 33-kDa peptide was cleaved by rhADAM10 at both the physiological Arg-Leu site and a nonphysiological Gln-Gln site. In contrast, rhADAM8 cleaved the 33-kDa peptide only at the nonphysiological Gln-Gln and Glu-Gln sites. Recombinant human ADAM17, the closest family member of ADAM10, did not possess activity toward either the 37-or 33-kDa peptide but was able to cleave a peptide derived from the cleavage site of TNF-␣ (data not shown).
To examine the ability of ADAM10 to cleave CD23 in a cellbased assay, we transiently co-transfected ADAM10 and CD23 into HEK293T cells. Wild-type ADAM10 (A10), ADAM10 containing a point mutation (E385A) that renders the proteinase inactive (A10EA), or vector alone was co-transfected with C-terminal HA-tagged CD23 (CD23-HA). After 48 h, the conditioned medium was harvested, and cells were washed and solubilized in radioimmune precipitation assay buffer. Samples from each transfection were separated by SDS-PAGE, transferred to PVDF membranes, and analyzed by Western blotting for soluble and membrane-bound CD23-HA. The top and bottom panels in Fig. 1A show the conditioned medium and cell lysates, respectively, probed with an anti-HA antibody for the presence of CD23-HA. Transfection of A10 enhanced the accumulation of CD23-HA in the supernatant. Both the A10EA and the vector control transfectants exhibited a background level of CD23 shedding that was partially inhibited using the broad spectrum metalloproteinase inhibitor TAPI-2. Quantification of the band intensity in each lane from multiple experiments indicated that transfection with A10 increased shedding by about 50% over vector control transfection (Fig. 1B). In contrast, transfection with the catalytically inactive A10EA did not increase sCD23 production relative to the vector control. Cotransfection of HEK293T cells with ADAM8 (A8) and CD23-HA resulted in a marked increase in sCD23 in the medium (Fig. 1C) as reported previously (24). In contrast, cotransfection of ADAM17 (A17) marginally affected shedding of CD23-HA from these cells.
The expression of endogenous CD23 is positively regulated by IL-4 in many types of leukocytes. However, the effect of IL-4 on ADAM10 expression is unknown. We cultured U937 promonocytic leukemia cells in the presence of IL-4 for 24 h and analyzed the expression of ADAM10 in cell lysates by Western blotting (Fig. 2, top panel). IL-4 slightly increased the amount of ADAM10 present in the lysates. However, it was a much stronger positive regulator of CD23 (Fig. 2, middle  panel). Increases in sCD23 upon exposure to IL-4 are more  likely due to the elevated expression of CD23 than to changes in the levels of ADAM10. The effect of inhibitors of metalloproteinases on CD23 shedding was assessed in U937 promonocytic leukemia cells, Ramos cells (a human B-cell line), and primary cultures of human peripheral blood B-cells. The cell lines were cultured for 24 h in the presence of IL-4 and each inhibitor. The amount of sCD23 in the conditioned medium was then determined by ELISA. In both cell lines, the broad spectrum metalloproteinase inhibitor GM6001 appeared to inhibit 50 -60% of CD23 shedding (Fig. 3,  A and B). A second broad spectrum inhibitor, TAPI-2, was slightly more potent than GM6001 (70% inhibition) in U937 cells (Fig. 3A). Recombinant TIMPs 1, 2, and 3 all exhibited modest inhibitory activity toward CD23 shedding in U937 cells, with TIMPs 1 and 3 being more potent than TIMP2. In Ramos cells TIMP1 inhibited CD23 release (Fig.  3B) to a similar extent as done in U937 cells.
The recombinant prodomain of ADAM10 (amino acids-(23-213), wtA10-(23-213)) is a potent, selective inhibitor (K i ϭ 48 nM) of ADAM10 in vitro and in cell culture. 4 Mutation of the cysteine switch region (C173S) improves the stability of the reagent and maintains the potency (K i ϭ 36 nM) and selectivity toward ADAM10. At concentrations of 600 and 300 nM the C173S mutant (A10-(23-213)) inhibited shedding of CD23 from U937 cells by about 30% and from Ramos cells by about 20% (Fig. 3, A  and B). At lower concentrations the inhibitory activity diminished in this cellular assay of CD23 shedding. A truncated prodomain (A10-(23-181)), based on the full-length Cys-to-Ser mutant of ADAM10, was a poor inhibitor of the metalloproteinase in vitro (K i ϭ Ͼ5 M) and weakly inhibited sCD23 production (Fig. 3, A and B).
Primary cultures of peripheral blood B-cells isolated from normal healthy adults were treated either with wtA10-(23-213) or vehicle control and cultured with or without IL-4 for 12 h at 37°C. The cultures were then stained with fluorochrome-labeled anti-CD19 and anti-CD23 and analyzed by flow cytometry. In the absence of IL-4, treatment of the cultures with wtA10-(23-213) induced a shift in the percentage of CD23/ CD19 double-positive cells from 38% in the diluent control sample to 56% in the wtA10-(23-213)-treated sample (Fig. 3C). Conversely, the percentage of CD19-positive, CD23-negative cells decreased from 53% in the diluent control sample to 36% in the wtA10-(23-213)-treated sample. Treatment with proteolytically inactive A10-(23-181) did not affect the percentage of positive cells (data not shown). The density (determined by mean fluorescence intensity) of CD23 on the double-positive cells was found to be insensitive to the presence of wtA10-(23-213). IL-4-treated cells responded differently to wtA10-  in that treatment with wtA10-(23-213) did not affect the percentage of CD23/CD19 double-positive cells in the sample, but it increased the density of CD23 present on the CD23/CD19 double-positive cells (Fig. 3D). The IL-4-stimulated, wtA10-(23-213)-treated CD19ϩ cells exhibited 2.5-fold greater mean density of cell surface CD23 than the vehicle control-treated cells. Prolonged treatment of IL-4-stimulated cells (36 h) with wtADAM10-(23-213) increased the percentage of cells that were CD23-positive from 83% in the diluent control to 98% in the inhibitor-treated sample (data not shown).
The ADAM family of proteinases is also involved in the shedding of cell surface proteins that is stimulated by various small molecule agonists. Cells cultured overnight with IL-4 were  washed and treated with various agonists for 45 min at 37°C. After centrifugation, the sCD23 in the cleared supernatants was determined by ELISA (Fig. 4A). Both PMA and lysophosphatidic acid (LPA) strongly induced shedding. Other reported agonists of shedding such as ATP and ionomycin only weakly induced shedding of CD23 from U937 cells. Sphingosine 1-phosphate did not stimulate shedding. Both A10-(23-213) and TAPI-2 were used to evaluate the contribution of ADAM10, and of metalloproteinases in general, to the stimulated shedding of CD23 (Fig. 4B). The weakly induced shedding by ATP and ionomycin was inhibited slightly more so by A10-(23-213) than the unstimulated control, whereas inhibition by TAPI-2 was relatively constant. In the samples strongly stimulated by LPA and PMA, A10-(23-213) was a poorer inhibitor of CD23 release than in the unstimulated control. Although TAPI-2 inhibited PMA-induced shedding in a manner comparable with the unstimulated control, it was a somewhat weaker inhibitor of LPA-induced shedding.
Mutant versions of ADAM proteins that lack metalloproteinase domains have been shown to function as dominantnegative alleles that suppress shedding of proteinase substrates (26 -28). We examined the effect of metalloproteinase-deleted alleles of ADAM8 (A8⌬MP), ADAM10 (A10⌬MP), and ADAM17 (A17⌬MP) on the shedding of CD23 from U937 cells. The mutant alleles were cloned into a retroviral expression vec-tor (pQCXIP) and transduced into U937 cells by retroviral delivery. Puromycin-resistant cells were then stimulated with IL-4 overnight, and the conditioned medium was assayed for the presence of sCD23 (Fig. 5A). Both A8⌬MP and A10⌬MP exhibited significantly reduced (20 -25%) production of sCD23 compared with the vector control cells. The A17⌬MP-transduced cells produced amounts of sCD23 similar to vector control cells. These results are consistent with both ADAM8 and ADAM10 being CD23 sheddases and with ADAM17 having no activity toward this protein.
We also found that A10⌬MP physically associated with CD23 (Fig. 5B). A C-terminal FLAG-tagged ADAM10 construct (A10⌬MP-FLAG) was co-transfected with CD23-HA into HEK293T cells. The transfected cells were cultured for 48 h and then washed with PBS and lysed in co-immunoprecipitation buffer. The extracts were then cleared by centrifugation, and the supernatants were immunoprecipitated using polyclonal sera against the respective epitope tags. The precipitated complexes were separated by SDS-PAGE and transferred to PVDF membranes. When CD23-HA and A10⌬MP-FLAG were co-transfected, immunoprecipitation of CD23-HA resulted in the co-precipitation of A10⌬MP (Fig. 5B, top left  panel). In the absence of transfected CD23-HA, A10⌬MP-FLAG was not detectable following precipitation. In both lysates A10⌬MP-FLAG was expressed, although A10⌬MP-FLAG expression levels were considerably higher in the absence of CD23-HA (Fig. 5B, lower left panel). Immunoprecipitation of A10⌬MP-FLAG resulted in the co-precipitation of CD23-HA in co-transfected cells (Fig. 5B, upper right panel). In the absence of transfected A10⌬MP-FLAG, a residual amount of CD23-HA still precipitated. The expression level of CD23-HA was similar in both lysates.
Endogenous ADAM10 in U937 cells was depleted using RNA interference. A construct encoding a hairpin siRNA specific for human ADAM10 (29) was cloned into a retroviral RNA interference vector and transduced into U937 cells along with a control siRNA. The stably transduced cells exhibited ADAM10 protein levels that were depleted ϳ50% compared with the control cells (Fig. 6A). Overnight stimulation of these cells with IL-4 followed by assay of the conditioned media using a CD23  ELISA revealed that the amount of CD23 produced was reduced by 35% in the conditioned medium from the ADAM10 siRNA-treated cells versus the control cells (Fig.  6B). Both the ADAM10 siRNA cells and the control cells were cultured with and without IL-4 and/or TAPI-2 for 24 h. The concentrated conditioned medium and whole cell lysates were separated by SDS-PAGE and transferred to PVDF membrane for Western blotting with CD23 antiserum (Fig. 6C, top two panels) or actin antiserum (loading control; Fig. 6C, bottom panel). In the absence of IL-4, very little CD23 was present in either the conditioned medium (Fig. 6C, top panel) or the cell lysates (middle panel). In the presence of IL-4, CD23 was induced in both the conditioned medium and in the cell lysate. In the conditioned medium the CD23 appeared to migrate with an apparent molecular mass of 37 kDa. In the cell lysate, CD23 appeared to migrate at ϳ45 kDa. The accumulation of the 37-kDa form of CD23 in the conditioned medium was inhibited using the broad spectrum metalloproteinase inhibitor TAPI-2. Conversely, treatment of IL-4-stimulated cells with TAPI-2 led to an increase in the amount of the higher molecular mass form of CD23 associated with the cells. Compared with the control cells, ADAM10-depleted cells produced less of the 37-kDa soluble form of CD23 than the control cells. However, the ADAM10-depleted cells possessed more of the cell-associated 45-kDa form of CD23 than the wild-type cells. Quantification of the blots revealed that the ratio of sCD23 to mCD23 in the ADAM10depleted cells was ϳ0.4, and that of the control cells was 1.4.
We also assessed the rate of release of soluble CD23 from ADAM10-depleted versus wildtype cells. Cells were stimulated for 24 h and then washed with medium, resuspended in fresh medium, and incubated at 37°C s for 1-7 h. The conditioned medium at each time point was then used to assay soluble  CD23 production by ELISA. A plot of CD23 concentration in the conditioned medium versus time shows that the rate of sCD23 production is lower in the ADAM10-depleted cells than in the control cells (Fig. 6D). The control cells produced CD23 at a rate of 0.19 Ϯ 0.04 ng⅐ml Ϫ1 ⅐h Ϫ1 . In contrast, the ADAM10depleted cells released soluble CD23 at a rate of 0.11 Ϯ 0.02 ng⅐ml Ϫ1 ⅐h Ϫ1 , and TAPI-2-treated control cells released sCD23 at a rate of 0.02 Ϯ 0.01 ng⅐ml Ϫ1 ⅐h Ϫ1 .

DISCUSSION
Proteolytic shedding of membrane proteins is an evolutionarily conserved post-translational modification that regulates the physiological activity of numerous proteins (30). The ADAM family of disintegrin/metalloproteinases has been shown to catalyze many of these cell surface endoproteolytic events. The cellular and physiological processes that are directly regulated by ADAMs include growth factor activation, chemokine mobilization, cell fate specification, and cell adhesion (31)(32)(33). Release of the 33-kDa sCD23 fragment from the human B-cell line RPMI 8866 has been shown to be mediated by membrane-associated metalloproteinase activity (23). ADAM8 has been shown capable of cleaving CD23 when cotransfected in HEK293 cells (24). However, the peptide sequence used to test activity of ADAM8 on CD23 was not derived from the known 33-and 37-kDa cleavage sites of CD23 (16). Despite the potential role of sCD23 in allergic and inflammatory diseases, the identity of the proteinase activities responsible for generating sCD23 remains elusive.
We investigated shedding of CD23 in human cells because the functional significance of mouse sCD23 is unclear. Serum IgE levels are elevated in CD23 knock-out mice indicating that CD23 is a negative regulator of IgE in vivo (34). However, transgenic mice that over-express mouse sCD23 exhibit no apparent IgE phenotype (35). Several biochemical differences between human and mouse CD23 may explain this disparity. Mouse CD23 is 55% identical to human CD23 at the amino acid level. The location and identity of the cleavage sites of mouse sCD23 are not conserved in human CD23 (36). Human sCD23 stimulates IgE synthesis in vitro through binding CD21 (18). Recent structural data predict that mouse sCD23 lacks a CD21 binding site (10). In addition, mouse sCD23 binds to IgE with an affinity that is Ͼ10-fold weaker than human CD23-IgE interaction (36). Because of these disparities, mice may not be adequate models for the function of sCD23 in IgE homeostasis.
Identification of the cellular proteinase that cleaves a particular substrate is facilitated when the physiological cleavage site of that substrate is known. The peptide cleavage data indicate that the ADAM10 catalytic domain possesses the catalytic capability to cleave CD23 directly at sites that are known to be physiologically relevant. There is evidence that cleavage of protein substrates by ADAMs likely involves ADAM-substrate interactions outside of the catalytic domain (37). We have demonstrated that a metalloproteinase-deficient mutant of ADAM10 co-immunoprecipitates with CD23, suggesting that ADAM10 and CD23 are present within the same complex in the cell. Thus, there is strong support for a direct action of ADAM10 on CD23.
Cell-based assays support the cellular role of ADAM10 in shedding CD23. Although overexpression of ADAM10 enhanced shedding of CD23, inhibition of ADAM10 function through either the use of specific inhibitors, dominant-negative ADAM10 constructs, or depleting ADAM10 by siRNA led to a reduction in CD23 shedding. However, depletion of cellular ADAM10 was not complete in cells transduced with the ADAM10 siRNA. Because targeted interference with ADAM10 function did not reduce CD23 shedding as much as the use of the broad spectrum metalloproteinase inhibitor TAPI-2, inhibition of its activity may have been incomplete. Shedding of integral membrane proteins may take place within intracellular compartments as well as at the cell surface (38). Although ADAM10-(23-213) is a potent, specific ADAM10 inhibitor in vitro, its polypeptide composition may render it cell-impermeable and thus may not efficiently inhibit ADAM10 in a cellbased assay. The residual ADAM10 activity may account, at least part, for the remaining CD23 sheddase activity in these cells.
A second reason for the failure of targeted inhibition of ADAM10 to inhibit CD23 shedding completely in the cellbased assays may be the presence of other proteinases that are functionally redundant with ADAM10. This phenomenon has been observed for other ADAMs substrates (39). For example, utilizing both cell-based assays and peptide cleavage assays, it has been demonstrated that the amyloid precursor protein, APP, can be cleaved by ADAM8, ADAM9, ADAM10, and ADAM17 (40). Mouse embryonic fibroblasts from Adam10 Ϫ/Ϫ mice process APP normally (41). However, the phenotypes resulting from the overexpression of ADAM10 as well as dominant-negative ADAM10 in a mouse model of Alzheimer disease are consistent with ADAM10 being a physiological APP sheddase (42). Stimulated shedding of CD23 from U937 cells by different small molecule agonists was differentially inhibited by the A10-(23-213), indicating that different stimuli activate different CD23 processing activities. This cellular behavior has been observed with other ADAM substrates such as CD44 (43). Both the Ramos and U937 cells express RNA transcripts for almost every proteolytically active member of the ADAM family including several previously characterized CD23 sheddases. The fact that CD23 is processed physiologically at several sites, and that several ADAMs are expressed in the cell types that express CD23, makes it probable that CD23 shedding is a process controlled by several enzyme activities.
In summary, we have shown that ADAM10 is a physiological CD23 sheddase. The targeted inhibition of ADAM10 in vivo may induce the accumulation of mCD23 and reduce levels of sCD23. Because mCD23 inhibits IgE synthesis and sCD23 stimulates IgE synthesis, alteration of the proteolytic balance between mCD23 and sCD23 through targeted inhibition of ADAM10 may prove to be an important therapeutic intervention point for controlling allergic disease.
Addendum-While this manuscript was under review, an article was published (45) that found ADAM10 to be the major CD23 sheddase activity in mice.