CD93 is a highly glycosylated transmembrane protein expressed on monocytes, neutrophils, endothelial cells, and stem cells. Antibodies directed at CD93 modulate phagocytosis, and CD93-deficient mice are defective in the clearance of apoptotic cells from the inflamed peritoneum. In this study we observe that CD93, expressed on human monocytes and neutrophils, is susceptible to phorbol dibutyrate-induced protein ectodomain shedding in a time- and dose-dependent manner. The soluble fragment found in culture supernatant retains the N-terminal carbohydrate recognition domain and the epidermal growth factor repeats after ectodomain cleavage. Importantly, a soluble form of the CD93 ectodomain was detected in human plasma, demonstrating that shedding is a physiologically relevant process. Inhibition of metalloproteinases with 1,10-phenanthroline inhibited shedding, but shedding was independent of TNF-α-converting enzyme (a disintegrin and metalloproteinase 17). Phorbol dibutyrate-induced CD93 shedding on monocytes was accompanied by decreased surface expression, whereas neutrophils displayed an increase in surface expression, suggesting that CD93 shed from the neutrophil surface was rapidly replaced by CD93 from intracellular stores. Cross-linking CD93 on human monocytes with immobilized anti-CD93 mAbs triggered shedding, as demonstrated by a decrease in cell-associated, full-length CD93 concomitant with an increase in CD93 intracellular domain-containing cleavage products. In addition, the inflammatory mediators, TNF-α and LPS, stimulated ectodomain cleavage of CD93 from monocytes. These data demonstrate that CD93 is susceptible to ectodomain shedding, identify multiple stimuli that trigger shedding, and identify both a soluble form of CD93 in human plasma and intracellular domain containing cleavage products within cells that may contribute to the physiologic role of CD93.
Protein ectodomain cleavage, or shedding, is a common feature of adhesion molecules that is generally mediated by metalloproteases (1). Shedding modulates homing properties of activated leukocytes (2) and releases bioactive ectodomain products. For example, soluble TNF-α, TGF-α, TGF-β, and members of the epidermal growth factor (EGF)3 family are all derived from shedding of transmembrane proteins (3). Another consequence of shedding is the release of the intracellular domain (ICD) fragment after regulated cleavage of the transmembrane domain (4). The ICD fragment of Notch, CD44, and others relocates to the nucleus and regulates gene transcription (5). Therefore, shedding affects numerous aspects of cell biology and subsequent physiology via both intracellular and extracellular pathways.
Shedding is largely mediated by zinc-dependent metalloproteinases of the matrix metalloproteinase, membrane type-matrix metalloproteinase or a disintegrin and metalloproteinase (ADAM) family. TNF-α-converting enzyme (TACE) (ADAM17) is the most widely investigated sheddase and is responsible for cleavage of numerous transmembrane proteins, including L-selectin, TNF-α, TGF-α, TGF-β, members of the EGF family, and the amyloid precursor protein (6). The majority of mice made deficient in TACE die between embryonic day 17.5 and the first day after birth, illustrating the importance of ectodomain shedding in mammalian development (7). However, TACE is not a universal sheddase, because a variety of transmembrane proteins, including CD44, are cleaved by distinct metalloproteinases (8).
CD93 (C1qRP, AA4.1 (mouse)) is a 100,000 Mr type 1 transmembrane glycoprotein that has been implicated in the regulation of cell-cell interactions during development and in the efficiency of phagocytosis (9, 10, 11, 12). Mice deficient in CD93 display an impairment in the uptake of apoptotic cells in vivo. In addition, an IgM anti-CD93 mAb (R3) triggers phagocytosis when cells are plated on the immobilized Ab and inhibits phagocytosis when cells are treated in the fluid phase (13). However, the mechanisms responsible for these effects on phagocytosis are unknown (10). CD93 has been implicated in regulating adhesive processes, because ligation of CD93 with the F(ab′)2 of the mAb mNI-11 stimulates homotypic aggregation of LPS-stimulated U937 cells, and homotypic aggregation is inhibited by Abs against LFA-1 or ICAM-1 (14). The same Ab triggers rapid spreading of HUVEC cells when immobilized on a surface, and this spreading is inhibited by cytochalasin D (such spreading is not detected with control samples after HUVEC exposure to anti-CD44) (15). CD93 is expressed on the endothelium and on circulating myeloid cells and platelets, consistent with a role for CD93 in cellular homing to sites of inflammation (12, 16). Furthermore, CD93 expression defines the earliest human bone marrow stem cells because it is expressed on CD34-negative and -positive hemopoietic and hepatic precursors, and CD93 is a common marker for murine B cell development (17, 18).
Previous studies demonstrated that glycosylation of CD93 stabilized surface expression and that under conditions of induced hypoglycosylation, increased levels of CD93 were detected in culture medium, suggesting that CD93 was proteolytically cleaved from the surface (19). Because shedding has been implicated as an important mechanism for regulation of other adhesion molecules with structural and functional similarities to CD93 (e.g., L-selectin (2) and CD44 (8)), we chose to investigate cell stimuli that might induce CD93 shedding. The work described in this study shows that CD93 is susceptible to metalloproteinase-mediated shedding induced by the phorbol ester, phorbol dibutyrate (PDBu), independent of the glycosylation state. Interestingly, surface cleavage of CD93 occurs in both human monocytes and neutrophils, but surface CD93 expression is actually increased in neutrophils, possibly due to an intracellular store of CD93 within neutrophil granules. Importantly, soluble CD93 (sCD93) was detected in normal human plasma, suggesting that the cleavage event is physiologically relevant, and cytoplasmic tail cleavage products were detected in cell lysates, indicating that the liberated CD93-ICD may also have biologic activity. In addition, CD93 shedding was induced by cross-linking with mAbs and the inflammatory mediators, TNF-α and LPS, indicating physiologic pathways that trigger this event. These data demonstrate a novel mechanism of regulated expression of CD93 that may have implications in stem cell development and inflammation.
Materials and Methods
Reagents and Abs
R139 and R3 were generated by immunization with C1q-binding proteins isolated from U937 cell extracts as previously described (13, 20). The polyclonal anti-CD93 cytoplasmic tail Ab 1150 was generated against the C-terminal 11 aa of CD93 as previously described (21). Polyclonal anti-CD93 1157 was generated by immunization with sCD93 produced in insect SF9 cells as previously described (16). The anti-L-selectin mAb, Dreg 56, was purchased from eBiosciences. FITC- or PE-conjugated secondary F(ab′)2 Abs were obtained from Jackson ImmunoResearch Laboratories. TNF-α was purchased from PeproTech. C5a was a gift from R. DiScipio (La Jolla Institute for Molecular Medicine, San Diego, CA). TNF-α protease inhibitor (TAPI) 1 and -2 were obtained from Peptides International. R0-31-9790 and R0-32-7315 were obtained from Hoffmann La Roche. Ultra Pure LPS was purchased from List Biological Laboratories. All media were obtained from Invitrogen Life Technologies unless otherwise noted. All other reagents were purchased from Sigma-Aldrich at the highest quality unless stated otherwise.
Cells and plasma
Human monocytes and lymphocytes were purified by counterflow elutriation using a modification of the technique of Lionetti et al. (22) as described previously (23). Plasma from normal healthy donors was saved at the time of elutriation and centrifuged at 16,000 × g for 10 min to remove platelets before use. In some experiments, IgG was depleted from plasma by overnight incubation at 4°C with 120 μl of GammaBind G-Sepharose (Amersham Biosciences) to 200 μl of plasma. Neutrophils were purified from human blood as previously described (24) and resuspended in HBSS+/0.25% human serum albumin (HSA; American Red Cross, distributed by FFF Enterprises) with 20 mM HEPES. ldlD cells were a gift from Dr. M. Krieger (Massachusetts Institute of Technology, Cambridge, MA) and were transfected with CD93 (ldlD-CD93) as previously described (19). ldlD-CD93 were maintained in serum-free medium, Chinese hamster ovary-serum-free medium II, supplemented with 100 U/ml penicillin G sodium/100 μg/ml streptomycin sulfate (Invitrogen Life Technologies) and 400 μg/ml G418 sulfate (Cellgro). To reconstitute wild-type glycosylation, 20 μM galactose and 400 μM N-acetylgalactosamine were added to culture medium 48 h before experiments were conducted. HEK293T cells were cultured in DMEM with 10% FBS (HyClone), 100 U/ml penicillin G sodium/100 μg/ml streptomycin sulfate (Invitrogen Life Technologies), 1× nonessential amino acids (Invitrogen Life Technologies), and 10 mM HEPES, pH 7.4.
For analysis of surface expression by flow cytometry, monocytes were resuspended at 2 × 106/ml in RPMI 1640/10% heat-inactivated FBS, 10 mM HEPES and stimulated as described in the figure legends. For ELISA studies, monocytes were resuspended at densities ≥8 × 106/ml before activation, and neutrophils were resuspended at ≥3 × 107/ml. For immunoprecipitation studies, monocytes were treated in serum-free HL1 medium (BioWhittaker). For inhibitor treatment, 5 mM 1,10-phenanthroline (dissolved in methanol to 2 M) was added immediately before addition of PDBu. For TACE inhibitor studies, inhibitors were added 30 min before addition of PDBu. All TACE inhibitors were dissolved in DMSO, except for TAPI-2, which was dissolved in HL1 medium (BioWhittaker). For Ab cross-linking experiments, Abs were coated on two-well Lab-Tek chamber slides (Nalge Nunc International) at 50 μg/ml for 2 h in 0.1 M carbonate buffer, pH 9.5. Chamber slides were washed with sterile PBS, and 2 × 106 cells in 1 ml of RPMI 1640/10% FCS/10 mM HEPES were added, centrifuged at 70 × g for 3 min, and incubated at 37°C for various time periods. Monolayers were washed with ice-cold PBS, and adherent cells were harvested by scraping in 0.5 ml of extraction buffer (0.3% Nonidet P-40, 10 mM triethanolamine, 1 mM CaCl2, 1 mM MgCl2, 0.15 M NaCl, 0.1 M PMSF, and 1/200 protease inhibitor mixture set III (Calbiochem)) on ice. Cell supernatants and wash containing nonadherent cells were spun down at 1,000 rpm for 10 min, washed with ice-cold PBS, and added to the cell lysates. Cells were lysed for 1 h on ice and centrifuged at 14,000 rpm for 15 min at 4°C, and soluble material was analyzed by ELISA or Western blot.
Cells (1 × 106) were washed twice in ice-cold FACS buffer (HBSS+, 0.2% BSA, and 0.2% sodium azide) and incubated with primary Abs (3 μg of IgM/R3 or 1 μg of IgG1/Dreg 56) for 30 min on ice with shaking every 10 min. Cells were washed twice in ice-cold FACS buffer and incubated with 2 μl of FITC- or PE-conjugated secondary Abs (F(ab′)2) for 30 min on ice, with shaking every 10 min. Cells were washed three times and either analyzed immediately or fixed overnight in 1% formaldehyde (prepared fresh) before analysis. Analysis was performed using the FACSCalibur apparatus (BD Biosciences) and the CellQuest program.
CD93 sandwich ELISA using anti-CD93 mAbs R139 and R3 was performed as previously described (19). To estimate the amount of CD93 in picomolar concentrations, values were compared with a standard curve of U937 cell extract. U937 cells were previously shown to contain 8.1 × 104 molecules of CD93/cell (19). In some experiments, values in picomolar concentrations were divided by the total number of cell equivalents. To detect recombinant CD93 fusion proteins anti-V5 (Invitrogen Life Technologies) was used at a 1/1000 dilution to capture recombinant CD93 ectodomain fragments, and R3 was used to detect. To determine whether sCD93 contained a cytoplasmic tail, an ELISA was used in which 1150 (rabbit polyclonal anti CD93 cytoplasmic tail) was used at 20 μg/ml to capture, and R3 was used to detect.
Immunoprecipitation and Western blot
Tissue culture supernatants were precleared for 2 h with 10 μg of IgG2b (ICN Pharmaceuticals) and 25 μl of a 1/1 slurry of GammaBind G-Sepharose (Amersham Biosciences). CD93 was immunoprecipitated from precleared tissue culture supernatants by incubation with 10 μg of R139 for 1 h at 4°C, followed by a similar incubation with 25 μl of a 1/1 slurry of GammaBind G-Sepharose. The beads were washed three times in extraction buffer (described above) and boiled in nonreduced SDS-PAGE sample buffer. For Western blot analysis, proteins were resolved by SDS-PAGE and transferred to polyvinylidene difluoride (PVDF). After blocking in 5% milk/TBST (20 mM Tris, 150 mM NaCl, and 0.1% Tween 20), blots were probed for 2 h at room temperature with primary Abs. Blots were washed and probed with secondary HRP-conjugated Abs for 1 h at room temperature and developed using ECL (Amersham Biosciences).
Expression of rCD93 extracellular domains
The CD93-CRD was amplified from pcDNA3.1-CD93 (11) using the following primers: 5′-GGAATTCAGAGGGCCACACAGAGACCG and 3′-GATATCTGAAGCT-GAACTTGCACAC. The CD93 CRD-EGF5 was amplified using the same 5′ primer and 3′-GATATCTGGTGCAAGAGACCCC. The endogenous CD93 signal sequence was included in the constructs. The PCR products were subcloned into pGEM-T (Promega) and sequenced at the University of California-Irvine core facility. The products from pGEM-T were cloned into a modified pENTR-11 (Invitrogen Life Technologies) vector that had been engineered to remove its start codon. Target genes were transferred to pcDNA-DEST40 via λ phage site-specific recombination as described by the manufacturer of GATEWAY cloning technology (Invitrogen Life Technologies). HEK 293T cells were transfected with the indicated plasmids or, as a control, full-length CD93 in pCDNA3.1 using LipofectAMINE (Invitrogen Life Technologies) according to the manufacturer’s instruction. Supernatants were used as a source of recombinant proteins. Proteins were purified by nickel chromatography according to the manufacturer’s protocol (Invitrogen Life Technologies).
Immunofluorescence and confocal microscopy
Human neutrophils (2 × 105) in 1 ml of HBSS+/0.25% HSA/20 mM HEPES were plated on sterilized coverslips in a 12-well plate and incubated at 37°C for 1 h. Cells were washed twice for 5 min each time with PBS and fixed in 3.7% formaldehyde/PBS for 10 min at room temperature. Cells were washed twice more with PBS and incubated overnight at 4°C before staining. For permeabilization, cells were incubated with 0.1% Triton/PBS for 5 min. The immunostaining procedure performed was previously described (16). Briefly, cells (permeabilized and nonpermeabilized) were incubated with 2% BSA/2% normal donkey serum in PBS (blocking solution) for 1 h at room temperature. Next, cells were labeled with either an anti-CD93 rabbit polyclonal Ab (1157) or a rabbit IgG control Ab at 5 μg/ml in blocking solution (1 h at room temperature), followed by Cy3-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch Laboratories; 1/500 in blocking solution; 1 h at room temperature). Coverslips were mounted with Vectashield (Vector Laboratories). Neutrophils were visualized using a Zeiss LSM 510 META confocal microscope.
CD93 is shed from the surface of PDBu-stimulated monocytes
To determine whether CD93 was susceptible to ectodomain cleavage, human monocytes were treated with increasing concentrations of the phorbol ester PDBu, a known potent inducer of shedding (25). A decrease in cell surface CD93 was detected by flow cytometry with increasing concentrations of PDBu, with a maximum decrease in surface expression at 50 nM PDBu (Fig. 1,A). To assess the kinetics of the PDBu-induced decrease in surface expression, monocytes were treated with 50 nM PDBu, and surface expression was assessed at 2, 5, 10, 20, and 60 min. The decrease in cell surface expression was significant at 20 min after addition of PDBu (p = 0.02, by paired Student’s t test) and was also enhanced at 60 min (80.6% shedding; n = 3; Fig. 1,B). To determine whether sCD93 was present in medium from PDBu-stimulated monocytes, medium was analyzed in a sandwich ELISA using anti-CD93 mAbs R139 and R3. The sCD93 was detected at higher levels (6- to 17-fold) in medium from PDBu-stimulated cells compared with medium from nonstimulated cells. This activity in the ELISA was depleted when the medium was absorbed with anti-CD93 (R139), but not with the isotype control Ab (IgG2b), verifying that the ELISA was specific (Fig. 1,C; n = 2). Full-length CD93 expression decreased over time after stimulation with PDBu, whereas a gradual increase in smaller cell-associated fragments containing the CD93 cytoplasmic tail was detected by Western blot (Fig. 1,D), indicative of proteolytic processing of CD93. The presence of a soluble form of CD93 in tissue culture supernatants was confirmed by immunoprecipitation of CD93 (using R139) from monocyte supernatants after treatment with 50 nM PDBu. A 75,000 Mr band and a 50,000 Mr band reactive with anti-CD93 (R3; Fig. 1, arrows) were apparent after treatment with PDBu (but not without PDBu; n = 2; Fig. 1 E).
Soluble CD93 contains both the CRD domain and the EGF repeats
Because both anti-CD93 mAbs R3 and R139 recognized the soluble fragment(s) of CD93 in the culture supernatant from PDBu-stimulated cells, recombinant proteins encoding either the CRD domain of CD93 or the CRD domain followed by the EGF repeats (CRD-EGF5; Fig. 2,A) were expressed and used to map the interaction site of the Abs to determine the nature of the soluble fragment. The proteins were expressed in HEK293T cells, and supernatants were tested by Western blot and ELISA. When a sandwich ELISA was performed using a capture Ab recognizing the V5 epitope tag found on both recombinant fusion proteins, reactivity with R3 was detected for both fusion proteins, whereas reactivity with R139 was detected only with the V5-CRD-EGF5 fusion protein, indicating that R3 recognizes the CRD domain and R139 recognizes a region within the EGF repeats (Fig. 2,B). Furthermore, after purification by nickel chromatography, both proteins were recognized with R3 by immunoblot analysis (Fig. 2,B, inset). Also, the fusion protein encoding the CRD-EGF5 displayed reactivity in an R3/R139 sandwich ELISA, whereas no reactivity by ELISA was demonstrated for the CRD alone (Fig. 2 C). These data demonstrate that R3 and R139, Abs that modulate phagocyte activity (20, 26), recognize distinct domains within CD93 and suggest that sCD93 consists of the CD93 N terminus and contains both the CRD and EGF5 repeats, because the mAbs recognize these distinct protein domains.
Soluble CD93 is detected in human plasma/serum
To determine whether sCD93 was produced in vivo, human plasma was analyzed for the presence of CD93 by ELISA. Plasma from three healthy donors contained measurable levels of sCD93, and these levels were similar among the three donors (Fig. 3,A). The activity in the ELISA was specific, because it was lost after absorbance with R139, and it was competed by the addition of R139 or R3, but not isotype control Abs (Fig. 3 B). Soluble CD93 was also detected in human serum at levels similar to those in plasma (data not shown). The concentration of sCD93 was estimated to be 0.2 nM based on the number of molecules per sample in a standard curve of U937 cell extracts that expresses ∼8.1 × 104 molecules/cell (19). The sCD93 in serum and monocyte culture medium after PDBu stimulation was not reactive with the anti-CD93 cytoplasmic tail Ab, as determined by ELISA, consistent with ectodomain proteolysis (data not shown).
CD93 shedding is mediated by a metalloproteinase, but is independent of TACE (ADAM17)
Because metalloproteinases are responsible for most of the reported ectodomain cleavage (1), 1,10-phenanthroline, a zinc-chelating compound and inhibitor of metalloproteinases, was tested for its ability to inhibit PDBu-stimulated shedding of CD93. Monocytes treated with PDBu in the presence of 1,10-phenanthroline were completely resistant to PDBu-induced ectodomain cleavage, as assessed by flow cytometry (Fig. 4, A and B). Furthermore, sCD93 could not be detected in supernatants from PDBu-stimulated monocytes in the presence of 1,10-phenanthroline (Fig. 4,C), and CD93 remained cell associated (Fig. 4,D). These data suggest that a metalloproteinase is responsible for the PDBu-induced cleavage event. TACE (ADAM17) is the sheddase responsible for the cleavage of numerous diverse transmembrane molecules (6); therefore, TACE inhibitors were used to determine whether CD93 was cleaved by TACE after PDBu stimulation. Incubation of monocytes with four different hydroxamic acid-based TACE inhibitors, TAPI-1, TAPI-2, R0-31-9790, and R0-32-7315, had no significant effect on PDBu-stimulated shedding of CD93. PDBu-stimulated cleavage of L-selectin on lymphocytes was inhibited by all TACE inhibitors, as expected, verifying that the inhibitors were functional (Fig. 4 E).
CD93 is shed from human neutrophils
Because CD93 is also expressed on human neutrophils, albeit at ∼5-fold less surface density than monocytes (27) (Fig. 4,B vs Fig. 5,C), we determined whether similar shedding properties were evident on this cell type. There was less total cell-associated CD93 detected in PDBu-stimulated human neutrophils compared with untreated control cells, as determined by ELISA (51.2–61.5% range of three experiments; Fig. 5,A) and Western blot (Fig. 5,D). Furthermore, when cell supernatants were analyzed for the presence of sCD93, 3.75- to 17-fold (range of three experiments) more sCD93 was detected in culture medium from PDBu-stimulated cells compared with unstimulated cells (Fig. 5,B), demonstrating that CD93 on neutrophils is also susceptible to PDBu-induced shedding. However, in contrast to monocytes, PDBu-stimulated neutrophils displayed an increase in cell surface CD93 expression (220 ± 69%; n = 4; Figs. 5,C and 8,C). These data indicate that CD93 is shed from PDBu-stimulated neutrophils and also suggests that CD93 may be stored in neutrophil granules, which are released in response to PDBu. In support of this hypothesis, confocal analysis of neutrophils stained with anti-CD93 (1157) demonstrated prominent intracellular staining in permeabilized cells (Fig. 5,E, lower right panel), whereas nonpermeabilized cells showed predominantly plasma membrane-associated CD93 (Fig. 5 E, upper right).
Glycosylation of CD93 does not alter PDBu-induced cleavage
Previous studies showed that CD93 is heavily O-glycosylated, and that hypoglycosylation of CD93 resulted in a loss of surface CD93 (19). Glycosylation of CD93 can be regulated in the CD93-expressing CHO ldlD cell line, because growth in medium in the absence of galactose and N-acetyl galactosamine (GalNAc) is nonpermisive for glycosylation, whereas in the presence of galactose and GalNAc, glycosylation is restored. To determine whether the glycosylation state of CD93 contributed to PDBu-induced shedding, the effect of PDBu on surface CD93 expression of ldlD-CD93 cells grown in the presence or the absence of galactose and GalNAc was tested. As shown previously, in the absence of galactose and GalNAc, CD93 surface expression was markedly decreased (average, 60.6% decrease; n = 3; data not shown). However, PDBu treatment of cultured glycosylation deficient cells (without sugars) resulted in similar decreases in relative CD93 expression (51.3 and 44.3% in the presence or the absence of sugars, respectively; n = 3; Fig. 6) as that in control cells. These data verify that glycosylation of CD93 affects cell surface expression/stability, but the PDBu-induced shedding event is independent of and distinct from the instability induced by hypoglycosylation.
Cross-linking CD93 with immobilized Ab induces shedding
To mimic potential ligation of CD93, monocytes were cultured on anti-CD93 Abs (or isotype control Abs), and cell-associated CD93 was measured by ELISA. After 60 min of adherence to anti CD93, cell-associated CD93 was decreased by 75.0 ± 13% on R139 compared with IgG2b (n = 3), and by 94.1% on R3 compared with IgM (n = 1) (Fig. 7,A); it remained depressed for at least 18 h in the presence of cross-linking anti-CD93 mAb (decrease of 85.6 ± 7.6% for R139 and 72.4 ± 19.5% for R3; n = 3; data not shown). The decrease in cell-associated CD93 at 60 min was partially inhibited with 1,10-phenanthroline (19.2 and 29.8% from two separate experiments; data not shown) consistent with ectodomain shedding. Analysis of cell lysates by Western blot demonstrated generation of cell-associated CD93 cytoplasmic tail-containing cleavage products in response to Ab cross-linking (Fig. 7,B), verifying that cross-linking stimulated proteolytic cleavage of CD93. However, sCD93 was not detected in supernatants from these adherent cells, even in response to PDBu. This is probably due to the 8- to 10-fold fewer cells per milliliter in the adhesion assay in this study compared with the cell suspension assay (Figs. 1, 4, and 8), resulting in sCD93 levels in supernatants that were below the range of sensitivity of the ELISA.
Inflammatory stimuli trigger CD93 shedding from monocytes and alter surface expression on neutrophils
Because the CD93 expression profile (monocytes, neutrophils, and endothelial cells) and its shedding by cell activation are consistent with a role for CD93 in leukocyte extravasation/inflammation, we investigated the influence of inflammatory mediators on CD93 shedding. Human monocytes were treated with the proinflammatory cytokine TNF-α or with the bacterial cell wall component LPS. After 60 min, TNF-α and LPS stimulated shedding of CD93 from monocytes, as assessed by the presence of sCD93 in culture medium (Fig. 8,A), although the amount of shed CD93 was 10–25% of that seen with PDBu stimulation. FACS analysis demonstrated that CD93 surface expression after stimulation with TNF-α and LPS was not significantly altered (Fig. 8 B), consistent with the lower degree of shed receptor and suggesting a modulated response to alternative stimuli.
In contrast to monocyte CD93 surface expression and consistent with the up-regulation of CD93 expression upon PDBu treatment of neutrophils, TNF-α as well as the inflammatory mediators, C5a and platelet-activating factor (PAF), induced a rapid increase (within 15 min) in CD93 expression on human neutrophils (Fig. 8,C). LPS also induced an increase in CD93 expression on neutrophils, but with a slower kinetic (by 30 min). In contrast, these stimuli triggered a decrease in L-selectin expression, suggesting that CD93 and L-selectin may have opposing functions in neutrophil homing (Fig. 8 D).
This study demonstrates that CD93 on human monocytes and neutrophils is susceptible to protein ectodomain cleavage. The phorbol ester, PDBu, a known inducer of ectodomain cleavage, triggered CD93 cleavage in a time- and dose-dependent manner (Fig. 1). Soluble CD93 was detected in tissue culture supernatants after exposure to PDBu, and two anti-CD93-reactive species were identified by Western blot at 75 and 50 kDa in monocyte supernatants (Fig. 1,E). Soluble CD93 was detected in human plasma from healthy donors (Fig. 3), consistent with a physiologic role for sCD93, and the CD93 intracellular domain was detected in cells after shedding. Cleavage of CD93 appears to be mediated by a metalloproteinase, because it was inhibited by the metalloproteinase inhibitor 1,10-phenanthroline (Fig. 4). Cross-linking CD93 with two immobilized mAbs recognizing different extracellular domains of CD93 resulted in the release of CD93 (Fig. 7), suggesting that ligand binding may induce CD93 shedding. However, this decrease in full-length CD93 expression was only partially inhibited by 1,10-phenanthroline, and the ICD-containing cleavage fragments were not as readily detected after Ab cross-linking compared with PDBu stimulation, suggesting that cross-linking may also stimulate other forms of CD93 degradation insensitive to the metalloprotease inhibitor. Incubation with the proinflammatory cytokine, TNF-α, or the bacterial cell wall component, LPS, also induced CD93 shedding on monocytes, suggesting that sCD93 may be released at sites of inflammation in vivo (Fig. 8).
Although CD93 was originally described as a C1q receptor that mediated enhanced phagocytosis, recent studies demonstrate that CD93 does not bind to C1q (12) and that CD93 is not required for the C1q-mediated enhanced phagocytic activity (10). McGreal et al. (28) have postulated that CD93, similar to platelet-endothelial cell adhesion molecule 1, is involved in leukocyte homing based on a shared tissue distribution pattern with platelet-endothelial cell adhesion molecule 1 and because a CD93 extracellular domain Fc fusion protein bound to activated endothelium, suggesting that there is a CD93 ligand expressed on the endothelium (12). Therefore, generation of sCD93 induced by proinflammatory mediators (e.g., TNF-α or LPS) or rapid alterations in CD93 expression on activated neutrophils may alter the homing properties of monocytes and/or neutrophils at sites of inflammation. Interestingly, TNF-α, LPS, PAF, and C5a all induced a rapid and nearly complete loss of surface L-selectin expression on neutrophils (Fig. 8,D) while inducing an elevated level of CD93 (Fig. 8,C), suggesting that these two molecules may have opposing functions in neutrophil migration. Furthermore, the enzyme(s) responsible for cleavage of CD93 differs from the L-selectin sheddase (TACE/ADAM17; Fig. 4 D). Shedding independent of TACE/ADAM17 was also reported for CD44 (8), an adhesion molecule with some similarities in structure and function to CD93 (29).
One consequence of ectodomain cleavage is the subsequent regulated cleavage of the transmembrane domain, resulting in the liberation of the ICD into the cytoplasm. Although some ICD fragments are susceptible to proteosome-mediated degradation, others have been shown to act as transcription factors or to act in complex with transcription factors to alter gene expression. Interestingly, the cytoplasmic domain of CD93 contains a highly charged juxtamembrane domain, predicted to contain a nuclear localization signal (30), suggesting that the CD93-ICD may be involved in gene regulation. Furthermore, CD93-ICD fragments are readily detected in human monocyte cell lysates after PDBu treatment or Ab cross-linking (Figs. 1,D and 7 B), implying a relative lack of susceptibility to proteosome-mediated degradation.
The soluble fragment of the CD93 extracellular domain may have biological activity, because numerous ectodomain cleavage products are biologically active (e.g., TNF-α, TGF-α, TGF-β, and members of the EGF family). Interestingly, Botto et al. (10) showed that CD93-deficient mice were defective in the clearance of apoptotic cells from the inflamed peritoneum; however, the mechanism remains elusive. Perhaps sCD93 is a bridging molecule, responsible for linking the apoptotic cell to the phagocyte, similar to C1q, protein S, thrombospondin, milk fat globule-EGF factor 8 protein, and others (31), and is the component responsible for enhanced clearance of apoptotic cells under inflammatory conditions, when sCD93 would be expected to be present.
The identification of the susceptibility of CD93 to ectodomain cleavage under a variety of conditions and the observation that the soluble fragment of CD93 exists in plasma from normal healthy donors open numerous avenues to study the functional significance of CD93 and the mechanism by which CD93 modulates phagocytosis and adhesion. Although not addressed in this study, CD93 is widely expressed on endothelium, so it will be important to determine whether endothelial CD93 is susceptible to ectodomain cleavage and to characterize the resulting changes in surface expression of CD93 after activation of endothelium. The observations presented in this study warrant further investigation regarding the biological activities of the soluble fragment(s) of CD93 found in plasma, and likely at foci of inflammation, as well as investigations to elucidate the biological activity of the CD93 ICD after ectodomain cleavage.
We are grateful to Ozkan Yazan for excellent technical assistance, and Meiying Zhou for help with the initial shedding experiments. We thank the staff of the University of California-Irvine General Clinical Research Center for help with obtaining human blood for monocyte purification.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by National Institutes of Health Grant AI41090 (to A.J.T.). Support for obtaining human blood products used in this study was provided in part by Public Health Service Research Grant M01RR00827 from the National Center for Research Resources.
Abbreviations used in this paper: EGF, epidermal growth factor; GalNAc, N-acetyl galactosamine; HSA, human serum albumin; ICD, intracellular domain; PAF, platelet-activating factor; PDBu, phorbol dibutyrate; TAPI, TNF-α protease inhibitor; PVDF, polyvinylidene diflouride; sCD93, soluble CD93; TACE, TNF-α-converting enzyme; ADAM, a disintegrin and metalloproteinase.