The ectodomains of many proteins located at the cell surface are shed upon cell stimulation. One such protein is the heparin‐binding EGF‐like growth factor (HB‐EGF) that exists in a membrane‐anchored form which is converted to a soluble form upon cell stimulation with TPA, an activator of protein kinase C (PKC). We show that PKCδ binds in vivo and in vitro to the cytoplasmic domain of MDC9/meltrin‐γ/ADAM9, a member of the metalloprotease–disintegrin family. Furthermore, the presence of constitutively active PKCδ or MDC9 results in the shedding of the ectodomain of proHB‐EGF, whereas MDC9 mutants lacking the metalloprotease domain, as well as kinase‐negative PKCδ, suppress the TPA‐induced shedding of the ectodomain. These results suggest that MDC9 and PKCδ are involved in the stimulus‐coupled shedding of the proHB‐EGF ectodomain.
The extracellular domain of a number of membrane proteins can be cleaved proteolytically causing release into the medium (Massague and Pandiella, 1993; Rose‐John and Heinrich, 1994). This proteolytic processing, also referred to as ‘ectodomain shedding’, is observed in growth factors (Massague and Pandiella, 1993), growth‐factor receptors (Vecchi et al., 1996), cell‐adhesion molecules (Feehan et al., 1996), extracellular matrix proteins (Subramanian et al., 1997; Werb, 1997) and other membrane proteins such as the β‐amyloid precursor protein (Esch et al., 1990). Ectodomain shedding of membrane proteins changes the fate, location and mode of action; thus it affects the biological activities of membrane proteins. For example, the ectodomain shedding of TNF‐α (Mohler et al., 1994), FAS‐ligand (Tanaka et al., 1998) and the homing receptor L‐selectin (Feehan et al., 1996) has been implicated in inflammation, whereas the ectodomain shedding of βAPP (Haass and Selkoe, 1993) has been implicated in the pathogenesis of Alzheimer's disease (Sisodia and Price, 1995). The ectodomain shedding of membrane receptors results in the generation of soluble competitors for their own ligands (Fernandez‐Botran, 1991). In the case of membrane‐anchored growth factors, ectodomain shedding can convert them into diffusible factors and greatly influence their functions (Massague and Pandiella, 1993). On the one hand, for example, the membrane‐anchored form of spitz, a TGF‐α‐like molecule in Drosophila, is inactive biologically and is cleaved proteolytically to yield a soluble active form (Golembo et al., 1996). On the other hand, the membrane‐anchored forms of c‐kit ligand (Flanagan et al., 1991) and ephrins (ligands of EPH tyrosine kinase receptor, Davis et al., 1994) are fully functional whereas the soluble forms exhibit little or no biological activity. Thus, ectodomain shedding is an important regulatory step in the function of membrane proteins involved in cell–cell communication in development, cell differentiation and tissue maintenance.
Heparin‐binding EGF‐like growth factor (HB‐EGF) is a member of the epidermal growth factor (EGF) family (Higashiyama et al., 1991), which encompasses a number of structurally homologous mitogens including EGF, TGF‐α, vaccinia virus growth factor (Carpenter and Wahl, 1990), amphiregulin (Shoyab et al., 1989), β‐cellulin (Shing et al., 1993) and epiregulin (Toyoda et al., 1995). Like EGF, TGF‐α and amphiregulin, HB‐EGF binds to and stimulates the phosphorylation of the EGF receptor. HB‐EGF is synthesized as a membrane‐anchored precursor protein of 208 amino acids composed of signal peptide, heparin‐binding, EGF‐like, transmembrane and cytoplasmic domains (Higashiyama et al., 1991). Although the membrane‐anchored form of HB‐EGF (proHB‐EGF) is cleaved on the cell surface to yield a soluble growth factor of 75–86 amino acids, a considerable amount of proHB‐EGF remains uncleaved on the cell surface (Goishi et al., 1995). Importantly, proHB‐EGF is not only a precursor of the soluble form but is also biologically active in itself; proHB‐EGF forms a complex with both CD9 (Mitamura et al., 1992; Iwamoto et al., 1994) and integrin α3β1 (Nakamura et al., 1995), both localized at cell–cell attachment sites (Nakamura et al., 1995), and transduces biological signals in a nondiffusible manner to neighboring cells (Higashiyama et al., 1995) as has been shown for TGF‐α (Brachmann et al., 1989; Wong et al., 1989) and colony‐stimulating factor (Stein et al., 1990). Moreover, although secreted mature HB‐EGF is a potent mitogen for a number of cell types (Higashiyama et al., 1991), the membrane‐anchored form may act as a negative regulator of cell proliferation (Miyoshi et al., 1997; Takemura et al., 1997). Thus, the processing of the juxtamembrane domain of proHB‐EGF to the soluble HB‐EGF means the conversion of the mode of action of this growth factor from juxtacrine to paracrine or the switching to the opposite activity for cell growth. ProHB‐EGF also acts as the specific receptor for diphtheria toxin (DT) and mediates the endocytosis of the receptor‐bound toxin (Naglich et al., 1992; Iwamoto et al., 1994). Interestingly, proHB‐EGF is cleaved rapidly to soluble HB‐EGF by treatment with TPA, suggesting the involvement of a cellular signaling pathway involving protein kinase C (PKC) (Goishi et al., 1995). The regulated shedding of membrane proteins has also been shown for a number of other membrane proteins including βAPP, TGFα, IL‐6 and receptor‐like protein tyrosine phophatases (Pandiella and Massague, 1991a,b; Arribas et al., 1996; Aicher et al., 1997). However, the mechanism of the ectodomain shedding of membrane proteins remains totally obscure.
Metalloprotease–disintegrins, or the ADAM family of proteins, make up a recently discovered protein family of membrane‐anchored glycoproteins comprising a pro‐domain, metalloprotease domain, a disintegrin domain, a cystein‐rich domain, an EGF‐like domain, a transmembrane domain and a cytoplasmic domain (Wolfsberg et al., 1995; Blobel, 1997). These proteins have been shown to be involved in cell–cell interactions, cell adhesion and the processing of proteins. More than 20 family members have been identified in a variety of species including mammals, Xenopus, Drosophila and Caenorhabditis elegans. The first identified proteins of the ADAM family were fertilin α and β (also called PH‐30α and β, or ADAM1 and 2), sperm surface proteins involved in fertilization by binding to unknown receptors on the egg (Wolfsberg et al., 1995; Blobel, 1997). Another ADAM protein, meltrin‐α (ADAM12), has been implicated in myoblast fusion or myogenesis (Yagami‐Hiromasa et al., 1995). The metalloprotease activity of ADAM family members was first revealed for TNF‐α converting enzyme (TACE), responsible for the release of TNF‐α from the cell surface by proteolytically cleaving the precursor form (Black et al., 1997; Moss et al., 1997). More recently, Kuzbanian (Kuz) in Drosophila, whose mammalian homolog is ADAM10 (Howard et al., 1996), has been suggested to be involved in the proteolytic cleavage and activation of Notch receptor (Pan and Rubin, 1997; Chan and Jan, 1998).
To understand the mechanism of the TPA‐induced shedding of the proHB‐EGF ectodomain, we first examined the role of PKC isotypes and identified PKCδ as the PKC isotype involved in the regulated shedding. Interaction screening of a cDNA library using purified PKCδ as a probe allowed us to identify MDC9 as a specific binding protein for PKCδ. The overexpression of MDC9 resulted in the shedding of proHB‐EGF without TPA, and the MDC9 mutants of the metalloprotease domain inhibited TPA‐induced shedding. These results provide evidence to suggest that MDC9 is involved in the processing of proHB‐EGF and that a direct interaction between PKCδ and MDC9 is involved in the regulated shedding.
Involvement of PKCδ in the TPA‐induced shedding of the ectodomain of proHB‐EGF
Previous studies on African green monkey kidney Vero cells overexpressing human proHB‐EGF (Vero‐H cells) have shown that membrane‐bound proHB‐EGF is processed by a phorbol ester, TPA‐dependent mechanism, to yield soluble bioactive HB‐EGF (Goishi et al., 1995). Treatment of Vero‐H cells with TPA results in a reduction in the cell‐surface DT receptor and cell‐surface immunoreactivity against anti‐HB‐EGF antibody, #H6, directed against the ectodomain of proHB‐EGF. In parallel with the loss of cell surface immunoreactivity, the amounts of soluble 14 kDa and 19 kDa HB‐EGF detected by the anti‐HB‐EGF antibody #2998 increase in the culture medium. Concomitantly, TPA induces the loss of proHB‐EGF juxtacrine activity and an increase in soluble HB‐EGF paracrine activity (Goishi et al., 1995). Importantly, these responses occur within 30 min (Goishi et al., 1995; Figure 1A and B). Furthermore, staurosporine and H7 inhibit both the TPA‐induced and constitutive shedding of the pro‐HB‐EGF ectodomain (Goishi et al., 1995). These results suggest the presence of a protease acting on the cell surface whose activity is regulated by an unknown mechanism through a TPA‐responsive protein kinase such as PKC.
As a first step in analyzing the mechanism of the shedding of the proHB‐EGF ectodomain, we confirmed the above observation by using immunofluorescence microscopy with #H6, an antibody against the ectodomain of HB‐EGF. Vero‐H cells were first treated with anti‐proHB‐EGF antibody #H6 directed against the ectodomain of proHB‐EGF, and fixed and visualized using a second antibody, revealing the presence of proHB‐EGF immunoreactivity on the outer surface of Vero‐H cells. As shown in Figure 1A, staining is concentrated at the cell–cell contact region in addition to the entire cell surface, confirming the previous results (Goishi et al., 1995). The addition of TPA results in the disappearance of the immunoreactivity on the cell surface within 30 min, consistent with the previous report (Figure 1A and B).
In order to obtain direct proof for the involvement of PKC and to obtain information about the PKC isotypes involved in the shedding of proHB‐EGF, we next examined the PKC isotypes expressed in Vero‐H cells. As shown in Figure 1C, Vero‐H cells express at least three TPA‐responsive PKC isotypes, PKCα, PKCδ and PKCε. Treatment of the cells with TPA for 12 h results in decreases in the amounts of both PKCδ and PKCε, whereas the amount of PKCα does not change. Although the meaning of this selective down‐regulation remains unknown, the results suggest that at least two PKC isotypes, PKCδ and PKCε, are activated in Vero‐H cells upon TPA treatment.
In order to clarify the involvement of PKC, we expressed the active mutant of PKC isotypes and examined cell surface immunoreactivity against the ectodomain of proHB‐EGF by indirect immunofluorescence. Figure 1D shows examples where Vero‐H cells expressing the activated forms of tag‐PKC isotypes are visualized using an anti‐tag (T7) antibody by indirect immunofluorescence (green). At the same time, the proHB‐EGF ectodomain (red) was also visualized as shown in Figure 1A. These PKC mutants show co‐factor independent kinase activities in vitro and their expression mimics the effect of TPA in vivo (Ueda et al., 1996). Interestingly, cells that overexpress active PKCδ (tag‐DR144/145A) but not active mutants of PKCα (tag‐αR22A/A25E) and PKCε (tag‐εK155A/R156A/A159E) show a loss of proHB‐EGF ectodomain staining. Figure 1E shows the quantified results. The PKC mutants used in the present study contain an N‐terminal tag sequence, permitting direct comparison of the amounts of PKC mutants expressed in the respective cells. Clearly, the results indicate that the ectopic expression of active PKCδ results in proHB‐EGF ectodomain shedding in the absence of TPA.
Next we examined the effect of the ectopic expression of a PKCδ kinase‐knockout mutant (DRKA) containing a point mutation at the ATP‐binding site that results in greatly diminished kinase activity and the inhibition of PKCδ‐dependent AP1 activation (Hirai et al., 1994). To evaluate the expression levels of the respective PKCδ and PKCδ mutant, we monitored cells expressing PKCδ mutants using an anti‐PKCδ antibody. As shown in Figure 2A and B, the expression of DRKA results in the inhibition of TPA‐induced proHB‐EGF ectodomain shedding, whereas the expression of wild‐type PKCδ has no effect in the absence of TPA, indicating that the kinase activity of PKCδ is required for the shedding of the proHB‐EGF ectodomain. Further, these results support the notion that PKCδ is involved in this process.
Identification of MDC9 as a PKCδ‐specific binding protein/substrate
To analyze the mechanism of the involvement of PKCδ in the shedding of the proHB‐EGF ectodomain, we searched for binding proteins specific for PKCδ. Our strategy was to screen a mouse fibroblast NIH 3T3 cDNA expression library using purified recombinant autophosphorylated PKCδ as a probe (Izumi et al., 1997). Secondary screening of the cDNA clones for putative PKCδ‐binding proteins using other PKC isotypes as probes revealed that most of the cDNA clones encode proteins that bind to multiple PKC isotypes (data not shown). These proteins include MARCKS, a known physiological substrate for PKC, and SRBC, a protein whose mRNA is induced upon serum depletion, both of which bind to PKCα and PKCε as well as to PKCδ in vitro (Fujise, et al., 1994; Izumi et al., 1997; data not shown). Among >10 such cDNA clones, clone G1 encoded a protein that showed binding specificity to PKCδ on blot overlay assay (data not shown).
A search of the sequence database revealed that the G1 sequence encodes the C‐terminal 127 amino acid residues of MDC9/meltrin‐γ/ADAM9. MDC9 is a member of the ADAM family of proteins that share disintegrin and metalloprotease domains in addition to other structural motifs (Figure 3A) (Yagami‐Hiromasa et al., 1995; Weskamp et al., 1996). The cloned cDNA encodes the exact C‐terminal sequence implicated as the cytoplasmic domain. Considering the structural motif of the proposed extracellular domain of MDC9 and the recent finding that a protein responsible for the processing of TNFα, TACE, is a member of the ADAM family, the present finding of a direct protein–protein interaction between PKCδ and MDC9 suggests a possible functional link between PKCδ and MDC9, and further suggests their involvement in the regulated shedding of the proHB‐EGF ectodomain. To examine these possibilities, we next isolated a full‐length cDNA for MDC9 and raised an antibody against the C‐terminal domain of MDC9.
The initial characterization of the MDC9 mRNA and proteins revealed that MDC9 is a ubiquitously expressed membrane‐spanning protein with a C‐terminal cytoplasmic domain containing two SH3‐binding motifs (Weskamp et al, 1996). The surface labeling of NIH 3T3 cells also revealed the presence of MDC9 at the cell surface (Weskamp et al, 1996). As shown in Figure 3B, a full‐length cDNA for MDC9 introduced into COS cells directed the appearance of two protein bands migrating at ∼84 kDa and 120 kDa. The two bands are detected in various cell lines including Vero‐H cells and COS cells (Figure 3B), consistent with the earlier observation (Weskamp et al, 1996).
To confirm the interaction between MDC9 and PKCδ in vivo, we expressed the tag‐MDC9 and PKC isotypes in COS cells, immunoprecipitated MDC9, and examined the immunoprecipitates for the presence of PKC isotypes. Figure 3C shows that PKCδ but not PKCα or PKCε, co‐precipitates with tag‐MDC9, indicating that the cytoplasmic domain of MDC9 specifically associates with PKCδ in vivo. Similar experiments using PKCα/δ and PKCδ/α chimeras, AD335 and DA318, revealed that MDC9 associates with AD335 but not with DA318. AD335 contains the regulatory domain of PKCα linked to the kinase domain of PKCδ, suggesting that the kinase domain of PKCδ is responsible for the interaction with MDC9.
To determine the MDC9 sequence required for the interaction with PKCδ, we constructed a series of glutathione S‐transferase (GST) fusion proteins containing the cytoplasmic domain of MDC9 as illustrated in Figure 3A. Figure 3D shows the results of a blot overlay assay where GST fusion proteins were fixed to the membrane and probed with autophosphorylated PKCδ, showing that PKCδ binds directly to the cytoplasmic domain of MDC9. Further, except for GST–MDC9‐743/827, all the C‐terminal deletion mutants associated with PKCδ, indicating that a very short region (amino acid residues 719–745) of MDC9 is sufficient for the interaction with PKCδ. Interestingly, this shortest mutant also interacts with PKCα on blot overlay assay, whereas the whole cytoplasmic domain fails to associate efficiently with PKCα (data not shown). Thus, the 25 amino acid sequence just downstream of the transmembrane domain contains the PKC‐binding sequence, whereas the sequence that determines the specificity for PKCδ is not included in this region. This region contains one threonine and two serine resides surrounded by basic residues (Figure 3A), a consensus sequence for PKC phosphorylation. In fact, GST–MDC9 fusion proteins are efficiently phosphorylated by purified PKCδ as shown in Figure 3E. These results clearly show that the cytoplasmic domain of MDC9 is the specific binding protein and substrate for PKCδ in vitro.
Overexpression of MDC9 results in the processing of proHB‐EGF and the release of bioactive soluble HB‐EGF
To test the possibility that MDC9 is involved in the processing of proHB‐EGF, we first examined the effect of the ectopic expression of MDC9 on Vero‐H cells. Forty‐eight hours after transfection with plasmids encoding full‐length mouse MDC9 cDNA, the proHB‐EGF ectodomain was visualized. As shown in Figure 4C, a and b, in cells expressing MDC9 cDNA (green), HB‐EGF ectodomain staining (red) is markedly reduced even in the absence of TPA treatment. The control transfection with the expression vector (pEF‐BOS) gave a result similar to those obtained with nontransfected cells in the presence or absence of TPA (data not shown). The percentage of cells expressing the proHB‐EGF ectodomain was quantitated and is summarized in Figure 4D. In the absence of TPA, all cells express the proHB‐EGF ectodomain, whereas the overexpression of MDC9 reduced the expression to 35%. Thus, the ectopic overexpression of MDC9 results in the processing of proHB‐EGF in the absence of TPA.
To obtain further biochemical proof that the overexpression of MDC9 results in the shedding of the proHB‐EGF ectodomain, we transfected Vero cells with MDC9 cDNA and isolated stable transformants expressing MDC9. The isolated clone, Vero‐MDC9, expresses ∼10 times more MDC9 than parental Vero cells but much less MDC9 than transiently expressing cells (data not shown). The effect of MDC9 on the shedding of the proHB‐EGF ectodomain was then examined in these cells. We measured both the amount of proHB‐EGF on the cell surface and the amount of HB‐EGF secreted into the medium by measuring DT‐binding activity. The amount of proHB‐EGF on the cell surface of Vero–MDC9 was ∼60% that of parental Vero cells under steady state culture conditions (Figure 5A), whereas the amount of HB‐EGF secreted into the medium from Vero–MDC9 was 2.5–3 times higher than from Vero cells (Figure 5B). When the cells were treated with TPA, most of the proHB‐EGF on the cell surface was processed rapidly (Figure 5A) and secreted into the medium (Figure 5B). Western blotting using an anti‐HB‐EGF neutralizing antibody showed that three species of soluble HB‐EGF with molecular masses of ∼14 kDa, 18 kDa and 19 kDa are secreted into the conditioned medium from Vero and Vero‐MDC9 cells (Figure 5C). That is similar to the HB‐EGF forms secreted from Vero‐H cells as reported previously (Goishi et al., 1995).
Finally, we tested whether the overexpression of MDC9 results in the release of mitogenically active soluble HB‐EGF. To examine this, Vero‐H cells were transfected with plasmids encoding MDC9 cDNA or with control vector plasmids and the mitogenic activity of the conditioned media was determined by measuring the rate of DNA synthesis of DER cells as described in Materials and methods. DER cells, similar to EP170.7 cells (Pierce et al., 1988), are a stable transformant of 32D cells expressing EGF receptor, and thus respond only to EGF‐receptor ligands in the absence of IL‐3. Although DER cells allow the efficient detection of the mitogenic activity of HB‐EGF, they also respond to other EGF‐receptor ligands. Vero‐H cells might secrete other mitogenic factors in addition to HB‐EGF. To measure the mitogenic activity attributable to HB‐EGF, we used CRM197 as a specific inhibitor of the mitogenic activity of HB‐EGF (Mitamura et al., 1995) and determined the CRM197‐inhibitable mitogenic activity. As shown in Figure 5D, the conditioned medium of MDC9‐transfected Vero‐H cells contains 2–3 times more mitogenic activity than that of vector‐transfected Vero‐H cells. The mitogenic activity attributed to other growth factors was <20% in each experiment. The results, together with those for the transfection of MDC9 to Vero‐H cells, led us to conclude that the ectopic expression of MDC9 results in marked enhancement of proHB‐EGF processing to yield biologically active soluble HB‐EGF.
Mutations in the metalloprotease domain result in the generation of dominant‐negative MDC9 mutants that suppress the TPA‐induced processing of proHB‐EGF
The above results suggest the involvement of both PKCδ and MDC9 in the regulated shedding of proHB‐EGF. To investigate whether the MDC9 protease activity is necessary for the processing of proHB‐EGF, several mutants of MDC9 were constructed (Figure 4A).
The expressions of these mutants were confirmed in COS cells (Figure 4B). When compared with endogenous MDC9 (Figure 3B), the overexpressed wild‐type MDC9 shows a slightly more intensely stained 120 kDa band and a less intense 84 kDa band. Although the nature of the two bands remains to be clarified in future experiments, the lower bands are absent from ΔMP and H347,351A mutants, both of which lack essential sequences in the protease domain. A similar situation occurs for ΔPKCBD, which contains an intact protease domain that is absent from the ΔPKCΔMP mutant. It is thus tempting to speculate that the lower bands correspond to the proteolytically processed form and that the protease activity of MDC9 is involved in the processing. In addition to the difference in the lower bands, the upper band of ΔPKCBD and the corresponding bands in ΔMP and ΔPKCΔMP show doublets.
Next we examined the effect of the overexpression of these mutants on the ectodomain shedding of proHB‐EGF in Vero‐H cells. As shown in Figure 4C, none of the mutants examined, except ΔPKCBD, has any significant effect on ectodomain shedding in the absence of TPA, in clear contrast to the results for wild‐type MDC9 (Figure 4C, a and b) and ΔPKCBD (Figure 4C, c and d). Importantly, the TPA‐induced shedding of proHB‐EGF is inhibited by the overexpression of ΔMP (Figure 4C, g and h). H347,351A also suppresses TPA‐induced shedding in Vero‐H cells (Figure 4C, o and p). In order to compare the effect of mutants more quantitatively, the percentage of proHB‐EGF‐positive cells was determined, and results are summarized in Figure 4D. These results indicate that ectodomain shedding caused by exogenous MDC9 requires the protease activity of MDC9 and that the deletion of the PKC‐binding domain does not affect ectodomain shedding. The dominant‐negative effects of MDC9 mutants lacking essential sequences at the metalloprotease domain on TPA‐induced shedding support the notion that endogenous MDC9 or its relative is involved in the shedding of the proHB‐EGF ectodomain. The ΔPKCΔMP mutant, which lacks both the MP‐ and PKC‐binding domains, has dominant negative effects very similar to those of the ΔMP and H347,351A mutants (Figure 4C, k and l); the dominant negative effects of ΔMP are not caused by the titration of PKCδ.
Overexpression of the MDC9 cytoplasmic domain (Cyto) also has a dominant‐negative effect on the TPA‐induced shedding of the proHB‐EGF ectodomain (Figure 4C, s and t, and Figure 4D). This effect was not observed for the CytoΔPKCBD mutant, which lacks the PKC‐binding domain (Figure 4C w and x, and Figure 4D), suggesting the involvement of PKC‐BD in this effect. The simplest explanation is that the PKC‐BD titrates the cellular PKC required for TPA‐induced ectodomain shedding, consistent with notion that the TPA‐induced ectodomain shedding requires endogenous PKCδ.
Another MDC9 mutant, ΔPro‐rich, which lacks the cytoplasmic proline‐rich domain, has a moderate effect on ectodomain shedding in the absence of TPA (Figure 4D). Considering the results for ΔPKCBD, which shows effects very similar to wild‐type MDC9, the results with ΔPro suggest the importance of the proline‐rich sequence for the shedding of the proHB‐EGF ectodomain.
Regulated shedding of the membrane ectodomain
Proteolytic processing of the extracellular domain of a variety of membrane proteins seems to be a rather general strategy for regulating the capacity of cells to respond to extracellular stimuli. However, our knowledge of the molecules involved in this process is quite limited. Although there is no apparent sequence similarity in the cleavage sites, the presence of mutant cell lines defective for the shedding of at least two unrelated molecules, βAPP and proTGF‐α, suggests that ectodomain shedding of these two molecules may share a common component (Arribas and Massague, 1995). Recent findings on the processing of TNF‐α and Notch have revealed that members of the ADAM family of proteases are involved in proteolytic processing (Howard et al., 1996; Black et al., 1997; Moss et al., 1997; Pan and Rubin, 1997). However, the mechanism for their regulation remains unknown. Ectodomain shedding is observed constitutively under normal culture conditions, but can often be stimulated by activators of protein kinase C such as TPA. Thus, a PKC‐dependent pathway is thought to be involved in ectodomain shedding (Pandiella and Massague, 1991a; Goishi et al., 1995).
In the present study, we provide evidence that the regulated processing of proHB‐EGF requires PKCδ and MDC9, a member of the ADAM family of metalloproteases, and that it involves a direct interaction between PKCδ and the cytoplasmic domain of MDC9. Overexpression of the constitutive active mutants of PKCδ, but not those of PKCα or PKCε, cause the shedding of the proHB‐EGF ectodomain in the absence of TPA in Vero‐H cells overproducing human proHB‐EGF. Furthermore, the TPA‐induced ectodomain shedding of proHB‐EGF is suppressed by the overexpression of a kinase knockout mutant of PKCδ. These results support the notion that PKCδ is specifically involved in the TPA‐induced shedding of the proHB‐EGF ectodomain in Vero‐H cells. To analyze the mechanism further, we searched for PKCδ‐binding proteins and found MDC9 as a PKCδ‐specific binding protein. PKCδ, but not PKCα or PKCε, interacts with the whole cytoplasmic domain of MDC9 in COS cells. The interaction between PKCδ and MDC9 requires the kinase domain of PKCδ and a 25 amino acid cytoplasmic sequence of MDC9 is sufficient for the direct interaction with PKCδ in vitro. Importantly, overexpression of MDC9 results in the shedding of the proHB‐EGF ectodomain in Vero‐H cells. MDC9‐dependent shedding of the proHB‐EGF ectodomain, as judged by the disappearance of the ectodomain epitope or DT‐binding activity from the cell surface, parallels the appearance of DT‐binding activity and growth‐stimulating activity in the culture medium, confirming the occurrence of physiological processing. Furthermore, MDC9 mutants lacking the protease domain, and those with point mutations in the protease domain, not only fail to induce ectodomain shedding but also suppress TPA‐induced shedding, supporting the notion that endogenous MDC9 or related members of the ADAM family are involved in the process. These results are consistent with the hypothesis that MDC9 and its protease activity are involved in TPA‐induced ectodomain shedding of proHB‐EGF in Vero‐H cells.
Proteases required for the shedding of the proHB‐EGF ectodomain
Earlier studies have suggested that a metalloprotease(s) is involved in the TPA‐induced processing of proHB‐EGF (Lanzrein et al., 1995). In Vero‐H cells, a hydroxamic‐acid‐based metalloprotease inhibitor, KB‐R8301, inhibits the constitutive and TPA‐induced processing of proHB‐EGF with an 50% inhibitory concentration (IC50) of ∼500 nM (data not shown). Since KB‐R8301 and other hydroxamic‐acid‐based metalloprotease inhibitors have IC50 values of ∼1–10 nM for most MMPs (Moss et al., 1997; Yamamoto et al., 1998), the major enzyme involved in the ectodomain shedding of proHB‐EGF in Vero and Vero‐H cells is not a known member of the MMP family. We also observed that tissue inhibitors of metalloprotease‐1 (TIMP‐1) do not inhibit the ectodomain shedding of proHB‐EGF in Vero‐H cells even at concentrations of 10 μg/ml (data not shown), supporting the notion that MMP members are not involved in the process.
In the present study, we show that MDC9 is involved in the TPA‐induced shedding of the proHB‐EGF ectodomain. Furthermore, mutants with deletions in the metalloprotease domain and point mutations in the zinc‐binding motif strongly suggest the requirement of protease activity for the processing of proHB‐EGF. However, we have not succeeded in demonstrating the actual proteolytic activity of MDC9 under cell‐free conditions in which the soluble form of MDC9 secreted from COS cells is incubated with either solubilized proHB‐EGF or with a synthetic peptide containing the juxtamembrane sequence of proHB‐EGF. Thus, it remains unclear whether MDC9 cleaves proHB‐EGF directly. As shown in blood‐clotting systems or in the caspase family system in programmed cell death, related proteases often constitute a protease‐cascade reaction. It is, therefore, conceivable that ectodomain shedding is also regulated by such a protease‐cascade reaction. If this is the case, the direct interaction with PKCδ supports the notion that MDC9 might be the initial protease in TPA‐induced ectodomain shedding of proHB‐EGF.
It will be important to determine the cleavage site of HB‐EGF, which is released from cells that overexpress MDC9. The cleavage site of proHB‐EGF for ectodomain shedding in vivo is still uncertain. Previous results, shown by amino‐acid‐composition analysis of C‐terminal chymotryptic peptide of 19 kDa HB‐EGF released by TPA treatment of Vero‐H cells, suggested that the cleavage site would be at Pro149–Val150 (Goishi et al., 1995). However, it is also possible that after cleavage at C‐terminal position by primary enzyme, additional enzymes may further process secreted HB‐EGF by removing Val150 as discussed previously (Suzuki et al., 1997). In any case, a comparison of cleavage sites of HB‐EGF released from nontransfected and MDC9‐transfected cells will be required for further corroboration of a role of MDC9 in HB‐EGF shedding.
Using an in vitro assay involving a fusion protein in which alkaline phosphatase (AP) replaced the transmembrane and cytoplasmic domains of HB‐EGF, Suzuki et al. (1997) reported that MMP‐3 is capable of cleaving proHB‐EGF at juxtamembrane domains. A more recent report involving an intact cell system suggested the presence of an ionomycin‐stimulated pathway for the shedding of the proHB‐EGF ectodomain, and that this pathway is not sensitive to PKC inhibitors or to prolonged treatment with TPA, suggesting that this pathway operates independently of the pathway involving PKC (Dethlefsen et al., 1998). Multiple distinct pathways would exist for the regulation of ectodomain shedding of proHB‐EGF, as has been suggested in the case of TGF‐α (Pandiella and Massague, 1991b). The relationship between these findings and ours remains to be clarified in future experiments.
Regulation of inside‐out signaling through PKCδ and MDC9
The sequence and domain structure of MDC9 and other ADAM family metalloproteases suggest that they are synthesized as catalytically inactive zymogens. The inactive state could be maintained through a cysteine switch mechanism and the protein converted to the active forms by unknown mechanisms (Loechel et al., 1998). In the case of meltrin‐α (ADAM12), cleavage of the prodomain at a site for furin‐like endopeptidases results in the conversion to a proteolytically active protein (Loechel et al., 1998). Most ADAMs, including MDC9, also have furin‐cleavage sites at the putative boundary between the prodomain and the metalloprotease domain. Thus, it is possible that TPA treatment may induce the cleavage of MDC9 at the furin site to generate the active form. Furin or a furin‐like enzyme has been suggested to be involved not only in N‐terminal processing of human proHB‐EGF (Nakagawa et al., 1996), but also in the cleavage and activation of proHB‐EGF‐bound diphtheria toxin (Tsuneoka et al., 1993). However, we think it unlikely that furin is involved in the TPA‐induced activation of MDC9 for the following reason. When MDC9 proteins, endogenous or overexpressed, were analyzed by SDS–PAGE and Western blotting, two bands with estimated sizes of 120 kDa and 84 kDa were seen as shown in Figure 3B, and these probably represent the proform and processed form, respectively. TPA treatment does not significantly change the pattern of MDC9 processing (data not shown). Thus, the PKC‐dependent activation of MDC9 seems to involve an as yet unidentified mechanism.
The dominant negative effect of Cyto, but not CytoΔPKCBD (Figure 4C and D), suggests that PKCBD is involved in TPA‐induced shedding. One explanation is that the endogenous PKCδ required for this process is titrated out by the overexpression of Cyto, supporting our hypothesis for the requirement of PKCδ for PKC‐induced shedding. The dominant‐negative effects of the MDC9 mutant ΔPKCBDΔMP, which lacks both PKCBD and the metalloprotease domain, suggest the involvement of an additional MDC9 sequence, other than PKCBD, in TPA‐induced shedding. Since CytoΔPKCBD shows no such effect, some sequence in the extracellular domain must be involved. Interestingly, proHB‐EGF forms a complex with integrin α3β1 (Nakamura et al., 1995). Therefore, although the role of the disintegrin domain has not been clarified in processing by ADAM family metalloproteases, it is tempting to speculate that the disintegrin domain of MDC9 plays some role, such as the recognition or binding to proHB‐EGF, in TPA‐induced shedding. Clarification of the sequence in the extracellular domain of MDC9 required for the dominant negative effect will provide further insight into the mechanism of TPA‐induced shedding.
One plausible role for PKCδ in the triggering of proHB‐EGF shedding is the recruitment of MDC9 to proHB‐EGF. If these proteins localize at different membrane domains under steady‐state conditions, the movement of MDC9 to a site that allows interaction with other proteases or with proHB‐EGF itself might be essential for shedding. The binding and subsequent phosphorylation of the cytoplasmic domain of MDC9 by PKCδ may trigger this spatial change in MDC9. This model may be able to explain why the overexpression of wild‐type MDC9, and even ΔPKCBD, results in shedding in the absence of TPA. Under overexpression conditions, MDC9 and its mutants may be missorted or mislocalized in the cells. The mislocalized MDC9 may act in the processing of the proHB‐EGF ectodomain without PKC activation. With regard to the localization and movement of MDC9, the effect of the overexpression of MDC9–ΔPro‐rich is interesting in that it has a moderate effect on ectodomain shedding in the absence of TPA, suggesting the involvement of proline‐rich domains. The cytoplasmic tail of MDC9 contains two proline‐rich sequences that can bind to the SH3 domains of Src, and may function as a SH3 ligand domain (Weskamp et al., 1996). The identification of molecules that bind to the proline‐rich domain of MDC9 might be important.
In conclusion, PKCδ and MDC9, which binds PKCδ at the cytoplasmic domain, are involved in the TPA‐induced processing of the proHB‐EGF ectodomain. A number of protein ectodomains are known to be processed in response to TPA treatment, suggesting the possible involvement of these proteins in the processing of membrane ectodomains in a variety of biological systems.
Materials and methods
12‐O‐Tetradecanoylphorbol 13‐acetate (TPA) was purchased from Nacalai Tesque Co., Ltd (Kyoto, Japan). KB‐R8301 was obtained from Kanebou Co., Ltd (Japan). CRM197, a non‐toxic mutant of diphtheria toxin, was prepared as described previously (Uchida et al., 1973).
PKC expression vectors encoding PKCα (YK504), PKCδ (M241), PKCε (YK529), DR144/145A (constitutive active mutant of PKCδ) and DRKA (dominant negative mutant of PKCδ) have been described previously (Hirai et al., 1994; Ohno et al., 1994). Tag‐αR22A/A25E (SRHis‐αR22A/A25E), tag‐DR144/145A (SRHis‐DR144/145A) and tag‐εK155A/R156A/A159E (SRHis‐εK155A/R156A/A159E) encode constitutive active mutants of PKCα, PKCδ and PKCε, respectively (Ueda et al., 1996), which fuse downstream of the six histidine residues and a 12 amino acid sequence from the T7 gene 10 leader sequence. AD335 or DA318 are chimeric mutants that fuse rabbit PKCα (amino acids 1–318) and mouse PKCδ (amino acids 335–674) or mouse PKCδ (amino acids 1–335) and rabbit PKCα (amino acids 318–672), respectively.
Rabbit anti‐human proHB‐EGF antibody #H6 was raised against a synthetic peptide corresponding to amino acids 54–73 of proHB‐EGF (Iwamoto et al., 1994). Goat anti‐HB‐EGF neutralizing antibody was purchased from R & D Systems. Monoclonal antibodies (mAb) of anti‐T7 tag antibody were purchased from Novagen. The monoclonal antibodies (mAb) of anti‐PKCα (P16520), anti‐PKCδ, (P36520) and PKCε (P14820) were purchased from Transduction Laboratories (TDL). The polyclonal antibody of anti‐PKCδ (δ5) was raised against a synthetic peptide corresponding to amino acids 656–673 of the rat PKCδ (Mizuno et al., 1991), however, mAb of PKCδ (P36520) was raised against N‐terminal regulatory region of human PKCδ. Rabbit anti‐MDC9 antibody was raised against a synthetic peptide corresponding to amino acids 795–809 of the mouse MDC9. HRP‐conjugated sheep anti‐rabbit IgG and anti‐mouse IgG were purchased from Amersham.
Isolation of a cDNA clone encoding MDC9/meltrinγ
Mouse PKCδ (400 ng) purified from recombinant baculovirus‐infected Sf21 cells was labeled by autophosphorylation with 32P in 155 μl of kinase buffer containing 20 mM Tris–HCl, 5 mM MgCl2, 10% glycerol, 0.6 mCi/ml [γ‐32P]ATP and 40 μg/ml phosphatidylserine pH 7.5, at 30°C for 2 h. After the phosphorylation reaction, [32P]‐labeled PKCδ was separated from unreacted ATP by gel filtration and used as a probe (106 c.p.m./ml) to screen an NIH 3T3–λEXlox cDNA expression library by an overlay method as described (Izumi et al., 1997). The initial isolate (clone G1) encoded amino acid residues 719–845 of mouse MDC9. A cDNA fragment that was initially identified as an RT–PCR product encoding a disintegrin domain of MDC9/meltrinγ (Yagami‐Hiromasa et al., 1995) was used as a probe to isolate its full‐length cDNAs. They were screened in a cDNA library of a mouse myoblast cell line, C2, made in ZAP II (Stratagene).
Overlay assay using [32P]‐labeled PKCδ as a probe
E.coli lysates containing PKCδ binding proteins were subjected to SDS–PAGE and blotted onto a PVDF membrane. After treatment with a 5% skim milk solution, the PVDF membrane was incubated with [32P]‐labeled PKC (106 c.p.m./ml), diluted in 50 mM Tris–HCl pH 7.5, 0.5 M NaCl, 50 μg/ml phosphatidylserine and 1% bovine serum albumin (BSA) at room temperature for 5 h. Excess ligand was removed by washing the PVDF membrane with Tris‐buffered saline (TBS) and the membrane was subjected to autoradiography.
In vitro phosphorylation of GST–MDC9 proteins by PKCδ
GST–MDC9 proteins were purified on Glutathione–Sepharose 4B (Pharmacia). The reaction mixture contained 20 mM Tris–HCl pH 7.5, 5 mM Mg(OAc)2, 10 μg/ml leupeptin, 50 ng/ml TPA, 25 μg/ml phosphatidylserine, 3.74 ng PKCδ and 50 nM of GST–MDC9 proteins in a total volume of 20 μl. The reaction was started by the addition of 20 μM ATP and 0.5 mCi [γ‐32P]ATP and the mixture was incubated at 30°C. After 5 min of incubation, the reaction was stopped by the addition of 5 μl of 5×Laemmli's SDS‐sample buffer. Proteins were then separated on SDS–PAGE and subjected to autoradiography.
COS cells in 10 cm dishes were suspended in 200 μl of lysis buffer containing 20 mM Tris–HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 50 mM NaF, 1 mM Na3VO4, 10 μg/ml leupeptin, 1 mM PMSF, 1.8 μg/ml aprotinin and 1% Triton X‐100. After 30 min incubation on ice, the lysates were clarified by centrifugation at 14 000 r.p.m. for 30 min and incubated with antibodies preabsorbed on Protein G–Sepharose (Pharmacia) for 1 h at 4 °C. The immunocomplexes on Sepharose were washed six times with lysis buffer, after which the proteins were separated by SDS–PAGE, transferred to PVDF membranes and probed with the indicated antibodies.
Construction of MDC9 mutants
An E.coli expression vector for GST–MDC9 was constructed using the expression vector pGEX‐3X (Pharmacia) and cDNA encoding the MDC9 cytoplasmic region. Deletion mutants of the MDC9 cytoplasmic region were constructed by PCR using clone G1 as a template. The mammalian expression vector for MDC9 was constructed using expression vector pEF‐BOS (Mizushima and Nagata, 1990) (pEF‐BOS–MDC9). MDC9ΔPKCBD (pEF‐BOS–MDC9ΔPKCBD), ΔPro‐rich (pEF‐BOS–MDC9ΔPro‐rich), ΔMP (pEF‐BOS–MDC9ΔMP) and ΔPKCBDΔMP (pEF‐BOS–MDC9ΔPKCBDΔMP) were constructed by the deletion of amino acid residues 719–743, 811–845, 256–394, 256–394 and 719–743, respectively. MDC9 H347,351A (pEF‐BOS–MDC9 H347,351A) has alanines substituted for the conserved histidines at positions 347 and 351. A PstI–HincII fragment of MDC9 cDNA was cloned into pBlueScript, and mutations for two histidines were generated by site‐directed mutagenesis (Quikchange Site‐Directed Mutagenesis Kit, Stratagene, La Jolla, CA) using the primers 5′‐CCATTGTTGCTGCTGAATTGGGGGCTAACCTTGGAATG‐3′ and 5′‐CATTCCAAGGTTAGCCCCCAATTCAGCAGCAACAATGG‐3′.
Finally the mutated PstI–HincII fragment was ligated into the corresponding sites of pEF‐BOS‐MDC9. MDC9cyto (SRHis–MDC9cyto) was constructed using clone G1 (amino acids 719–845 of mouse MDC9) ligated into SRHis vector (Ueda et al., 1996). MDC9 cytoΔPKCBD (SRHis–MDC9cytoΔPKCBD) was constructed using SRHis–MDC9cyto, which encodes amino acid residues 744–845 of mouse MDC9.
Cell culture and transfection
Vero, Vero‐H and Vero‐MDC9 cells were maintained in modified Eagle‘s medium (MEM) with nonessential amino acids (MEM‐NEAA) supplemented with heat‐inactivated 10% fetal calf serum (FCS). Vero‐H cells are a stable transformant of Vero cells that overexpress human proHB‐EGF (Goishi et al., 1995). Vero–MDC9 cells were isolated transfecting mouse MDC9 cDNA into Vero cells. COS‐7 cells from monkey kidney were cultured in Dulbecco's MEM (DMEM) supplemented with 10% FCS. DER cells, 32D cells that stably express EGF receptor, were grown in RPMI 1640 medium containing 10% FCS and 5% WEHI‐3 cell‐conditioned medium as a source of IL‐3. Transfection was carried out with plasmids by electroporation (Bio‐Rad, Gene Pulser) according to the manufacture's instructions.
Western blot analysis
For sample preparation, the cell in culture dishes were directly solubilized by using Laemmli sample buffer (Figures 1C, 3B and 4B) (Laemmli, 1970). Western blot analysis was performed by one‐dimensional electrophoresis followed by electrophoretic transfer to PVDF membranes, which were then incubated with various antibodies. Antibodies were detected by chemiluminescence ECL (Amersham).
Cells plated on coverslips were incubated with 5 μg/ml of rabbit anti‐HB‐EGF antibody #H6 in skim milk solution (20 mM Tris–HCl pH 7.2, 0.2 mM CaCl2, 0.2 mM MgCl2, 150 mM NaCl and 5% skim milk) for 2 h on ice, washed three times with PBS(+) and fixed with 3.7% formaldehyde in PBS(−) for 1 h on ice. Cells were washed twice with PBS(−), incubated with 0.1 M glycine/50 mM Tris–HCl pH 8.0, for 1 h at 4°C, and overnight with skim milk solution at 4°C. ProHB‐EGF on the cell surface was detected by incubating with Cy3‐conjugated goat anti‐rabbit IgG (Amersham) for 1 h at 4°C and washing three times with PBS(−). In order to detect the expression of PKC isotypes, MDC9 or derived mutants, cells were permeabilized with 0.1% Triton X‐100 at 4°C for 5 min and incubated with an appropriate antibody for 1 h at 4°C. After washing three times with TBST (20 mM Tris–HCl pH 7.5, 150 mM NaCl, 0.05% Tween 20), the cells were incubated with fluorescein isothiocyanate (FITC)‐conjugated goat anti‐mouse IgG (EY Laboratories) for 1 h at room temperature. The cells were washed three times with TBST and observed under a fluorescence microscope. Images were captured with a Princeton Instruments digital camera and manipulated with Adobe Photoshop software.
The percentage of proHB‐EGF‐positive cells was determined by counting the number of proHB‐EGF‐positive cells (red fluorescence) among cells concomitantly expressing products of transfected cDNA (green fluorescence). The values were determined based on the results obtained in at least two independent transfections and at least 100 independent cells positive for marker proteins were examined in each experiments for the clear reduction (not complete disappearance) of the staining of proHB‐EGF ectodomain. The scoring was performed in a completely blind manner.
The binding of 125I‐labeled DT to cells was measured as described previously (Mekada and Uchida, 1985). Typically, cells transfected with plasmids were cultured with or without 64 nM TPA in binding medium (MEM supplemented with 10% FCS and 20 mM HEPES–NaOH pH 7.2) for 30 min at 37°C. The medium was then replaced with fresh cold binding medium containing 10 μg/ml heparin and the cells were incubated with 100 ng/ml 125I‐labeled DT (107 c.p.m./μg) for 8 h at 4°C. The cells were washed and the cell‐associated radioactivity was measured. Nonspecific binding of 125I‐labeled DT was assessed in the presence of a 100‐fold excess of unlabeled DT. Specific binding was determined by subtracting the nonspecific binding from the total binding obtained with 125I‐labeled DT alone. Nonspecific binding accounted for <10% of total binding.
To determine the amount of soluble HB‐EGF, conditioned medium was harvested, centrifuged to remove cell debris and incubated with 50 μl of heparin–Sepharose CL‐6B (Pharmacia LKB Biotechnology, Sweden) for 2 h at 4°C. After washing twice with PBS(+) containing 1mg/ml BSA, the gels were suspended in washing solution and incubated with 100 ng/ml 125I‐labeled DT for 6 h at 4°C with gentle rotation. The gels were washed and the radioactivity associated with the gels was counted. Specific binding was determined as described above.
Measurement of mitogenic activity
The mitogenic activity of secreted HB‐EGF was measured by co‐culture assay as follows. Vero‐H cells were transfected with control plasmids pEF‐BOS or plasmids encoding MDC9 and then seeded into 24‐well plates at a density of 5×104 cells/well. After incubation for 24 h at 37°C, the medium was replaced with fresh RPMI 1640 medium supplemented with 10% serum. Then a Transwell chamber (Millipore Corporation, Bedford, MA) was set in each well and DER cells were inoculated into the inside of the Transwell chambers (1×104 cells/well). Cells were cultured for 24 h at 37°C and then incubated with [3H]thymidine (2 μCi/ml) for 4 h. The DER cells were then harvested and the radioactivity incorporated into DNA was determined as described previously (Higashiyama et al., 1995). To determine the mitogenic activity attributable to growth factors other than HB‐EGF, the rate of DNA synthesis was also determined in the presence of CRM197 (Mitamura et al., 1995). The mitogenic activity derived from HB‐EGF was calculated by subtracting the radioactivity obtained in the presence of CRM197 from that obtained without CRM197.
E.Mekada is supported in part by a grant from The Research for the Future Program, the Japan Society for the Promotion of Science (JSPS) (Project No. 97L00303), and a Grant‐in Aid for Scientific Research, The Ministry of Education, Science, Sports and Culture (No. 09480198). S.Ohno is supported in part by a grant from The Research for the Future Program, JSPS (Project No. 96L00305), and a Grant‐in Aid for Scientific Research, The Ministry of Education, Science, Sports and Culture. Y.Izumi is a Research Fellow of JSPS for Young Scientists.
- Copyright © 1998 European Molecular Biology Organization