Tumor necrosis factor alpha converting enzyme (TACE) is the metalloprotease-disintegrin responsible for the ectodomain shedding of several proteins, including tumor necrosis factor α. Using the yeast two-hybrid system, we identified the scaffolding protein synapse associated protein 97 (SAP97) as a binding partner of the cytoplasmic domain of TACE. By deletions and site-directed mutagenesis, we demonstrated that this interaction involved the PDZ3 domain of SAP97 and the extreme C-terminal amino-acid sequence of TACE. This interaction as well as the identification of the specific domains involved was confirmed in vitro by affinity purification and in mammalian cells by co-immunoprecipitation and alteration of localization analyzed by immunofluorescence microscopy. In addition, confocal microscopy showed that endogenous TACE and SAP97 colocalized in some intracellular areas of COS-7 cells and CACO-2 cells. Furthermore, overexpression of SAP97, unlike that of a mutant form of SAP97 deleted for its PDZ3 domain, altered the ability of TACE to release its substrates. Altogether, these results demonstrate an interaction between TACE and SAP97, which may have a functional implication for the regulation of TACE shedding activity.
Metalloprotease-disintegrins (also called ADAMs) are originally type I transmembrane glycoproteins. Their basic structure is phylogenically well conserved and consists of multiple domains with specific roles: a pro-domain that represses the activity of the metalloprotease-like domain (Loechel et al., 1999; Milla et al., 1999; Roghani et al., 1999) followed by a disintegrin-like domain and a cysteine-rich domain, both with adhesive properties (Cho et al., 1998; Iba et al., 2000; Iba et al., 1999; Linder and Heinlein, 1997; Nath et al., 1999; Yuan et al., 1997; Zhang et al., 1998), a transmembrane domain and a cytoplasmic tail domain that binds cell signaling component (Izumi et al., 1998; Roghani et al., 1999). ADAMs have been implicated in numerous cellular processes (Primakoff and Myles, 2000; Yamamoto et al., 1999), some of which involve shedding of cell-surface molecules.
Tumor necrosis factor alpha converting enzyme (TACE) is an ADAM (ADAM17) originally described as the main enzyme responsible for tumor necrosis factorα (TNF) release from membranes (Black et al., 1997; Moss et al., 1997). Subsequently, TACE was implicated in the ectodomain shedding of TNF receptors p55 (TNFR1), p75 (TNFR2) (Peschon et al., 1998; Reddy et al., 2000) and several other proteins (Moss et al., 2001).
The regulation of TACE activity is poorly understood. On the basis of the effect of phorbol esters on the internalization and degradation of cell-surface TACE (Doedens and Black, 2000), a multilevel regulation of TACE activity involving changes in membrane targeting or changes in the interaction of the enzyme with regulatory proteins was proposed (Reddy et al., 2000). The cytoplasmic domain of TACE contains several signaling motifs such as SH3 ligand domains and a protein kinase C phosphorylation site (Black et al., 1997). Mitotic arrest deficient 2 protein (MAD2) was identified as a partner of TACE cytoplasmic tail (Nelson et al., 1999), and this finding suggests a potential relationship between TACE and the cell cycle, but does not provide information on the regulation of TACE activity. The extracellular signal-regulated kinase binds to the cytoplasmic tail of TACE and phosphorylates threonine 735 (Diaz-Rodriguez et al., 2002). Recently, the protein-tyrosine phosphatase PTPH1 was shown to interact with the C-terminus of TACE and was suggested to be a negative regulator of TACE levels and function (Zheng et al., 2002).
To gain insights into the regulation of TACE activity, we searched for binding partners that interact with its cytoplasmic domain. Yeast two-hybrid experiments allowed us to identify the synapse associated protein 97 (SAP97) as a potential binding partner of the TACE cytoplasmic tail. This interaction was confirmed biochemically in vitro and demonstrated in COS-7 cells by co-immunoprecipitation and immunofluorescence microscopy. Confocal microscopy revealed some overlapping clusters of endogenous TACE and SAP97 in COS-7 cells and in CACO-2 cells. A functional implication of this interaction was suggested by the fact that overexpression of SAP97 decreased the release of TACE-processed substrates, whereas overexpression of a mutant form of SAP97 that did not bind TACE did not have an effect.
Materials and Methods
cDNA fragments were cloned using total RNA from PMA-treated HL-60 cells. RT PCR was performed using the single tube RT PCR kit from Qiagen; PCR products were cloned into pGEM vector (Promega) and sequenced. For yeast two-hybrid system experiments, the sequence coding for TACE cytoplasmic tail was introduced in frame with the GAL4 DNA-binding domain of the pAS2-1 vector (Clontech) between EcoRI and SalI sites to give TACE (from amino acid 695 to amino acid 824 cf. Fig. 1). Point mutations in the C-terminal sequence of TACE were introduced by PCR (Fig. 2). This sequence was introduced into pAS2-1 and was called TACEm. For pull-down experiments, the sequence coding for the TACE cytoplasmic tail was introduced into pGEX 4T1 (Pharmacia). The sequence from the disintegrin domain to the cytoplasmic domain of TACE (from amino acid 475 to 824) was introduced into pHooK vector (Invitrogen), generating haTACE, which possesses a dominant-negative activity on endogenous TACE (Solomon et al., 1999). A similar construct with point mutations at the C-terminal extremity of TACE (cf. Fig. 3) was called haTACEm. Both constructs possess the Murine Ig kappa-chain V-J2 signal peptide and the HA epitope from the pHook vector.
All SAP97-cDNA-containing plasmid vectors were generated from pEGFP SAP97 as described elsewhere (Wu et al., 1998). For in vitro translation, the sequence of the pET28b (Novagen) containing the T7 promoter and the ribosome-binding site was introduced in the pEGFP-containing SAP97 sequence. For two-hybrid experiments, SAP97 sequences were introduced into pGADC1 plasmid vector (James et al., 1996), generating SAP97 fused to the GAL4 activation domain.
Complete coding regions of TNF, TNFR1 and TNFR2 cDNAs were introduced into pcDNA3 plasmid vector (Invitrogen).
Yeast two-hybrid screen
The yeast reporter strain PJ69a (James et al., 1996) was co-transformed with a human placental cDNA library cloned into the pACT2 vector (Clontech) and pAS-TACE (described above).
Selection was made by growth on histidine-, adenine-, leucine- and tryptophane-free media. LacZ gene expression was determined with a colorimetric filter assay. cDNA clones from positive colonies were isolated, transferred into XL1 Blue bacteria and identified by cDNA sequencing.
Binding assays in vitro
GST fusion proteins were overexpressed in BL21 Escherichia coli. Cells were lysed by sonication and GST fusion proteins were purified using glutathione-Sepharose 4B beads as recommended by the manufacturer (Amersham Pharmacia Biotech). The amount of cell lysate incubated with the beads was adjusted so that the level of the different GST fusion proteins eluted from the glutathione-agarose beads was similar. After extensive washes with PBS, the beads were washed once and suspended in 500 μl of binding buffer (75 mM NaCl, 20 mM Tris-HCl pH 7.4, 0.1 mM EDTA, 2.5 mM MgCl2, 0.75 mg/ml BSA, 0.1% Tween 20, 1 mM DTT).
In vitro transcriptions and translations experiments were performed using T7 TNT Quick Coupled Reticulocyte Lysate System (Promega). Biotinylation of the neosynthesized protein was performed using TranscendTM (Promega).
10 μl of reticulocyte lysate (containing the biotinylated protein) was added to the glutathione-agarose bound GST fusion protein in binding buffer and incubated for 4 hours at 4°C on a rotator, then washed four times with PBS and incubated in 20 μl 1× SDS sample buffer for 5 minutes at 95°C. The supernatant was submitted to a SDS polyacrylamide gel electrophoresis according to Laemmli (Laemmli, 1970). After electrotransfer onto a PVDF membrane, biotinylated proteins were detected using a streptavidin alkaline phosphatase conjugate.
Cell culture and transfection
COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum, 2 mM glutamine, 100 i.u./ml penicillin and 100 μg/ml streptomycin. CACO-2 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 20% (v/v) fetal bovine serum, 0.1 mM non-essential amino acids, 2 mM glutamine, 100 i.u./ml penicillin and 100 μg/ml streptomycin. Transient transfections were performed with Polyfect reagent (Qiagen). Analyses were made 48 hours after transfection.
Immunoprecipitation of overexpressed proteins
Cells were lysed with 1% (v/v) Nonidet P40, 20 mM Tris-HCl, pH 7.4, 75 mM NaCl, protease inhibitors and 10 μM of the metalloprotease inhibitor active on TACE RU36156. Lysates were preclarified by centrifugation and incubated for 1 hour with Protein A/G plus agarose (Santa Cruz Biotechnology) and for the next 2 hours with Protein A/G plus agarose coated with antibodies (anti-HA epitope or anti-GFP from Santa Cruz Biotechnology). Beads were washed four times with lysis buffer and incubated in 20 μl of 1× SDS sample buffer for 5 minutes at 95°C. The supernatant was submitted to western blot analysis to detect either overexpressed SAP97 using the anti-SAP97 monoclonal antibody (kindly provided by C. Garner, University of Alabama, Birmingham) or the endogenous TACE using the cytoplasmic-domain-specific TACE antibody from Santa Cruz Biotechnology (# C-15).
Immunoprecipitation of endogenous proteins
The interaction between TACE and SAP97 was maintained by lowering the pH of the lysis buffer to 6.8. To decrease non-specific binding, the preclarified lysate was concentrated using a Centricon YM-100, which removes proteins with a mass below 100 kDa and the retentat was diluted to the initial volume with lysis buffer. The lysate was then treated as described above except that the antibody used for immunoprecipitation was anti-SAP97 (Santa Cruz Biotechnology).
Detection was made using either the cytoplasmic-domain-specific TACE antibody or the ectodomain-specific TACE antibody form R&D Systems Europe (Lille, France) (clone: 111633).
Localization of haTACE was performed in COS-7 cells. Cells were washed three times with PBS (with calcium and magnesium), fixed for 5 minutes in 1% paraformaldehyde, permeabilized with 0.2% Triton-X 100 in PBS for 20 minutes, washed with PBS 1% BSA and incubated with anti-HA epitope antibody for 1 hour. After extensive washes, rhodamine-labeled secondary antibody was incubated for 1 hour. Cells were washed and observed by epifluorescence microscopy. Localization of endogenous TACE and SAP97 was performed in COS-7 and CACO-2 cells. Cells were fixed, permeabilized and incubated with anti-SAP97 for 1 hour, washed and incubated for 1 hour with a fluorescein isothiocyanate (FITC)-conjugated secondary antibody, washed again and incubated for 1 hour with a phycoerithrin (PE)-conjugated antibody against TACE (R&D Systems, clone: 111633). Analyses were made by confocal microscopy using a Leica PCS SP2 microscope.
Flow cytometry analysis
Surface expression of the overexpressed form of TACE (haTACE) on COS-7 cells was analyzed by flow cytometry using FITC-conjugated monoclonal antibody against HA epitope (clone BMG-3F10 from Roche). Cells assigned to intracellular staining were treated according to the specification of the Intrastain kit from Dako Cytomation (Trappes, France). FITC-labeled cells were analyzed on a XL-cytofluorograph (Coulter Electronics Inc.) at 488 and 525 nm, which correspond to excitation and detection wavelengths, respectively.
N-linked carbohydrate residues were removed using either peptide:N-glycosidase F (PNGase F) or endoglycosidase H (EndoH) (New England Biolabs) as described by the manufacturer. COS-7 cells were directly lysed in the reaction buffer furnished with the enzyme. Lysates were heated at 100°C for 10 minutes; aliquots were adjusted for deglycosylation and treated with or without 1000 units of PNGase F or EndoH at 37°C for 8 hours. Samples were then separated by SDS-PAGE and analyzed by western blotting.
Total protein of cell lysates was assayed using the bicinchoninic acid protein assay kit from Sigma. Total amount (free and bound forms) of TNF, TNFR1 and TNFR2 were assayed in cell lysates and in culture media after 24 hours of accumulation, according to the specification of their respective enzyme-linked immunosorbent assays kits (R&D Systems).
Interaction of TACE with SAP97 in the yeast two-hybrid system
The cytoplasmic tail of TACE was used as bait in two-hybrid screens to identify interacting partners from a placental cDNA library. Five different clones were isolated on the basis of both nutritional selection andβ -galactosidase activity. Only three of them contained a cDNA coding for cytoplasmic proteins. One of these clones (which was originally isolated in duplicate) contained a fragment of SAP97 cDNA coding for a protein from amino acid 422 to the stop codon (amino acid 904). This protein contained the third PDZ domain (PDZ3), the SH3 domain and the guanylate-like kinase (GK) domain of SAP97 (Fig. 1) and was called SAP97DPDZ1-2. Full-length SAP97 cDNA cloned in frame of the GAL 4 activation domain of pGADC1 allowed also the growth of yeast on a selective media when co-transformed with pAS-TACE (Fig. 1A). This result suggests that TACE and SAP97 possess the ability to interact in vivo and that the cytoplasmic region of TACE and at least the PDZ3, SH3 and GK domains of SAP97 are involved in this interaction.
Analysis of the interaction of TACE with SAP97
Our two-hybrid screens showed that the sequence of SAP97 from amino acid 422 to amino acid 904 is involved in the binding with the cytoplasmic tail of TACE. A mutant form of SAP97 deleted for the last 165 amino acids (85% of the GK domain) was also able to interact with TACE as suggested by the result of a two-hybrid screen (data not shown). Thus, the cytoplasmic tail of TACE binds SAP97 between its amino acids 422 and 739. This region of SAP97 contains the PDZ3 domain, the SH3 domain and 25 amino acids of the GK domain. We narrowed down this region by deleting the PDZ3 (SAP97DPDZ3) or SH3 (SAP97DSH3) domains from full-length SAP97 and testing the interaction of these mutants forms with TACE cytoplasmic tail using two-hybrid system (Fig. 1A). Mutant forms of SAP97 were correctly produced in yeast (Fig. 1B), and both deletions slowed down the growth of yeast on selective media. However, PDZ3 domain deletion was the most effective in reducing the growth of yeast (Fig. 1A). The involvement of the PDZ3 domain of SAP97 in the binding to TACE cytoplasmic tail was also investigated using an in vitro biochemical binding assay. In vitro translated and biotinylated fragments of SAP97 (Fig. 1C, lanes 1 and 2) were incubated with glutathione-agarose bound GST-TACE cytoplasmic tail and submitted to pull-down experiments. The GST-TACE cytoplasmic tail retained more SAP97DPDZ1-2 than SAP97DPDZ1-3 (Fig. 1C, compare lanes 3 and 4). None of these forms binds the GST alone (Fig. 1C, lanes 5 and 6). These results suggest a privileged role of the PDZ3 domain of SAP97 in the interaction with TACE.
PDZ domains are multifunctional protein-protein recognition modules involved in the clustering of signaling molecules and play an important role in organizing protein networks on membranes (Fujita and Kurachi, 2000; Harris and Lim, 2001). In many cases, PDZ domains specifically bind a motif occurring at the C-terminus of target proteins (Harris et al., 2001). To test if the extreme C-terminal sequence of TACE could be responsible for the interaction with SAP97, we modified this sequence and we analyzed the binding of TACE to SAP97 using the two-hybrid system. Yeast were co-transformed with pGADC1 coding for the full-length SAP97 and with pAS2-1 TACE (coding for the wild-type cytoplasmic tail of TACE: 821E-T-E-C824) or with pAS2-1 TACEm (coding for a mutated version of the cytoplasmic tail of TACE: 821D-A-E-C824). Yeast expressing TACE and SAP97 were able to grow during nutritional selection unlike those expressing TACEm and SAP97 (Fig. 2A). However, TACEm fused to GAL4 DNA-binding domain was produced in yeast as efficiently as TACE (Fig. 2B). This finding suggests that the C-terminal extremity of TACE is involved in the interaction with SAP97.
Interaction of TACE and SAP97 in mammalian cells
COS-7 cells constitutively express detectable amounts of TACE and SAP97 (see below). By size exclusion chromatography, SAP97 and TACE (immature and mature forms) co-eluted in a range of molecular masses between 250 kDa and 150 kDa (data not shown). As the molecular mass of each protein is between 90 and 120 kDa, this suggests that these proteins exist in complexes. We used a specific antibody to immunoprecipitate SAP97 from COS-7 cells lysate, and the presence of endogenous TACE in the resulting immune complex was assessed by western blots with two different antibodies. The cytoplasmic-domain-specific TACE antibody allowed us to detect proteins migrating at the level of the immature (with the prodomain) and mature (without the prodomain) forms of TACE (Fig. 3A); an unidentified fast migrating protein was also detected. The ectodomain-specific TACE antibody, which mainly recognizes the mature form of TACE in NP-40-based cell lysate, allowed us to detect a protein that migrated at the level of the mature form of TACE (Fig. 3B). This result suggests that endogenous TACE and SAP97 are able to interact. Moreover, immature and mature forms of TACE were co-immunoprecipitated with overexpressed GFP-SAP97 but not with GFP-SAP97DPDZ3, suggesting that in mammalian cells, immature and mature forms of TACE interact with the PDZ3 domain of SAP97 (Fig. 3C). When a HA-tagged membrane-anchored cytoplasmic tail of TACE (haTACE) was overexpressed together with the full-length GFP-SAP97, both proteins were immunoprecipitated with an anti-HA epitope antibody (Fig. 3D). Mutations of the extreme C-terminal part of haTACE (haTACEm) abolished the co-immunoprecipitation, suggesting that in mammalian cells, the C-terminal extremity of TACE is engaged in the interaction with SAP97. Altogether, these data are strongly favor of an intracellular interaction between human TACE (immature and mature forms) and SAP97. This interaction involves the extreme C-terminal part of TACE and the PDZ3 domain of SAP97.
Intracellular localization of SAP97 and TACE
Localization of overexpressed SAP97 and TACE in COS-7 cells was assessed by epifluorescence microscopy. GFP-SAP97 exhibited a diffuse pattern (Fig. 4A), and this distribution was comparable to that of GFP-SAP97DPDZ3 (data not shown). haTACE showed also a diffuse staining pattern in COS-7 cells (Fig. 4B), which has already been described (Schlondorff et al., 2000). This distribution is similar to that of haTACEm (data not shown) and is, a priori, consistent with endoplasmic reticulum and/or surface protein localization. When overexpressed with haTACE, GFP-SAP97 accumulated as round aggregates (Fig. 4C). This particular distribution was clearly evident in more than 70% of the transfected cells and was not observed when GFP-SAP97 was coexpressed with haTACEm (Fig. 4D) or when the PDZ3 domain of GFP-SAP97 was deleted (Fig. 4E). This result suggests that the interaction between overexpressed GFP-SAP97 and haTACE, which involves their PDZ3 domain and the extreme C-terminal part, respectively, is responsible for their intracellular aggregation. To support this result, we examined, simultaneously, intracellular distributions of overexpressed GFP-SAP97 and haTACE and noticed that they mostly colocalized in so-called aggregates (Fig. 5). Altogether, these data show that association of overexpressed haTACE with GFP-SAP97 results in the alteration of their intracellular distribution.
The form of the human TACE that was overexpressed in COS-7 cells did not contain the prodomain and the catalytic domain (Materials and Methods). Since the prodomain was described to be necessary for the targeting of TACE (Milla et al., 1999) we investigated the cellular distribution of haTACE by analyzing its sensitivity to N-glycosidase F (PNGase F) and endoglycosydase H (EndoH). N-linked sugars are sensitive to PNGase F but most of them become resistant to EndoH after they are modified in the medial Golgi. We found that both enzymes were equally efficient in deglycosylating the overexpressed haTACE (Fig. 6A). This suggests that haTACE is present in the early secretory pathway (endoplasmic reticulum and/or proximal Golgi). By comparison, deglycosylation of the endogenous TACE showed, as already described (Schlondorff et al., 2000), that both immature and mature forms of TACE were sensitive to PNGase F. By contrast, only the immature form of TACE was sensitive to EndoH (Fig. 6B), suggesting that the immature form is present in the endoplasmic reticulum and proximal Golgi, whereas the mature form traverses the medial Golgi. Detection of the overexpressed haTACE by flow cytometry favours the hypothesis that the overexpressed haTACE cannot go through the Golgi apparatus. Less than 10% of the cells expressed a very small amount of haTACE at the cell surface (Fig. 6C), whereas, virtually all transfected cells (around 60%) expressed haTACE intracellularly (Fig. 6D). When co-transfected with SAP97 or SAP97DPDZ3, haTACE was always sensitive to deglycosylation by EndoH (data not shown), suggesting that haTACE is always present in the early secretory pathway. Taken together these data allowed us to conclude that the interaction between haTACE and GFP-SAP97, which can be visualized by the aggregation of these proteins, takes place in the early secretory pathway. However, since overexpressed haTACE cannot traverse the medial Golgi, we hypothesized that endogenous TACE and SAP97 interact in other cell compartments. In favor of this hypothesis are the results shown on Fig. 3A-C, which demonstrate that endogenous or overexpressed SAP97 can co-immunoprecipitate both immature (present in the early secretory pathway) and mature (that had traversed the medial Golgi) forms of TACE. To confirm the physiological relevance of this interaction, we localized endogenous TACE and SAP97 by immunofluorescence confocal microscopy. In COS-7 cells, the staining associated with the detection of each of these two proteins showed a diffuse pattern that significantly overlapped in some intracellular areas and at the cell lateral membrane (Fig. 7A). In the epithelial cell line CACO-2, SAP97 and TACE were produced (Fig. 7B) and colocalized at the lateral membrane at sites of cell-cell contacts (Fig. 7A).
Functional implication of TACE/SAP97 interaction
We investigated whether the overexpression of SAP97 affected the ability of endogenous TACE to release its substrates. COS-7 cells were co-transfected with TACE substrates (TNF, TNFR1 and TNFR2) together with different forms of SAP97, and the amount of substrates (both released and cell-associated) was measured. We first verified that the releases of TNF, TNFR1 and TNR2 from COS-7 cells were mainly due to TACE activity. The release of TACE substrates was reduced by treatment with the metalloprotease inhibitor RU36156 active on TACE (Gallea-Robache et al., 1997) and by overexpression of a dominant-negative form of TACE (haTACE mentioned above) (Table 1), which suggests that TACE is responsible for the shedding of the overexpressed TNF, TNFR1 and TNFR2 from COS-7 cells (Table 1). The coexpression of GFP-SAP97 together with TACE substrates did not alter the original intracellular distribution of GFP-SAP97 (data not shown) but reduced the amount of released TACE-processed substrates (Table 1) at least as efficiently as the dominant-negative form of TACE. This inhibitory property was lost when cells were co-transfected with TACE substrates and SAP97DPDZ3 (Table 1). Accordingly, cells co-transfected with SAP97 expressed more cell-associated TNF and TNFR2 (370±90 pg/μg and 55±5 pg/μg of cellular proteins, respectively) than cells co-transfected with SAP97DPDZ3 (190±20 pg/μg and 26±3 pg/μg, respectively). The amount of cell-associated TNFR1 assayed in cell lysates was below the detection limit of the assay. This result suggests that, when cells were co-transfected with SAP97, the low release of TACE-processed substrates is not due to an inhibition of their synthesis but rather to a downregulation of their cleavage. These data emphasize the specific inhibitory effect of overexpressed SAP97 on the release of TACE-processed substrates. Because deletion of the region of SAP97 involved in the binding with TACE (the PDZ3 domain) abrogates this inhibitory effect, it can be reasonably concluded that the interaction of overexpressed SAP97 with the endogenous TACE is responsible for the inhibitory effect on the release of TACE-processed substrates. Dual color cytometry analysis with a PE-labeled anti-TACE ectodomain antibody that detects TACE at the surface of GFP-positive cells did not show any significant difference in the amount of endogenous TACE exposed at the surface of GFP-SAP97- and GFP-SAP97DPDZ3-overexpressing cells (data not shown). This result suggests that the reduced TACE activity measured in SAP97 overexpressing cells was not due to a decreased amount of active TACE exposed at the cell surface.
This study was initiated to gain insights into TACE regulation through the identification of its potential cytoplasmic partners. The scaffolding protein SAP97 was identified by a yeast two-hybrid system as a potential TACE-interacting protein. Two-hybrid and in vitro experiments allowed us to localize elements necessary for this interaction in the third PDZ domain of SAP97, the PDZ3 domain (Fig. 1). Eukaryotic PDZ domains are defined as multifunctional protein-protein interaction modules that play a central role in organizing diverse cell signaling assemblies (Fujita and Kurachi, 2000; Harris and Lim, 2001). To our knowledge, none of SAP97 partners already described has been shown to interact with its PDZ3 domain. The neuronal G-protein-gated inwardly rectifying K (+) channel, Kir3.2c, binds to the second PDZ domain of SAP97 and becomes sensitive to G protein stimulation (Hibino et al., 2000). The strong inwardly rectifying potassium channels Kir2.x interacts also with the second PDZ domain of SAP97 (Leonoudakis et al., 2001). Weak binding of the GluR6/Kainate receptors to the first PDZ domain of SAP97 could account for its subcellular localization (Mehta et al., 2001). PDZ domains specifically bind to the motif occurring at the C-terminus of target proteins. Class I PDZ domains recognize the motif E-S/T-X-O (single letter amino acid code, X denotes any amino acid and O an hydrophobic amino acid); class II PDZ domains recognize the motif O-X-O; and class III PDZ domains recognize the motif X-X-C (Harris et al., 2001). Most of the PDZ domains fall into one of these classes (Fuh et al., 2000; Schneider et al., 1999; Stricker et al., 1997). The extreme C-terminal part of TACE (821E-T-E-C824) is involved in the binding to the PDZ3 domain of SAP97 (Fig. 2). However, this sequence is not strictly comparable to that involved in the binding of class I PDZ domains as the C-terminal cysteine is not purely a hydrophobic amino acid (usually valine). Despite the presence of a C-terminal cysteine, it is unlikely that the TACE sequence would bind to class III PDZ domains because amino acid replacement upstream of the cysteine abrogates the interaction of the PDZ3 domain of SAP97. From our results we cannot state that TACE is a ligand for an already defined class of PDZ domains. Recently, the C-terminus of TACE was shown to interact with the PDZ domain of the protein-tyrosine phosphatase PTPH1, which led the authors to suggest that the C-terminus of TACE represents a novel class I PDZ-binding sequence characterized by a terminal cysteine residue (Zheng et al., 2002).
The interaction between TACE and SAP97 and the identification of the elements responsible for this interaction were confirmed in COS-7 cells by co-immunoprecipitation of endogenous and overexpressed proteins (Fig. 3). Microscopic observation of overexpressed haTACE and GFP-SAP97 distribution into COS-7 cells confirmed the demonstration of a molecular interaction between these overexpressed proteins as both proteins were found colocalized in cells as round aggregates (Figs 4, 5). This particular aggregation was not observed when GFP-SAP97DPDZ3 was expressed or when the C-terminal extremity of the overexpressed haTACE was mutated. However, deglycosylation experiments suggest that, in contrast to the endogenous TACE, the overexpressed form of TACE does not traverse the medial Golgi (Fig. 6A). Therefore, the colocalization of overexpressed TACE and SAP97 reasonably argues for their interaction but remains uninformative as to where endogenous TACE and SAP97 interact. The fact that both endogenous immature (present in the early secretory pathway) and mature (that had traversed the medial Golgi) forms of TACE were co-immunoprecipitated with endogenous and overexpressed SAP97 (Fig. 3) suggests that the interaction also takes place after TACE had traversed the medial Golgi.
In COS-7 cells, immunofluorescence confocal microscopy revealed a diffuse pattern for both endogenous SAP97 and TACE that overlapped in some intracellular areas and at the cell lateral membrane. Consistent with previous results (Wu et al., 1998), we found that in the epithelial cell line CACO-2, SAP97 localized at the lateral membrane at cell-cell adhesion sites where it was reported to be associated with the cortical cytoskeleton (Reuver and Garner, 1998). Furthermore, in our study we showed a lateral membrane localization pattern of TACE that significantly overlapped that of SAP97.
Coexpression of SAP97 with potassium channel protein (Kv1) results in the formation of intracellular aggregates containing these two proteins and blocks Kv1 channels surface expression (Kim and Sheng, 1996; Tiffany et al., 2000). Endogenous SAP97 associates with a subset of GluRI early in the secretory pathway (in the endoplasmic reticulum or cis-Golgi) and dissociates at the plasma membrane (Sans et al., 2001); these data show that interactions involving SAP97 in the early secretory pathway are of physiological importance. Interestingly, the observed colocalization of overexpressed GFP-SAP97 and haTACE in the early secretory pathway is very similar to that described for SAP97 and Kv1 channels. Overexpression of SAP97 (which binds TACE) reduced the release of three different TACE-processed substrates, TNF, TNFR1 and TNFR2, whereas that of SAP97DPDZ3 (that does not bind TACE) did not have an effect (Table 1). By contrast, cells co-transfected with SAP97 expressed more cell-associated TNF and TNFR2 than cells co-transfected with SAP97DPDZ3. These data strongly suggest that the interaction between endogenous TACE and overexpressed SAP97 is involved in the downregulation of TACE-processed substrates release (reinforcing the idea that TACE and SAP97 interact) and that in our conditions the above mentioned substrates are essentially processed by TACE. This finding also suggests a probable functional implication of the association between endogenous TACE and SAP97.
In light of the literature mentioned above, it is tempting to envisage that the downregulation of the release of TACE-processed substrates triggered by the overexpression of SAP97 could be the consequence of an intracellular sequestration of endogenous TACE. However, using dual color cytometry, we did not measure significant change in the amount of endogenous TACE exposed at the surface of cells overexpressing GFP-SAP97 and GFP-SAP97DPDZ3, ruling out the possibility that this hypothetical intracellular retention of TACE modifies the amount of cell-surface active TACE.
The C-terminal sequence of TACE binds to the PDZ domain of the protein-tyrosine phosphatase PTPH1 (Zheng et al., 2002). However, this interaction was not necessary to allow overexpressed PTPH1 to decrease the level of TACE, which can result in a reduced amount of TNF accumulated in the cell culture media. The absence of phosphorylated tyrosine in the cytoplasmic sequence of TACE leads the authors to suggest that PTPH1 acts as a negative regulator of TACE levels and function, most probably, by dephosphorylating a yet to be defined effector. These and our results underline the ability of the C-terminal sequence of TACE to bind PDZ sequences.
The need for an appropriate organization of the cytoskeleton for the correct positioning of both the shedding machinery and the substrates at the cell membrane has been evoked (Mullberg et al., 2000) in two situations. One is related to the effect of cell adhesion and spreading on the shedding of membrane-anchored heparin-binding epidermal-like growth factor (Gechtman et al., 1999) and the other is related to the effect of inhibitors that block calmodulin binding to the cytoplasmic tail of L-selectin and accelerate the release of L-selectin from cells (Kahn et al., 1998). By analogy, we speculate that the already described scaffolding role of SAP97 is involved in the transport and/or the organization of TACE, which should have an impact on TACE activity. Further investigations to evaluate the physiological implication of the interaction between TACE and SAP97 on TACE activity will be of particular interest.
SAP97 expression vectors and monoclonal antibody recognizing SAP97 were generously provided by C. Garner (University of Alabama, Birmingham). We are indebted to E. Mas (INSERM U 559, Marseilles) for image acquisition and to S. Puget (Université de Toulon) for her contribution to yeast two-hybrid screens. This work was supported by funds of INSERM.
- Accepted February 6, 2003.
- © The Company of Biologists Limited 2003