|Home | About | Journals | Submit | Contact Us | Français|
The ectodomain of certain transmembrane proteins can be released by the action of cell surface proteases, termed secretases. Here we have investigated how mitogen-activated protein kinases (MAPKs) control the shedding of membrane proteins. We show that extracellular signal-regulated kinase (Erk) acts as an intermediate in protein kinase C-regulated TrkA cleavage. We report that the cytosolic tail of the tumor necrosis factor α-converting enzyme (TACE) is phosphorylated by Erk at threonine 735. In addition, we show that Erk and TACE associate. This association is favored by Erk activation and by the presence of threonine 735. In contrast to the Erk route, the p38 MAPK was able to stimulate TrkA cleavage in cells devoid of TACE activity, indicating that other proteases are also involved in TrkA shedding. These results demonstrate that secretases are able to discriminate between the different stimuli that trigger membrane protein ectodomain cleavage and indicate that phosphorylation by MAPKs may regulate the proteolytic function of membrane secretases.
The ectodomain of a number of transmembrane proteins can be released as a soluble fragment by the action of cell surface proteases, termed secretases (Ehlers and Riordan, 1991 ; Massagué and Pandiella, 1993 ; Hooper et al., 1997 ). Proteins whose ectodomains are shed include membrane-anchored growth factors (Massagué and Pandiella, 1993 ), some of their receptors (Downing et al., 1989 ; Porteu and Nathan, 1990 ; Prat et al., 1991 ; Yee et al., 1994 ; Cabrera et al., 1996 ; Vecchi et al., 1996 ), adhesion molecules (Kishimoto et al., 1989 ), ectoenzymes (Sadhukhan et al., 1999 ), and proteins such as the β-amyloid precursor protein (Selkoe, 1994 ; Hooper et al., 1997 ). Alterations in the cleavage of some of these membrane proteins may lead to disease. Thus, mutations in the ectodomain of the p75 tumor necrosis factor (TNF) receptor (p75TNFR) that decrease the shedding of its ectodomain have been linked to an autosomal dominant disease characterized by fever and severe local inflammation (McDermott et al., 1999 ). The pathogenesis of these disease signs probably involves impaired clearance of the p75TNFR resulting in a decreased buffering activity of the soluble receptor, together with an increased level of the membrane holoreceptor that facilitates TNF-α responses. Another example of the importance of membrane sheddases has been obtained in animal models of pseudoinfectious or cachectic states (Gearing et al., 1994 ; Mohler et al., 1994 ). Injection of inhibitors of the shedding of membrane proTNF-α into rodents prevented the release of soluble TNF-α into the serum and protected these animals against a lethal dose of endotoxin (Gearing et al., 1994 ; Mohler et al., 1994 ).
Pharmacological experiments using inhibitors of the different protease families pointed to metalloproteases as the enzymes responsible for membrane protein ectodomain cleavage (Gearing et al., 1994 ; McGeehan et al., 1994 ; Mohler et al., 1994 ). Later, by the use of these inhibitors (Moss et al., 1997 ) and in vitro peptide cleavage assays (Black et al., 1997 ), an enzyme, the proTNFα-converting enzyme (TACE), which participates in the solubilization of TNF-α from its precursor, was been isolated. Structurally, TACE is a type I membrane protein that contains several domains in the extracellular region, including disintegrin and metalloprotease domains, characteristic of the ADAM subfamily of metalloproteases (Blobel, 1997 ; Black and White, 1998 ). TACE may also participate in the cleavage of other transmembrane proteins, since cells from TACE-deficient animals fail to efficiently cleave protransforming growth factor α (proTGFα), l-selectin, the p75TNFR (Peschon et al., 1998 ), or β-amyloid precursor protein (Buxbaum et al., 1998 ). It is possible that other proteases may be involved in the regulation of membrane protein ectodomain cleavage. In fact, solubilization of the ectodomain of the angiotensin-converting enzyme is unaffected in fibroblasts derived from TACE-deficient animals (Sadhukhan et al., 1999 ); studies of the solubilization of proheparin-binding epidermal growth factor (EGF)-like growth factor (proHB-EGF) have indicated that other ADAM family members, such as MDC9/ADAM9/Meltrin-γ, may also participate in the release of soluble forms of membrane-bound molecules (Izumi et al., 1998 ).
A property of the activity of membrane secretases is its highly regulated nature. In fact, activation of intracellular second messenger systems, such as the protein kinase C (PKC) or intracellular Ca2+ pathways, up-regulates the activity of membrane secretases (Massagué and Pandiella, 1993 ). Recently, mitogen-activated protein kinase (MAPK) cascades have also been implicated in the regulation of the shedding of membrane proteins. Thus, solubilization of the membrane-anchored growth factors HB-EGF (Gechtman et al., 1999 ) and proTGFα (Fan and Derynck, 1999 ), as well as the adhesion molecule l-selectin (Rizoli et al., 1999 ), can be prevented by treatments that block MAPK activation. MAPK routes are characteristically organized into a three-kinase module that includes an MAPK, the upstream kinase MEK or MKK, which phosphorylates and activates MAPK, and the MEK kinase, which is responsible for the activation of MEK (Robinson and Cobb, 1997 ). Three major MAPK pathways have been described in mammals. The extracellular signal-regulated kinase (Erk) 1 and 2 route is activated by receptors for polypeptide growth factors, G protein-coupled receptors, or by directly stimulating intracellular pathways such as the PKC or calcium messenger systems (Garrington and Johnson, 1999 ; Widmann et al., 1999 ). The two other MAPK routes, the p38 and the Jun N-terminal kinase (JNK) pathways, are mainly triggered by cytokine and stress stimuli (Ip and Davis, 1998 ; Nebreda and Porras, 2000 ). All MAPKs possess overlapping substrate specificities. MAPKs are proline-directed kinases that phosphorylate target proteins at serine or threonine residues with a minimum sequence of Thr/Ser-Pro. In addition, a Pro residue in the −2 position facilitates the preference of MAPKs for their substrates (Widmann et al., 1999 ). Besides containing a potential phosphorylation site, MAPKs appear to recognize their targets by interaction with docking sites in substrates, which often include a positively charged region (Tanoue et al., 2000 ).
During studies aimed at elucidating the mechanisms controlling the regulated cleavage of the TrkA neurotrophin receptor, we found that multiple MAPK pathways act as intermediates in PKC- and stress-induced TrkA cleavage. PKC-induced cleavage was impaired in cells derived from mice expressing an inactive form of the secretase TACE. However, stress-induced cleavage occurred in these cells, suggesting that TrkA cleavage is a complex process regulated by different proteases and signaling pathways. The MAPK Erk and TACE coprecipitated, and Erk was able to phosphorylate TACE at threonine 735. These studies identified TACE as a novel Erk substrate and indicate that MAPKs, in addition to their roles as intermediates in membrane-nuclear signaling, may also regulate inside-out signaling by directly controlling the function of membrane secretases.
The mAb MGR12 was obtained from Dr. S. Ménard (Istituto Nazionale dei Tumori, Milan, Italy), the M220 anti-TACE antibody was from Dr. R.A. Black (Immunex, Seattle, WA), and the mAb anti-hemagglutinin (HA) was from BABCO (Richmond, CA). Various polyclonal antisera were raised by injecting peptides corresponding to the C terminus of human TrkA, TACE, or p38 into rabbits. The anti-panTrk antiserum has been described (Cabrera et al., 1996 ). The anti-TACE antiserum corresponded to the C terminus of human TACE (NH2-CKLQRQNRVDSKETE-COOH). The anti-p38 antiserum was raised to the peptide NH2-CSQERPTFYRQELN-COOH. The anti-pErk, anti-JNK, anti-pJNK, and anti-Erk2 were from Santa Cruz Biotechnology (Santa Cruz, CA). The anti-pp38 and the anti-pTP antibodies were from New England Biolabs (Beverly, MA). Immunoprecipitations and Western analyses were performed as described by Cabrera et al. (1996) . EGF was from Collaborative Research (Bedford, MA), and NGF was from UBI (Lake Placid, NY).
A TACE form mutated in the putative phosphorylation site that substituted threonine 735 for alanine (TACE-T735A) was generated by oligonucleotide-directed mutagenesis. Extracellularly HA-tagged TACE and TACE-T735A were obtained by insertion of a PCR-amplified HA-coding fragment into the naturally occurring EcoRI restriction site located in the ectodomain of TACE. The tagged constructs were subcloned into the pCDNA3 vector and were used to transfect 293 cells. Clones were selected with geneticin (500 μg/ml). In parallel, these forms of TACE were subcloned into the BamHI-XhoI sites of the pLZR-internal ribosomal entry site (IRES)-green fluorescent protein (GFP) retroviral vector. A dominant negative kinase dead form of Erk2 tagged at the N terminus with an HA epitope (HA-Erk2K52R) was provided by Dr. P. Crespo (Instituto de Investigaciones Biomédicas, Madrid, Spain). For the production of this dominant negative form in retrovirus, a 1.25-kb fragment obtained by digestion with HindIII-PstI, was blunt ended and ligated to the pLZR-IRES-GFP vector after XhoI digestion of the vector and blunt end creation with Klenow.
TACEΔZn/ΔZn fibroblasts (Peschon et al., 1998 ) were transfected by lipofection (LipofectAMINE, Life Technologies, Carlsbad, CA) and clones and pools selected with puromycin (4 μg/ml) or infected with retrovirus. For the latter, 293T cells were plated at 1.8 × 106 cells/60-mm dish. The following day, and ~5 min before transfection, 25 μM chloroquine was added to each plate. The transfection solution was DNA (2.5 μg of pMD-G, 5 μg of pNGVL-MLV-gag-pol, 3 μg of retroviral vector), 2 M CaCl2 (61 μl), and double distilled H2O to 500 μl. Once mixed, the solution was bubbled for 15 s before its addition to the cells. Twenty-four hours later, the medium was replaced with 3 ml of fresh “virus-collecting medium,” and retrovirus was recovered 1 d later. Target cells were then infected with viral supernatants containing 6 μg/ml Polybrene (Sigma, St. Louis, MO).
GST-Erk, GST-Elk, GST-TACE, or GST-TACE-T735A, were generated according to standard protocols (Guan and Dixon, 1991 ). For precipitation experiments, 30 μg of the corresponding uneluted GST-fusion protein were incubated for 1 h at 4°C in 25 mM HEPES, pH 7.4; 1 mM EDTA; and 1 mM dithiothreitol, before the addition of lysates from treated cells. The lysates were then incubated for 2 h with the beads at 4°C, and then complexes were washed three times with ice-cold phosphate-buffered saline (PBS). Electrophoretic and blotting analyses were performed as described above.
Extracts from HeLa or 293 cells treated with or without phorbol 12-myristate 13-acetate (PMA) were immunoprecipitated with the anti-Erk2 antibody, and immunocomplexes were washed three times with lysis buffer. A final wash was performed with kinase buffer (20 mM HEPES, pH 7.6; 20 mM MgCl2; 25 mM β-glycerophosphate; 0.1 mM sodium orthovanadate; 2 mM dithiothreitol). The immunocomplexes were then incubated for 30 min at 30°C with the substrate in the kinase buffer containing 15 μM ATP, 1 μCi of (γ-32P)ATP, and 10 μg of GST-TACE or GST-TACE-T735A. To detect whether GST-Erk was able to phosphorylate GST-TACE or GST-TACE-T735A, the kinase reaction was performed as described, but 10 μg of GST-Erk were included in the reaction. Samples were electrophoresed in 10% SDS-PAGE gels that were dried. Bands in gels were detected by autoradiography.
The method for two-dimensional phosphoamino acid analysis has been described in detail (Cooper et al., 1983 ). Briefly, in vitro phosphorylated GST-TACE or GST-TACE-T735A was excised from dried gels and rehydrated in 50 mM ammonium bicarbonate (1.2 ml). Then, 50 μl of β-mercaptoethanol and 10 μl of 10% SDS were added, and the samples were boiled for 5 min. Proteins were eluted overnight at 37°C and precipitated with trichloroacetic acid, and pellets were washed with ethanol. Samples were dissolved in 6 M HCl, incubated for 60 min at 110°C, and dried. The hydrolysis products were separated in thin-layer chromatography plates by electrophoresis at pH 1.9 (formic acid:acetic acid:water, 50:156:1794) for 35 min at 1.5 kV, followed by a second dimension run at pH 3.5 (acetic acid:pyridine:water, 10:1:189) for 22 min at 1.3 kV, and then exposed to autoradiography.
293 cells were incubated for 3 h in phosphate-free DMEM, and then fresh medium containing 0.5 mCi/ml [32P]orthophosphoric acid was added. Cells were labeled for 3.5 h and then treated as indicated. Monolayers were washed twice in PBS and lysed in 10 mM Tris, pH 8.0; 150 mM NaCl; 1% Triton X-100; 0.1 mM sodium orthovanadate; 1 mM NaF; 1 mM phenylmethylsulfonyl fluoride; 1 μg/ml pepstatin; 10 μg/ml aprotinin; and 1 μg/ml leupeptin. Samples were then immunoprecipitated for 2 h with the anti-TACE antibody. Protein A was added for the last 20 min, and immunocomplexes were washed three times with 10 mM Tris, pH 8.0; 150 mM NaCl; and 0.1% Triton X-100, once in the same buffer supplemented with 0.5 M NaCl, and twice more in the first buffer. Samples were electrophoresed in 6% SDS-PAGE gels and dried, and bands were detected by autoradiography.
TACEΔZn/ΔZn-TrkA cells were plated at a density of 106 cells/60-mm plate and infected with retrovirus that included the pLZR-IRES-GFP, pLZR-TACE-T735A-IRES-GFP, or pLZR-TACE-IRES-GFP vectors, as described above. The cells were infected for 24 h and then replated into two 100- or 60-mm dishes. Two days later, cells were treated or not for 30 min with PMA and then lysed to be immunoprecipitated with anti-panTrk antibodies, followed by Western analyses with the same antibodies.
To visually inspect TrkA cleavage, TACEΔZn/ΔZn-TrkA cells were infected with the wild-type TACE retrovirus, and then cells were plated on coverslips. Cells treated or not with PMA were then washed with PBS and fixed in p-formaldehyde. Monolayers were washed twice in PBS supplemented with 0.1% Triton X-100 (final concentration; PBST), blocked in PBST with 5% bovine serum albumin for 1 h at room temperature, and then incubated for 2 h with the MGR12 mAb anti-TrkA ectodomain antibody, and with the anti-panTrk polyclonal affinity-purified antibody. After two washes of 15 min each in PBST, the coverslips were incubated for 30 min with a mixture of anti-mouse-Cy5 and anti-rabbit-Cy3-conjugated secondary antibodies. The coverslips were washed three times, 5 min each, in PBST and mounted. Samples were then analyzed for the presence of TrkA by immunofluorescence using an LSM510 confocal laser scanning microscope (Zeiss, Thornwood, NY). To avoid interference between fluorescence signals, the images were captured under multitracking mode.
Regulated TrkA cleavage can be easily followed in Chinese hamster ovary (CHO) cells transfected with the cDNA coding for the human version of the receptor (CHOTrkA cells; Cabrera et al., 1996 ). Immunoprecipitation and Western blot analysis of CHOTrkA cell lysates with an anti-panTrk antiserum raised to the cytosolic C terminus of TrkA identified two forms of the receptor: gp110TrkA and gp140TrkA (Figure (Figure1A).1A). In addition to these holoreceptor forms, lower Mr fragments (p41 and p40) were identified by this antiserum in cells treated with the PKC activator PMA (Figure (Figure1A;1A; Cabrera et al., 1996 ; Diaz-Rodriguez et al., 1999 ). These lower Mr forms represent truncated forms of the receptor that contain the transmembrane and cytosolic domains (Cabrera et al., 1996 ; Diaz-Rodriguez et al., 1999 ).
During studies to evaluate whether the TrkA-processing protease was membrane bound, we noticed that treatment of cultured cells with high salt (1 M NaCl) stimulated processing of the receptor (Figure (Figure1A). 1A). Treatment of CHOTrkA cells with sorbitol mimicked the effect of NaCl on TrkA cleavage (Figure (Figure2;2; Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data), indicating that changes in medium osmolarity can stimulate cleavage independently of the effect of NaCl on membrane potential. Because changes in medium osmolarity had been reported to regulate the shedding of l-selectin through MAPK routes (Rizoli et al., 1999 ), the above data raised the possibility that TrkA cleavage induced under these treatments could be mediated by the activation of these routes. To further explore this possibility we used PD98059 and U0126, two drugs that inhibit Erk upstream-activating kinases (Dudley et al., 1995 ; Favata et al., 1998 ) and have been reported to affect the shedding of other membrane-bound molecules (Fan and Derynck, 1999 ; Gechtman et al., 1999 ). Preincubation with PD98059 or U0126 resulted in a significant inhibition of PMA-induced TrkA cleavage (Figure (Figure1B).1B). The extent of this inhibition varied from 30–90% depending on the cell line (see below; Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data). In addition, transfection of TrkA-expressing 293 cells (293TrkA cells) with a dominant negative form of Erk2 partially inhibited PMA-induced TrkA cleavage (Figure (Figure1C). 1C).
To investigate whether different MAPKs were involved in TrkA cleavage, specific inhibitors of the Erk1/2 and p38 MAPK routes were used, combined with an analysis of the activation of different MAPKs using phosphorylation state-specific antibodies. PMA treatment resulted in Erk1/2 phosphorylation (Figure (Figure2A,2A, second left panel). In contrast, this treatment did not significantly stimulate p38 or JNK phosphorylation (Figure (Figure2A,2A, fourth and sixth panels). Induction of stress by sorbitol or UV irradiation stimulated the p38 and JNK pathways (Figure (Figure2A,2A, fourth and sixth panels). Like PMA, sorbitol was also able to stimulate Erk1/2 phosphorylation (Figure (Figure2A,2A, second left panel). Erk1/2 activation by phorbol esters or sorbitol was prevented by preincubation with PD98059, indicating that both agents act on Erk upstream-activating kinases (Figure (Figure2A,2A, second central panel). In contrast, SB203580, an inhibitor of the p38 pathway that acts downstream of the p38 kinase (Cohen, 1997 ), did not prevent PMA or sorbitol-induced Erk1/2 activation (Figure (Figure2A,2A, second right panel). The action of sorbitol on Erk1/2 activation was not mediated by stimulation of PKC, as indicated by the failure of the PKC inhibitor bis-indolylmaleimide (BIM) to prevent Erk1/2 activation (Figure (Figure2C). 2C). BIM was also unable to prevent sorbitol-induced TrkA processing, indicating that the action of sorbitol on TrkA cleavage was independent of PKC.
Preincubation with PD98059 significantly inhibited PMA-induced TrkA cleavage (Figure (Figure2A,2A, top central panel). SB203580 largely inhibited sorbitol-induced TrkA cleavage, leaving unaffected the ability of PMA to induce TrkA cleavage (Figure (Figure2A,2A, top right panel). Preincubation with both PD98059 and SB203580 had a more profound inhibitory effect on sorbitol-induced TrkA cleavage than treatment with each inhibitory drug alone (Figure (Figure2B),2B), indicating that the action of sorbitol on TrkA cleavage probably depends on the activation of both p38 and Erk1/2. UV irradiation, which activated the p38 and JNK routes, also induced TrkA cleavage (Figure (Figure2A,2A, top left panel). This effect was abolished by preincubation with SB203580 (Figure (Figure2A,2A, top right panel), indicating that UV irradiation stimulated TrkA cleavage through a p38-dependent route.
Hydroxamic acid-derived metalloprotease inhibitors can prevent the release of the ectodomain of several membrane proteins (Blobel, 1997 ; Black and White, 1998 ). To investigate whether the different MAPK routes converged in the activation of analogous protease activities, we used the metalloprotease inhibitor BB3103. As shown in Figure Figure2D, 2D, preincubation of cells with this inhibitor prevented the stimulated cleavage of TrkA by PMA or sorbitol. The effect of BB3103 was dose dependent with IC50 values (250 nM for PMA and 800 nM for sorbitol, Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data) known to specifically act on metalloproteases (Gearing et al., 1994 ).
The protease TACE/ADAM17 has been isolated with metalloprotease inhibitors of the hydroxamic acid family (Black et al., 1997 ; Moss et al., 1997 ). This secretase has been implicated in the shedding of several membrane proteins (Peschon et al., 1998 ). To analyze whether this protease was involved in the cleavage of TrkA, we first investigated whether TrkA and TACE colocalized. For this, an antibody toward the intracellular C terminus of TACE was raised. In human 293 cells this antibody recognized two bands whose presence was prevented by preincubation with the peptide used for immunization (Figure (Figure3A).3A). The amount of the faster migrating form, which may represent immature or truncated TACE, was variable (Figure (Figure3,3, A and C). These TACE forms were also recognized by the M220 mAb, which has previously been shown to interact with the ectodomain of human TACE (Black et al., 1997 ; Doedens and Black, 2000 ). Immunofluorescence staining with the purified polyclonal antibody showed that TACE accumulated in several locations, including the plasma membrane, the perinuclear area, and other cytosolic locations where staining was characterized by a dotted pattern (Figure (Figure3B).3B). This staining was prevented by preincubation of the anti-TACE antibody with the TACE peptide, and no staining was observed when using another affinity-purified control antibody (Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data). TrkA staining at the cell periphery was more prominent than that of TACE, and the receptor also accumulated in an intracellular perinuclear region (Figure (Figure3B).3B). Merging of both immunofluorescence images indicated that TACE and TrkA colocalized at the plasma membrane and in the perinuclear area. Interestingly, TrkA staining was excluded from the regions of the cytoplasm where a dotted distribution of TACE was observed. Treatment of cells with PMA or sorbitol did not substantially affect the cellular distribution of TACE (Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data), even though a decrease in the amount of the protease could be detected in cells treated with PMA (Figure (Figure4;4; Doedens and Black, 2000 ).
To biochemically analyze whether TACE was involved in the cleavage of TrkA, we used a fibroblastic cell line (TACEΔZn/ΔZn cells) derived from mice homozygous for a form of TACE in which the metal binding pocket of the protease had been deleted (Reddy et al., 2000 ). This deletion leaves a continuous in-frame sequence. As a consequence, TACEΔZn/ΔZn cells expressed a TACE form with an Mr below that of TACE from wild-type animals (Figure (Figure3C).3C). The effect of agents that in CHOTrkA and 293TrkA cells stimulated holoreceptor cleavage by independent MAPK routes was analyzed in several distinct clones of TACEΔZn/ΔZn cells transfected with TrkA. Treatment with PMA did not induce cleavage of the holoreceptor in TACEΔZn/ΔZn-TrkA cells (Figure (Figure3D).3D). In these cells, high salt or sorbitol stimulated TrkA cleavage, indicating that cleavage mechanisms other than those induced by PKC were preserved in TACEΔZn/ΔZn-TrkA cells. PMA stimulated Erk1/2 phosphorylation in TACEΔZn/ΔZn-TrkA cells (Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data), demonstrating that the failure of the phorbol ester to induce TrkA cleavage was not due to inefficient activation of the Erk1/2 pathway.
To reconstitute TrkA cleavage in TACEΔZn/ΔZn-TrkA cells, we infected these cells with retrovirus that included a bicistronic vector coding for TACE and GFP (pLZR-TACE-IRES-GFP). Immunofluorescence experiments indicated that >99% of the cells that stained strongly for TACE also exhibited a strong GFP signal (Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data). Retroviral infection forced the expression of wild-type TACE to levels several times higher than those for endogenous TACE-ΔZn2+ (Figure (Figure3E,3E, bottom). In cells expressing wild-type TACE, treatment with PMA induced p41/40 generation, in contrast to cells infected with the retrovirus containing the empty vector pLZR-IRES-GFP (Figure (Figure3E,3E, top). That TACE could reconstitute TrkA cleavage in TACEΔZn/ΔZn-TrkA cells was also supported by immunofluorescence experiments (Figure (Figure3F).3F). For these experiments, we used the same TACEΔZn/ΔZn-TrkA clone that was used for the biochemical experiments. Immunofluorescence staining with the anti-TrkA antibodies showed that, together with TrkA-expressing cells, this clone also contained a cell population that had lost TrkA expression (Figure (Figure3F).3F). Cells were infected with the retrovirus containing the pLZR-TACE-IRES-GFP vector, treated (Figure (Figure3F,3F, bottom) or not (Figure (Figure3F,3F, top) with PMA. Then cells were analyzed for TrkA content by staining with the anti-ectodomain and anti-endodomain antibodies. As previously reported (Cabrera et al., 1996 ; Diaz-Rodriguez et al., 1999 ), TrkA ectodomain cleavage can be followed by loss of anti-ectodomain fluorescence at the cell periphery with preservation of peripheral staining with the anti-endodomain antibody. In cells expressing very low levels of TACE (Figure (Figure3F,3F, top and bottom, arrow) TrkA at the cell periphery could easily be detected by the anti-TrkA ectodomain antibody. This was observed in both untreated (top) and PMA-treated (bottom) samples. In contrast, in cells expressing TrkA and high GFP levels (Figure (Figure3F, 3F, bottom, asterisks), PMA induced a substantial decrease in TrkA ectodomain staining, while still preserving peripheral staining by the anti-TrkA endodomain antibody. Collectively, these results indicate that TACE acts as a protease that cleaves TrkA in response to PMA.
MAPKs are proline-directed kinases that phosphorylate target proteins at serine or threonine residues within a minimum consensus sequence of Thr/Ser-Pro (Schaeffer and Weber, 1999 ; Widmann et al., 1999 ). In addition, a Pro residue at the −2 position facilitates the preference of MAPKs for their substrates (Widmann et al., 1999 ). Inspection of the TACE intracellular domain showed the presence of a ProGln-Thr735-Pro motif that fits with a potential MAPK phosphorylation site, raising the possibility that TACE could be a direct MAPK substrate (Figure (Figure4A).4A). In vitro kinase assays using recombinant Erk (GST-Erk) or Erk immunoprecipitated from cells stimulated with phorbol esters showed that Erk could induce phosphorylation of a GST-intracellular domain of TACE fusion protein (Figure (Figure4A).4A). The efficiency of phosphorylation of TACE was lower than that of a classical Erk substrate such as Elk (Figure (Figure4A).4A). Substitution of threonine 735 for alanine in the intracellular domain of TACE (GST-TACE-T735A) strongly reduced the ability of anti-Erk immunoprecipitates to cause in vitro phosphorylation of the mutated GST-TACE fusion protein (Figure (Figure4B).4B). Phosphoamino acid analysis of GST-TACE and GST-TACE-T735A supported the idea that the major site phosphorylated by Erk was T735 (Figure (Figure4C). 4C). These analyses also demonstrated that PMA stimulated a serine kinase that coprecipitated with Erk and was able to stimulate serine phosphorylation of GST-TACE and GST-TACE-T735A. This serine kinase phosphorylated the part of the fusion protein corresponding to TACE, because GST alone failed to be phosphorylated by the anti-Erk immunoprecipitates or GST-Erk (Figure (Figure44A).
Immunoprecipitation of TACE from 293 cells labeled with 32P indicated that the metalloprotease was phosphorylated under resting conditions (Figure (Figure4D).4D). Treatment with PMA increased the phosphorylation of TACE (Figure (Figure4D)4D) and caused its down-regulation (Figure (Figure5B;5B; see also Doedens and Black, 2000 ). The resting and PMA-induced increases in TACE phosphorylation were inhibited by preincubation with PD98059 (Figure (Figure4D,4D, bottom), suggesting that the Erk pathway participates in the control of the phosphorylation of TACE in vivo. To analyze whether T735 was phosphorylated in vivo, we took advantage of the fact that T735 is the unique Thr residue followed by a Pro in the intracellular domain of TACE (Black et al., 1997 ; Moss et al., 1997 ) and used an antibody that allows detection of phospho-threonine-proline (pTP) epitopes. In vitro kinase experiments indicated that this antibody reacted with GST-Elk and GST-TACE, but only when ATP was present in the incubation buffer, indicating that the antibody recognized only the phosphorylated version of these fusion proteins (Figure (Figure5A). 5A). Recognition of GST-Elk or GST-TACE was, however, prevented when the anti-pTP antibody was preincubated with an excess of phosphothreonine (Figure (Figure5A).5A). In addition, the antibody failed to react with the GST-TACE-T735A mutant (Figure (Figure5B)5B) or with GST (Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data). Western blotting of anti-TACE immunoprecipitates from 293 cells with the anti-pTP antibody indicated that TACE was phosphorylated at T735 under resting conditions, and PMA treatment increased phosphorylation at this site (Figure (Figure5,5, C and D). Preincubation of PMA-treated extracts with an excess of the peptide used for the generation of anti-TACE antibodies (Figure (Figure5C)5C) or preincubation of the anti-pTP antibody with phosphothreonine (Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data) prevented the reaction of the anti-pTP antibody with TACE. Treatment of 293 cells with PD98059 decreased resting pT735P TACE phosphorylation and strongly prevented the PMA-induced increase in the phosphorylation of TACE at this site (Figure (Figure5D).5D). In agreement with the in vitro data, an HA-tagged version of TACE mutated at T735 failed to be recognized by the anti-pTP antibody when transfected into 293 cells (Figure (Figure5E). 5E).
The above data strongly indicated that TACE was phosphorylated by Erk1/2 at T735. However, because PD98059 is an inhibitor of MEK1/2 (Dudley et al., 1995 ), it was possible that a substrate of the latter, other than Erk1/2, or even MEK1/2, could be responsible for TACE phosphorylation at T735. To test this possibility a dominant negative form of Erk2 was used. This form was created by point mutation (K52→R52) in the ATP-binding pocket of the molecule. This mutation destroys the kinase activity of Erk2 but preserves MEK1/2-mediated dual phosphorylation at the Thr-Glu-Tyr microdomain within the Erk2 activation loop (see below). Infection of 293 cells with a retrovirus containing a vector for the expression of HA-tagged Erk2K52R (HA-Erk2K52R) resulted in the expression of the dominant negative form of the protein (Figure (Figure5F,5F, top) to levels above those of endogenous Erk2 (Figure (Figure5F,5F, second panel from top). Treatment with PMA caused a change in mobility of both endogenous Erk2 and infected HA-Erk2K52R (Figure (Figure5F,5F, second panel from top) and induced their dual phosphorylation (Figure (Figure5F,5F, third panel from top). In vitro kinase studies using GST-Elk as a substrate indicated that expression of HA-Erk2K52R strongly reduced Erk2 activation in response to PMA (Figure (Figure5F,5F, fourth panel from top). Expression of HA-Erk2K52R was also able to substantially decrease resting and PMA-induced TACE phosphorylation at T735 (Figure (Figure5F,5F, second panel from bottom).
In addition to PMA, activation of receptor tyrosine kinases, such as the EGF receptor (EGFR; Fan and Derynck,  or TrkA, Cabrera et al. ; Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data), has been reported to trigger ectodomain shedding. To investigate whether receptor tyrosine kinase activation could also induce phosphorylation of TACE at T735, 293TrkA cells were treated with EGF or nerve growth factor (NGF). In these cells, EGF (acting through the endogenous EGFR) or NGF (acting through the transfected TrkA receptor) caused Erk1/2 activation (Figure (Figure5G, 5G, middle) and induced TACE phosphorylation at T735 (Figure (Figure5G,5G, top). The effect of NGF on TACE phosphorylation at T735 was consistently more pronounced than that caused by EGF (Figure (Figure5G;5G; Díaz-Rodríguez, Montero, Esparís-Ogando, Yuste, and Pandiella, unpublished data), probably because of the higher complement of TrkA receptors of 293TrkA cells compared with the amount of endogenous EGFR. Retrovirus-mediated expression of TACE-T735A in 293TrkA cells substantially reduced the ability of NGF to induce TACE phosphorylation at T735 (Figure (Figure55H).
Because Erk was able to phosphorylate TACE, we investigated whether these molecules interacted. As shown in Figure Figure6A,6A, the treatment of cells with PMA favored interaction of Erk with GST-TACE. Erk also interacted with GST-TACE-T735A, although less efficiently than with GST-TACE. These results indicate that Erk and TACE interact in vitro and that T735 is important, but not critical, for the docking of Erk to the intracellular domain of TACE. To analyze whether Erk and TACE interacted in vivo, 293TrkA cells were incubated with PMA and then lysates were immunoprecipitated with anti-TACE or anti-proTGFα (used as a control) antibodies. Immunoprecipitates were analyzed by SDS-PAGE followed by Western blotting with anti-p-Erk (Figure (Figure6B,6B, left) or anti-TACE (Figure (Figure6B,6B, right) antibodies. p-Erk1/2 coprecipitated with the anti-TACE immunoprecipitates but not with the anti-proTGFα immunoprecipitates. The reverse use of these antibodies confirmed that Erk and TACE coprecipitated (Figure (Figure6C)6C) and their interaction was favored by the activation of Erk by PMA. TACE and Erk associated in TACEΔZn/ΔZn cells, as well as in 293TrkA cells treated with BB3103, indicating that TACE was not required in its active form for its association to Erk (Figure (Figure6D). 6D).
To investigate the functional importance of TACE phosphorylation at T735, TACEΔZn/ΔZn-TrkA cells were infected with TACE or TACE-T735A and TrkA cleavage in response to PMA analyzed by Western blotting. As shown above (Figure (Figure3E), 3E), wild-type TACE rescued PMA-induced TrkA cleavage (Figure (Figure7A).7A). TACE-T735A also rescued PMA-induced TrkA cleavage, although to a lesser extent than wild-type TACE (Figure (Figure7A).7A). Considering the rescue obtained with wild-type TACE as 100%, densitometric quantitation of different experiments (n = 7) indicated that TACE-T735A rescued cleavage of TrkA by 59% ± 12, i.e., 41% less than wild-type TACE. The extent of this inhibition was analogous to the effect that U0126 had in these cells on PMA-induced TrkA cleavage (Figure (Figure7B).7B). These results demonstrate that the phosphorylation of TACE at T735 facilitates the accumulation of truncated fragments of TrkA, especially in cells treated with PMA. In addition, these experiments indicate that other Erk-independent routes regulate TACE activity by PMA.
In this work we report that the Erk and p38 MAPK routes act as mediators in the cleavage of TrkA induced by PKC activation or stress, respectively. We also provide evidence indicating that PMA-induced TrkA cleavage involves Erk activation. Activation of Erk facilitates its interaction with the membrane secretase TACE. We identify TACE as a novel Erk substrate and show that phosphorylation of TACE at T735 is important for the regulation of TACE activity by Erk.
Previous work on TrkA has indicated that PKC enzymes regulate holoreceptor cleavage (Cabrera et al., 1996 ). The use of inhibitors of the Erk1/2 pathway demonstrated that Erk acted as an intermediate in the action of PKC on TrkA cleavage. However, because the block was only partial, these data indicate that PKC regulates TrkA shedding by Erk-dependent and Erk-independent routes, a conclusion also supported by the reconstitution data obtained with the TACE- T735A mutant. Cleavage of TrkA was also induced by UV, which did not significantly induce Erk1/2 activation but stimulated p38 and JNK, indicating that other MAPK routes may regulate TrkA cleavage. This was supported by the use of the p38 pathway inhibitor SB203580. This drug, which did not affect PMA-induced TrkA cleavage, had, however, a profound effect on UV-induced receptor processing, indicating that the p38 kinase pathway mediated the action of UV on TrkA cleavage. In addition to TrkA, the p38 route has also been implicated in the shedding of l-selectin (Rizoli et al., 1999 ) and proTGFα (Fan and Derynck, 1999 ), supporting the idea that this MAPK route may play a generic role in the control of membrane protein ectodomain cleavage. In a more ample context, the data obtained with different activators of cleavage together with inhibitory drugs of distinct MAPK routes suggest that MAPKs regulate the cleavage of membrane proteins by independent routes.
Our data answered the question of whether a single secretase could be activated by all the stimuli that induce cleavage. In cells expressing an inactive form of TACE, TrkA cleavage was defective, suggesting a role of this protease in PMA-induced TrkA shedding. That this secretase had a role in TrkA processing was further supported by the rescue of TrkA cleavage in response to PKC activation in TACEΔZn/ΔZn cells in which wild-type TACE was reintroduced. However, secretases other than TACE may also cleave membrane proteins, because osmotic stress was able to stimulate TrkA cleavage in these cells. These findings, besides suggesting a role of TACE as a mediator of PMA-induced TrkA cleavage, also indicate that the osmotic stress-activated protease is insensitive to PKC/Erk-induced activation but is a target of the p38 route. The nature of the p38-activated (and PKC/Erk-insensitive) secretase is presently unknown. Thus, an important conclusion that emanates from these data is that a single secretase or group of secretases may be activated by some (but not all) of the intracellular routes that trigger membrane protein ectodomain cleavage.
Insights into the mechanism by which MAPKs regulate membrane protein ectodomain cleavage have been obtained by the study of TACE as a target of MAPKs. In vitro analysis indicated that the protease was phosphorylated not only by recombinant Erk but also by endogenous Erk. Substitution of the potential MAPK phosphorylation site (Thr735 for Ala) in TACE profoundly inhibited the ability of recombinant TACE to be phosphorylated by activated Erk, suggesting that this residue could act as a target for MAPK phosphorylation. This was further supported by the in vivo phosphorylation studies, which showed resting, growth factor, and PMA-stimulated TACE phosphorylation at T735. Besides acting as a phosphorylation site, T735 was found to contribute to the interaction of TACE and Erk. In fact, association of Erk with TACE was favored by the presence of T735, indicating that T735 participates in the docking of Erk to the TACE intracellular domain. This latter role, together with the fact that T735 phosphorylation is modestly stimulated by PMA, indicates that phosphorylation of TACE at T735 may not be the only signal that controls its activity, and docking of Erk may also be important, perhaps allowing Erk to target other TACE-bound molecules.
A functional role of TACE phosphorylation at T735 was indicated by the effect of TACE and TACE-T735A on the amount of truncated fragments of TrkA. In TACEΔZn/ΔZn cells, TACE and TACE-T735A caused accumulation of the truncated fragments of TrkA, especially when cells were treated with PMA. However, the mutated form of TACE was less efficient than the wild type. This effect has also been found when analyzing the effect of TACE and TACE-T735A on the generation of truncated fragments of the membrane-anchored growth factor proNeuregulinα2c (J.C. Montero, L. Yuste, E. Díaz-Rodríguez, A. Esparís-Ogando, and A. Pandiella, unpublished data). How can TACE phosphorylation at T735 regulate the amount of the cell-bound truncated fragments? A mechanism could include the potential regulation of the catalytic activity of TACE by phosphorylation at T735. However, PMA-induced shedding of TNF-α has been reported to occur independently of the cytosolic tail of TACE (Reddy et al., 2000 ). On the other hand, TACE phosphorylation at T735 could regulate the amount of the cell-bound truncated fragments by mechanisms not related to the control of TACE activity. Thus, TACE could favor the stabilization of the truncated fragments by decreasing their turnover and the TACE-T735A mutant could be less effective. However, we have failed to detect any significant difference in the resting accumulation of truncated fragments of TrkA between TACEΔZn/ΔZn-TrkA cells infected with TACE or TACE-T735A. Another possibility could be the participation of T735 phosphorylation in the maturation of TACE with generation of the processed, active form of the enzyme. In this respect, it is interesting that the phosphorylated form of TACE detected in 293 cells corresponds to unprocessed TACE. Finally, another mechanism that could explain why TACE causes accumulation of TrkA and proNRGα2c fragments better than TACE-T735A could be related to a potential sorting defect of the mutated TACE. This latter possibility seems unlikely because cell surface immunoprecipitation of HA-tagged forms of TACE and TACE-T735A reveal identical exposure of wild type and the mutant protein at the cell surface (L. Yuste and A. Pandiella, unpublished observations). A detailed study of TACE phosphorylation, maturation, sorting/trafficking, and shedding activity will be required to address the above possibilities.
In summary, our data indicate that multiple MAPK routes independently regulate membrane protein ectodomain cleavage and show that secretases are able to discriminate between the ample spectrum of stimuli that induce cleavage. In addition, we identified TACE as a novel Erk substrate and described a potential mechanism of activation of TACE by Erk. The future identification of other functional secretases and their cellular targets will allow us to evaluate to which extent direct phosphorylation by MAPKs regulates the activity of other membrane secretases.
We particularly thank Drs. Jacques Peschon, and Roy. A. Black, from Immunex, for providing us with the TACEΔZn/ΔZn cells, the M220 antibody, and the mouse TACE cDNA. The supply of BB3103 by British Biotech is also acknowledged. We thank Dr. X. Bustelo for reading the manuscript and continuous encouragement. This work was funded by the European Community, the Fundación Ramón Areces, and the Spanish Ministry of Education and Culture. E.D-R. was supported by a fellowship from the Centro de Investigación del Cáncer. A.E-O. was supported by a postdoctoral contract from the Spanish Ministry of Education and Culture, and L.Y was supported by a predoctoral fellowship from the same Ministry.
Article published online ahead of print. Mol. Biol. Cell 10.1091/mbc.01–11–0561. Article and publication date are at www.molbiolcell.org/cgi/doi/10.1091/mbc.01–11–0561.