Proteolytic cleavage of the neural cell adhesion molecule by ADAM17/TACE is involved in neurite outgrowth

Authors


Address correspondence and reprint requests to Melitta Schachner, Zentrum für Molekulare Neurobiologie, Universität Hamburg, Martinistraße 52, D-20246 Hamburg, Germany. E-mail: melitta.schachner@zmnh.uni-hamburg.de

Abstract

The transmembrane and multidomain neural cell adhesion molecule (NCAM) plays important functional roles in the developing and adult nervous system. NCAM is proteolytically processed and appears in soluble forms in the cerebrospinal fluid and in serum under normal and pathological conditions. In this report, we present evidence that the metalloprotease a disintegrin and a metalloprotease (ADAM)17/tumour necrosis factor α converting enzyme (TACE) cleaves the polysialylated as well as the non-polysialylated transmembrane isoforms of NCAM, whereas the glycophosphatidylinositol-linked isoform of NCAM is not proteolytically cleaved. A truncated, enzymatically inactive mutant of TACE did not result in release of the NCAM110 cleavage product. Proteolytic cleavage was enhanced by a calmodulin-specific inhibitor and the actin-destabilizing agents cytochalasin D and latrunculin B. In contrast, the microtubule-stabilizing agent colchicine or microtubule-destabilizing agent paclitaxel did not affect the release of the 110-kDa fragment of NCAM. Neurite outgrowth from cerebellar microexplants was inhibited in the presence of the metalloprotease inhibitor GM 6001 on substrate-coated NCAM, but not on poly-l-lysine. Upon transfection of hippocampal neurones with an enzymatically inactive mutant of TACE, NCAM-stimulated neurite outgrowth was inhibited without affecting neurite outgrowth on poly-l-lysine, showing that proteolytic processing of NCAM by the metalloprotease TACE is involved in NCAM-mediated neurite outgrowth.

Abbreviations used
ADAM

a disintegrin and a metalloprotease

CHO

Chinese hamster ovary

CSF

cerebrospinal fluid

dTACE

deletion mutant of tumour necrosis factor-α-converting enzyme

DMEM

Dulbecco's modified Eagle's medium

Endo N

endoneuraminidase N

GFP

green fluorescent protein

GPI

glycosylphosphatidylinositol

NCAM

neural cell adhesion molecule

PBS

phosphate-buffered saline

PLL

poly-l-lysine

PSA

polysialic acid

SDS–PAGE

sodium dodecyl sulphate–polyacrylamide gel electrophoresis

TACE

tumour necrosis factor α converting enzyme

TNF

tumour necrosis factor

Neural cell adhesion molecule (NCAM) is involved in neuronal cell migration, neurite outgrowth, axon fasciculation, synaptogenesis, synaptic plasticity and emotional behaviour (for reviews see Schachner 1997; Ronn et al. 1998; Murase and Schuman 1999). In the central and peripheral nervous system NCAM is expressed at the surface of most, if not all, neural cell types. NCAM mediates neurite outgrowth by homophilic or heterophilic interactions between adjacent cells and with the extracellular matrix (for review see Crossin and Krushel 2000). This implicates NCAM both as a ligand and signal transducing receptor. NCAM, a member of the immunoglobulin superfamily, is composed of five immunoglobulin-like domains and two fibronectin-type III like repeats (for review see Cunningham et al. 1987). The three major isoforms are generated by alternative splicing of a primary transcript derived from a single gene and serve as neuritogenic ligands (Doherty et al. 1989, 1990) that have identical extracellular domains. Two isoforms, named NCAM180 and NCAM140, are transmembrane forms with an apparent molecular weight of 180 and 140 kDa respectively, whereas the third isoform of 120 kDa, NCAM120, is attached to the membrane via a glycosylphosphatidylinositol (GPI) anchor (He et al. 1986; Sadoul et al. 1986).

Besides these membrane-bound NCAM forms, several soluble NCAM forms with molecular weights ranging from 110 to 190 kDa have been observed in whole-brain supernatants (He et al. 1986; Nybroe et al. 1989; Probstmeier et al. 1989), neural cell culture supernatants (Sadoul et al. 1986; Nybroe et al. 1989), in cerebrospinal fluid (CSF) (Krog et al. 1992; Strekalova et al. 2005) and serum (Krog et al. 1992; Takamatsu et al. 1994). These studies showed that the predominant soluble form has an apparent molecular weight in the range of 110–115 kDa, whereas other forms with molecular weights ranging from 135 to 190 kDa are found in minor and variable amounts. The GPI-anchored NCAM120 is released from crude brain membranes (He et al. 1986) or from primary cerebellar neurones or astrocytes (Sadoul et al. 1986) upon treatment with phosphatidylinositol-specific phospholipase C. Incubation of brain membranes at 37°C results in the generation of a soluble 110-kDa fragment (Probstmeier et al. 1989). This soluble form of NCAM was identified as a binding partner of different extracellular matrix components (Probstmeier et al. 1992) and serves as a substrate cell adhesion molecule, as cells adhere to substrate-coated soluble NCAM. This cell adhesion was inhibited by the soluble NCAM fragment (Olsen et al. 1993). In addition, increased amounts of the fragment were found in perfusates of hippocampal slices after induction of long-term potentiation (Fazeli et al. 1994) and in CSF of patients with schizophrenia (Vawter et al. 2001) or dementia (Strekalova et al. 2005), underscoring a possible function in synaptic plasticity and behaviour. Thus, the mechanisms yielding the various soluble NCAM fragments and their functional roles remain to be elucidated in more detail in terms of cell biological consequences and behavioural outcomes in the intact organism. Interestingly, a transgenic mouse expressing the soluble extracellular domain of NCAM showed functional deficiencies and behavioural abnormalities suggesting that increased shedding of NCAM perturbs the synaptic connectivity of GABAergic interneurones, and may be involved in schizophrenia and other neuropsychiatric disorders (Pillai-Nair et al. 2005).

Recently, it has been shown that the release of NCAM is regulated by a metalloprotease and by ATP, and that the inhibition of release by a metalloprotease inhibitor reduces neurite outgrowth of hippocampal neurones (Hubschmann et al. 2005). Here, we investigated the role of proteolytic cleavage in the generation of the major soluble 110-kDa form and showed that generation of this form is mediated by a metalloprotease of the a disintegrin and metalloprotease (ADAM) family, namely ADAM17/tumour necrosis factor (TNF) α converting enzyme (TACE), and modulated by calmodulin inhibitors and agents stabilizing or destabilizing the actin cytoskeleton. Furthermore, we have demonstrated that NCAM-induced neurite outgrowth depends on this proteolytic cleavage.

Materials and methods

Reagents, antibodies and cDNA constructs

GM 6001, leupeptin, cytochalasin D, latrunculin B, paclitaxel and colchicine were purchased from Calbiochem (Bad Soden, Germany), and 1,10-phenanthroline from Sigma (Taufkirchen, Germany). The calmodulin inhibitor CGS 9343B was a gift from Novartis Consumer Health (Nyon, Switzerland). Polyclonal NCAM antibodies recognizing the extracellular domains of the three major isoforms of mouse NCAM were obtained from rabbits immunized with a protein A-purified NCAM fusion protein consisting of the extracellular domain of mouse NCAM and the Fc portion of human immunoglobulin (Chen et al. 1999). The antibody 735, directed against polysialic acid (PSA), and endoneuraminidase N (Endo N) were a gift from R. Gerardy-Schahn (University of Hannover, Hannover, Germany). The polyclonal antibody against ADAM17/TACE was a gift from S. J. Frank (University of Alabama, Birmingham, AL, USA). All secondary antibodies were obtained from Dianova (Hamburg, Germany). Rat NCAM140/pcDNA3 and rat NCAM180/pcDNA3 were provided by P. Maness, University of North Carolina, Chapel Hill, NC, USA). Rat NCAM120 cDNA (a gift from E. Bock, University of Copenhagen, Copenhagen, Denmark) was subcloned into the pcDNA3 vector (Invitrogen, Karlsruhe, Germany). The pcDNA3.1 vector encoding the mouse ADAM17/TACE was a gift from A. Pandiella (University of Salamanca, Salamanca, Spain).

Cell culture

Mouse neuroblastoma (Neuro 2a) cells, Chinese hamster ovary (CHO) cells and ADAM17/TACE-deficient and stably transfected ADAM17/TACE-positive fibroblasts from ADAM17/TACE-deficient mice (provided by Amgen, Thousand Oaks, CA, USA) were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum, 1 mm sodium pyruvate and antibiotics (penicillin 100 units/mL and streptomycin 100 µg/mL). For inhibitor studies cells (105−106 cells/well) were seeded into a six-well plate (NUNC, Wiesbaden, Germany) and grown in DMEM containing 10% fetal calf serum. After 24 h the medium was replaced with serum-free medium containing GM 6001 (10 µm), leupeptin (10 µm), CGS 9343B (10 µm), cytochalasin D (1 µm), colchicine (2 µm), latrunculin B (10 µm), paclitaxel (50 nm) or the corresponding solvents as controls. The cells were maintained for a further 24 h in the absence or presence of inhibitors. Cells and cell culture supernatants were collected separately. Cell culture supernatants were cleared by centrifugation for 1 h at 100 000 g and 4°C. After centrifugation, proteins from the supernatants were concentrated by acetone precipitation; one volume of each sample was incubated overnight with seven volumes of ice-cold acetone and then subjected to centrifugation for 30 min at 14 000 g and 4°C. Protein pellets were resuspended in sample buffer. Cells were homogenized in phosphate-buffered saline (PBS), pH 7.4, and centrifuged for 10 min at 1000 g and 4°C. For sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE), the supernatants were diluted with sample buffer.

Microexplant cultures

Microexplant cultures from mouse cerebella were prepared as described previously in detail (Kalus et al. 2003). Briefly, cerebella taken from 6–8-day-old C57BL/6J mice were forced through a Nitrex net with a pore size of 300 µm. Some 30–40 tissue pieces were plated on to poly-l-lysine (PLL)-coated glass coverslips (15 mm in diameter) or coverslips additionally coated with 2 µg/mL NCAM immunopurified from adult mouse brain (Pollerberg et al. 1985) and incubated in cell culture medium (minimal essential medium containing 10% horse serum, 10% fetal calf serum, 6 mm glucose, 200 µm l-glutamine, 50 U/mL penicillin, 50 µg/mL streptomycin, 10 µg/mL human transferrin, 10 µg/mL insulin and 10 ng/mL selenium). After 16 h, serum-free culture medium containing GM 6001 (20 µm), leupeptin (10 µm) or the corresponding solvent as control was added to the explants. After 24 h the explants were fixed and stained. Neurite outgrowth was quantified by measuring the length of the 10 longest neurites of 15 aggregates in each experiment with an IBAS Image analysis system (Kontron, Zeiss, Germany).

Cloning of the dominant-negative construct of ADAM17/TACE

First, a cDNA coding for the truncated mouse ADAM17/TACE from Ser474 to Cys827 was amplified by RT–PCR using mRNA isolated from mouse cerebellum as template and dADAM17/TACE up primer (5′-TTCTATGGATCCAGCAACAAGGTGTGGGC-3′; BamHI site underlined, ADAM17/TACE sequence in bold) and dADAM17/TACE down primer (5′-AAAGAGACAGAGTGCTAGGAATTCGCCAGC-3′; EcoRI site underlined, ADAM17/TACE sequence in bold). Three successive PCRs were run to amplify the truncated ADAM17/TACE following a Flag tag, a signal sequence and a Kozak sequence. In the first round, the PCR product obtained by RT–PCR was used as template and the dADAM17/TACE-3 primer (5′-GCACCCCGAGACTACAAAGACGATGACGACAAGAGCAACAAGGTGTGTGGCAACTCCAGG-3′; signal sequence in italics, Flag tag underlined, ADAM17/TACE sequence in bold) was used in combination with the dADAM17/TACE down primer. In the second round, the PCR product from the first round was used as template with the dADAM17/TACE-2 primer (5′-CTCATCCTGACCACTTTGGTGCCTTTCGTCCTGGCACCCCGAGACTACAAAGACGATGAC-3′; signal sequence in italics, Flag tag underlined) and dADAM17/TACE down. In the third round, the PCR product from the second round was taken for PCR using dADAM17/TACE down and the dADAM17/TACE-1 primer (5′-TTATTAGGATCCGCCACCATGAGGCGGCGTCGTCTCCTCATCCTGACCACTTTGGTGCCTTTC-3′; BamHI site underlined, Kozak sequence in bold, signal sequence in italics). The resulting PCR product was cloned into pcDNA3.1 (Invitrogen) using the BamHI and EcoRI sites. The plasmid construct containing the deletion mutant of mouse TACE (dTACE) lacking the prodomain and the catalytic domain was identical to the plasmid construct described previously for the human dTACE (Solomon et al. 1999).

Transient DNA transfection

Cells were transfected with the different vectors using Lipofectamine Plus (Invitrogen) following the manufacturer's instructions. After 24 h the transfection medium was replaced with serum-free medium and the cells were maintained for a further 24 h at 37°C. Cells and cell culture supernatants were collected separately. Cell culture supernatants were cleared by centrifugation for 1 h at 100 000 g and 4°C. After centrifugation, proteins from the supernatants were concentrated by acetone precipitation. Cells were homogenized in PBS, pH 7.4, and further centrifuged for 10 min at 1000 g and 4°C. The supernatants were subjected to western blot analysis as described previously (Kalus et al. 2003). Transfection of primary hippocampal neurones and determination of neurite length were carried out as reported previously (Dityateva et al. 2003).

In vitro assay of proteolytic processing

Total brains from adult mice were homogenized in 1 mm NaHCO3, 0.2 mm CaCl2, 0.2 mm MgCl2, 1 mm spermidine, pH 7.9, and centrifuged for 15 min at 600 g and 4°C. After re-centrifugation of the supernatant for 45 min at 25 000 g and 4°C, the pellet was resuspended in RPMI medium (PAA Laboratories, Cölbe, Germany) and incubated at 37°C with GM 6001 (10 µm), 1,10-phenanthroline (50 µm), leupeptin (10 µm), EDTA (5 mm), or the solvents as controls. The samples were then centrifuged for 1 h at 100 000 g and 4°C. Supernatants were concentrated by acetone precipitation and the resulting protein pellets were subjected to western blot analysis as described previously (Kalus et al. 2003).

Results

Soluble NCAM fragments are released from brain membranes by a metalloprotease

Previous studies have shown the existence of various soluble forms of NCAM. To determine whether the release of soluble NCAM fragments is due to proteolytic cleavage, a crude membrane fraction from young adult mouse brain was incubated at 37°C in the absence or presence of different protease inhibitors or chelators inhibiting divalent ion-dependent proteases. After pelleting the membranes by high-speed centrifugation, the supernatants were subjected to western blot analysis using a polyclonal NCAM antibody directed against the extracellular part of the three major isoforms of NCAM. A soluble 110-kDa fragment was mainly detectable in the supernatant in the absence of protease inhibitors (Fig. 1a, lane 1). After incubation of membranes in the presence of EDTA (Fig. 1a, lane 2) or in the presence of the metalloprotease inhibitor GM 6001 (Fig. 1a, lane 3), no soluble fragments were detectable in the supernatant. In the presence of the cysteine and serine protease inhibitor leupeptin (Fig. 1a, lane 4), the soluble 110-kDa fragment was present in the supernatant in amounts comparable to those observed in the absence of inhibitors. Only small amounts of this fragment were found in the supernatants obtained after incubation in the presence of 1,10-phenanthroline (Fig. 1a, lane 5), which inhibits metalloproteases by chelating divalent ions, in particular zinc. In summary, the results indicate that the soluble NCAM110 form is most probably generated by a zinc-dependent metalloprotease and not by a serine or cysteine protease.

Figure 1.

In vitro generation of soluble NCAM110 from brain membranes. A crude membrane fraction from mouse brain was incubated for 1 h in the absence (lane 1) or presence of EDTA (lane 2), GM 6001 (lane 3), leupeptin (lane 4) or 1,10-phenanthroline (lane 5). After centrifugation at 100 000 g the supernatants were subjected to western blot analysis using either a polyclonal antibody directed against NCAM (a) or a monoclonal antibody directed against PSA (b).  (c) The supernatants were preincubated in the absence or presence of Endo N before western blot analysis with PSA (lanes 1, 2) or NCAM (lanes 3, 4) antibodies. Lower panel in (a) shows intensities of the 110-kDa bands determined by densitometric quantification of three independent experiments (mean ± SD) and expressed as a percentage of the control (set to 100%).

Besides the major 110-kDa fragment, small amounts of a soluble 180-kDa form were observed in the supernatants in the absence of inhibitors (Fig. 1a, lane 1) and in the presence of leupeptin (Fig. 1a, lane 4), but not in the presence of EDTA, GM 6001 and 1,10-phenanthroline (Fig. 1a, lanes 2, 3 and 5). Western blot analysis using an antibody that recognizes α2,8-linked PSA showed that this 180-kDa NCAM form carried PSA and was most abundant in the supernatant of the control and in the presence of leupeptin (Fig. 1b, lanes 1 and 4). Only small amounts were found in the presence of GM 6001 and 1,10-phenanthroline (Fig. 1b, lanes 3 and 5), and it was not detectable in the presence of EDTA (Fig. 1b, lane 2). These data indicate that PSA-NCAM is also cleaved by a metalloprotease. After treatment of the supernatants with Endo N, an enzyme that specifically removes PSA from the protein backbone, the 180-kDa PSA-decorated form was no longer detectable (Fig. 1c, compare lane 2 with lane 1). The 110-kDa form was found in supernatants before or after treatment with Endo N (Fig. 1c, lanes 3 and 4), indicating that the transmembrane isoforms of PSA-NCAM are proteolytically cleaved, resulting in a 110-kDa NCAM form carrying PSA that migrates with an apparent molecular weight of 180 kDa on SDS–PAGE.

The soluble 110-kDa NCAM fragment is proteolytically released from cultured cells

To study proteolytic cleavage of NCAM expressed by live cells, mouse neuroblastoma Neuro 2a cells were grown in serum-free medium in the absence or presence of the protease inhibitors GM 6001 and leupeptin. These cells express the two transmembrane isoforms NCAM140 and NCAM180. Cell culture supernatants were subjected to western blot analysis after concentrating proteins by acetone precipitation. In the absence of protease inhibitors NCAM110 was found in the cell culture supernatant (Fig. 2a, lane 1). This form was also seen when cells were cultured in the presence of leupeptin (Fig. 2a, lane 3), but not in the presence of GM 6001 (Fig. 2a, lane 2), indicating that a metalloprotease, rather than a serine or cysteine protease, is involved in the generation of the soluble NCAM110 form.

Figure 2.

 Generation of soluble NCAM110 in neuroblastoma or transfected CHO cells. (a) Neuro 2a neuroblastoma cells were grown overnight in serum-free medium in the absence (lane 1) or presence of the protease inhibitors GM 6001 (lane 2) or leupeptin (lane 3). (b, c) CHO cells were transfected with a vector coding either for NCAM180 (lanes 1 and 2), NCAM140 (lanes 3 and 4) or NCAM120 (lanes 5 and 6) and maintained overnight in serum-free medium in the absence (lanes 1, 3 and 5) or presence (lanes 2, 4 and 6) of the protease inhibitor GM 6001. (a–c) Cells were pelleted and cell culture supernatants were cleared by centrifugation at 100 000 g. The cleared supernatants (a, c) and the cell pellets (b) were subjected to western blot analysis using a polyclonal NCAM-specific antibody. Lower panel in (a) shows intensities of the 110-kDa bands determined by densitometric quantification of three independent experiments and expressed as a percentage of the control (set to 100%).

Because neuroblastoma Neuro 2a cells express only the two transmembrane isoforms of NCAM, we investigated whether the GPI-anchored NCAM120 isoform is also released by proteolytic cleavage. NCAM-negative CHO cells were transfected with the NCAM180 (Fig. 2b, lanes 1 and 2), NCAM140 (Fig. 2b, lanes 3 and 4) and NCAM120 (Fig. 2b, lanes 5 and 6) isoforms. In culture supernatants of CHO cells transfected with NCAM180 and NCAM140, the 110-kDa fragment was seen in the absence of protease inhibitors (Fig. 2c, lanes 1 and 3), but was not detectable in the presence of GM 6001 (Fig. 2c, lanes 2 and 4). A 110-kDa fragment was not seen in the supernatant of cells transfected with NCAM120, but two soluble forms of NCAM with molecular weights of approximately 100 and 120 kDa were observed even in the presence of GM 6001 (Fig. 2c, lanes 5 and 6). These data indicate that the soluble 110-kDa form of NCAM is generated from the transmembrane isoforms of NCAM by proteolytic cleavage, whereas the soluble forms of NCAM derived from the GPI-anchored NCAM120 are generated by different processes.

ADAM17/TACE is involved in the generation of the soluble 110-kDa fragment

It has been shown in several studies that members of the ADAM metalloprotease family are involved in ectodomain shedding of cell recognition molecules and cell surface receptors. As ADAM17/TACE has a wide spectrum of different substrates, we analysed whether ADAM17/TACE mediates the cleavage of NCAM. ADAM17/TACE-deficient fibroblasts, which express NCAM140, were mock-transfected or transfected with a plasmid encoding ADAM17/TACE. Western blot analysis of the cell culture supernatants showed that NCAM110 was not detectable in culture supernatants of untransfected cells (Fig. 3a, lane 1), but was seen after transfection with ADAM17/TACE (Fig. 3a, lane 2), indicating that ADAM17/TACE cleaves the 140-kDa NCAM isoform.

Figure 3.

 Analysis of the involvement of ADAM17/TACE in the generation of soluble NCAM110. (a) ADAM17/TACE-deficient fibroblasts (lane 1) or ADAM17/TACE-deficient fibroblasts transfected with a vector coding for ADAM17/TACE (lane 2) were grown overnight in serum-free medium. (b, c) Neuro 2a neuroblastoma cells were mock-transfected (lane 1) or transfected either with a vector coding for functionally active ADAM17/TACE (lane 2) or a truncated inactive ADAM17/TACE (lane 3). The transfected cells were grown overnight in serum-free medium. After pelleting the cells and centrifugation at 100 000 g the cleared cell culture supernatants (a, c) were subjected to western blot analysis using a polyclonal NCAM-specific antibody and the cell pellets (b) were subjected to western blot analysis using a polyclonal TACE-specific antibody. (b) The individual TACE forms as previously described (Skovronsky et al. 2001) are shown. Lower panel in (c) shows intensities of the 110-kDa bands determined by densitometric quantification of three independent experiments and expressed as a percentage of the control (set to 100%).

Because it has been shown that transfection of a deletion mutant of human TACE lacking the pro-domain and catalytic domain results in cell surface expression of dTACE and inhibition of the proteolytic release of TNF-α and TNF receptor II (Solomon et al. 1999), we decided to use a similar enzymatically inactive, dominant-negative construct of mouse ADAM17/TACE (dTACE) to verify that ADAM17/TACE is involved in generation of the soluble 110-kDa fragment. Upon transfection of Neuro 2a cells with the empty vector (Fig. 3b, lane 1), NCAM110 was detectable in the cell culture supernatant (Fig. 3c, lane 1). Transfection with full-length ADAM17/TACE (Fig. 3b, lane 2) resulted in a small increase in the level of the 110-kDa fragment in the supernatant (Fig. 3c, lane 2). After transfection with a vector encoding the truncated, enzymatically inactive ADAM17/TACE (Fig. 3b, lane 3) only small amounts of this fragment were observed (Fig. 3c, lane 3). This result confirms that cleavage of transmembrane NCAM isoforms by ADAM17/TACE results in the generation of 110-kDa NCAM.

Involvement of calmodulin and actin cytoskeleton in release of the 110-kDa proteolytic fragment from cells in culture

Calmodulin regulates ectodomain shedding of cell surface proteins, such as TrkA, L-selectin, amyloid precursor protein and L1 (Kahn et al. 1998; Diaz-Rodriguez et al. 2000; Kalus et al. 2003). We therefore investigated whether generation of the 110-kDa fragment is also regulated by calmodulin. Neuro 2a cells were grown in the absence or presence of the calmodulin inhibitor CGS 9343B, and cell culture supernatants were then analysed to determine levels of the 110-kDa fragment in western blots. A pronounced increase in levels of the 110-kDa fragment was observed in the presence of the calmodulin inhibitor in comparison to the control without addition of the inhibitor (Fig. 4a, compare lanes 1 and 2). A similar result was obtained when ADAM17/TACE-deficient fibroblasts transfected with ADAM17/TACE were maintained in the presence of the calmodulin inhibitor (Fig. 4b, compare lanes 1 and 2).

Figure 4.

 Role of calmodulin and the actin cytoskeleton in the generation of soluble NCAM110. Neuro 2a neuroblastoma cells (a), ADAM17/TACE-deficient cells stably transfected with ADAM17/TACE (b) or NCAM-negative CHO cells transfected with a vector coding for NCAM140 (c) or NCAM120 (d) were grown in serum-free medium overnight in the absence (lane 1) or presence of the calmodulin inhibitor CGS 9343B (lane 2). (e) Neuro 2a cells were grown in serum-free medium overnight in the absence (lane 1) or presence of colchicine (lane 2), cytochalasin D (lane 3), latrunculin B (lane 4) or paclitaxel (lane 5). (a–e) The cleared cell-free culture supernatants were subjected to western blot analysis using a polyclonal NCAM-specific antibody. Intensities of the 110-kDa bands determined by densitometric quantification of three independent experiments and expressed as a percentage of the control (set to 100%).

This increase in levels of the 110-kDa fragment was also observed in cell culture supernatants of CHO cells transfected with NCAM140 maintained in the presence of the calmodulin inhibitor (Fig. 4c, compare lanes 1 and 2). These results indicate that proteolytic processing of the transmembrane NCAM isoforms by ADAM17/TACE is reduced by calmodulin. A major 100-kDa form and a minor 120-kDa form were released from CHO cells transfected with NCAM120, and the amount of these forms was not increased in the presence of the calmodulin inhibitor (Fig. 4d, lanes 1 and 2), indicating that generation of soluble forms derived from the GPI-anchored NCAM120 is not regulated by calmodulin.

The role of the cytoskeleton in regulation of proteolytic processing was investigated by maintaining Neuro 2a cells in the absence or presence of agents that stabilize or destabilize the actin or tubulin cytoskeleton. In comparison to the levels of the 110-kDa fragment in cell culture supernatant of the control without addition of such agents (Fig. 4e, lane 1), levels of the 110-kDa fragment were increased in the presence of the actin-destabilizing agents cytochalasin D and latrunculin B (Fig. 4e, lanes 3 and 4). In contrast, in the presence of the microtubule-destabilizing agent colchicine or microtubule-stabilizing agent paclitaxel, levels of the 110-kDa fragment were similar to control levels (Fig. 4e, lanes 2 and 5). These results indicate that proteolytic processing of the transmembrane NCAM isoforms is enhanced when the dynamics of the actin cytoskeleton, but not of the tubulin cytoskeleton, are disturbed, suggesting that proteolytic processing of the transmembrane NCAM isoforms is regulated by interaction with the actin cytoskeleton.

NCAM-induced neurite outgrowth depends on proteolytic cleavage

Cerebellar neurones from NCAM-deficient mice do not exhibit promotion of neurite outgrowth on substrate-coated NCAM in contrast to those that express NCAM (Niethammer et al. 2002 and references therein). This indicates that NCAM-induced neurite outgrowth is mainly mediated by homophilic interaction and signalling via cell surface-expressed NCAM. We investigated the effect of the metalloprotease inhibitor GM 6001 on NCAM-induced neurite outgrowth to determine whether processing of NCAM by a metalloprotease is required for NCAM-induced neurite outgrowth. Cerebellar explants were maintained in the presence or absence of GM 6001 and leupeptin either on PLL as a control substrate or on substrate-coated NCAM immunopurified from mouse brain. Explants maintained on PLL showed no significant difference in neurite length in the presence or absence of leupeptin or GM 6001 (Figs 5a, c and d). Explants maintained on substrate-coated NCAM showed a decrease in neurite length of approximately 50% in the presence of GM 6001 compared with explants maintained in the absence of this inhibitor (Figs 5b, e and f). Neurite outgrowth was not affected in the presence of leupeptin (Fig. 5b). Thus NCAM-mediated neurite outgrowth depends on the action of a metalloprotease but not on the action of a serine or cysteine protease.

Figure 5.

 Dependence of NCAM-induced neurite outgrowth on proteolytic processing. Cerebellar microexplant cultures were plated onto glass coverslips coated with PLL (a) or a combination of PLL and NCAM (b), and maintained in the absence or presence of GM 6001 or leupeptin. After incubation for 24 h at 37°C, the explants were fixed and stained. (a, b) Neurite outgrowth from the explants was quantified by measuring the 10 longest neurites of 15 aggregates in three independent experiments. Neurite lengths in the presence of GM 6001 (black bars) or leupeptin (white bars) are shown relative to lengths in the absence of inhibitors. *p < 0.005 by t-test. (c–f) Microexplants maintained on coverslips coated with PLL (c, d) or PLL and NCAM (e, f) for 24 h at 37°C in the absence (c, e) or presence of GM 6001 (d, f) are shown after fixation and staining. Scale bars 50 µm.

To further investigate whether TACE is involved in regulating neurite outgrowth, we transfected hippocampal neurones with an enzymatically inactive mutant of TACE. We chose hippocampal neurones because, in contrast to cerebellar neurones, these neurones could be easily and efficiently transfected by electroporation. Furthermore, it has been shown recently that NCAM-dependent neurite outgrowth of hippocampal neurones is regulated by a metalloprotease as indicated by reduction of neurite outgrowth by GM 6001 (Hubschmann et al. 2005). Hippocampal neurones transfected with inactive TACE showed a reduction in neurite outgrowth on substrate-coated NCAM in comparison to mock-transfected neurones, whereas transfected neurones grown on the control substrate PLL did not (Fig. 6).

Figure 6.

 NCAM-induced neurite outgrowth from hippocampal neurones depends on TACE. Hippocampal neurones were transfected with green fluorescent protein (GFP) alone (white bars) or co-transfected with an enzymatically inactive mutant of TACE (black bars). Transfected neurones were plated on to glass coverslips coated with PLL or a combination of PLL and NCAM (NCAM), and maintained at 37°C for 24 h. Neurite outgrowth in three independent experiments was quantified by measuring the total length of neurites in GFP-positive cells. *p < 0.005 by t-test.

Discussion

ADAM17/TACE specifically cleaves the transmembrane isoforms of NCAM

Proteolytic processing of cell surface proteins is of prime importance for regulating the functional properties of adhesion molecules (for reviews see Hooper et al. 1997; Schlondorff and Blobel 1999; Turner and Hooper 1999; Yoshida and Shiosaka 1999). Cleavage of recognition molecules at the cell surface has been implicated in neuronal migration, neurite outgrowth and synaptic plasticity (for reviews see Buxbaum et al. 1998; Madani et al. 1999; Galko and Tessier-Lavigne 2000a; Asahina et al. 2001; Spira et al. 2001). Extracellular metalloproteases are involved in remodelling of the extracellular matrix, are crucial regulators of cell function and have been implicated in diseases of the nervous system, such as multiple sclerosis and Alzheimer's disease. Two large families of zinc-dependent extracellular metalloproteases have been described: the matrix metalloproteases and the ADAMs. ADAM17/TACE cleaves a number of different cell surface proteins or receptors involved in a variety of biological processes. Among the numerous substrates of ADAM17/TACE are amyloid precursor protein (Buxbaum et al. 1998), L-selectin (Goddard et al. 2001), vascular cell adhesion molecule 1 (Garton et al. 2003), growth hormone receptor (Zhang et al. 2000), neurotrophin receptor p75 (Weskamp et al. 2004), receptor tyrosine kinase c-kit (Cruz et al. 2004), macrophage colony-stimulating factor receptor (Rovida et al. 2001), Notch (Brou et al. 2000), TNF receptors (Solomon et al. 1999), CD44 (Nagano et al. 2004) and prion protein (Vincent et al. 2001). Like ADAM10 that also cleaves, for instance, amyloid precursor (Lammich et al. 1999) and Notch (Parks et al. 2000), ADAM17/TACE is likely to function as an α-secretase and mutations in either protease cause severe defects in development and adulthood (Peschon et al. 1998; Killar et al. 1999). We provide several lines of evidence that the two transmembrane NCAM isoforms are cleaved by at least one metalloprotease, namely ADAM17/TACE. Cleavage by this metalloprotease results in the generation of a major 110-kDa fragment that is soluble, and found in cell culture supernatants and in the soluble fraction of brain homogenates.

The GPI-anchored NCAM120 isoform is not cleaved by metalloproteases

In contrast to the transmembrane NCAM isoforms, the GPI-anchored NCAM120 isoform is not cleaved of metalloproteases. This result is in contrast to recent findings by reported by Hubschmann et al. (2005), who showed that all isoforms of NCAM, including NCAM120, are cleaved by a metalloprotease. The reason for this discrepancy is unknown. Nevertheless, in the present study we obtained evidence that generation of soluble fragments derived from this NCAM isoform must involve other mechanisms, such as release from the cell surface by the phosphatidylinositol-specific phospholipase C (He et al. 1986; Sadoul et al. 1986) or by proteases other than metalloproteases. In vitro studies using the serine protease plasmin, the end product of the plasminogen–plasminogen activator cascade, indicate that NCAM may also be cleaved by this protease (Endo et al. 1998, 1999; Hoffman et al. 1998). However, the proteolytic products of plasmin activity had molecular weights between 60 and 90 kDa, possibly resulting from several cleavage sites within the extracellular domain of NCAM. It is noteworthy in this context that we did not obtain any evidence for the cleavage of NCAM by a serine protease. Nevertheless, it is possible that cleavage by plasmin takes place under certain physiological conditions. It is interesting in this respect that plasmin has also been implicated in cleavage of L1, leading to the generation and release of a defined 140-kDa fragment (Silletti et al. 2000). Two other proteases have been implicated in extracellular cleavage, namely the pro-protein convertase PC5a and ADAM10 of L1 (Kalus et al. 2003).

PSA-NCAM is also cleaved by ADAM17/TACE

The α2,8 PSA glycan is carried by the three major isoforms of NCAM, NCAM120, NCAM140 and NCAM180 (for review see Durbec and Cremer 2001). This glycan confers special features to the NCAM protein backbone in that it reduces NCAM homophilic binding by steric interference with NCAM protein–protein interactions and by its prominent hydration volume surrounding the highly negatively charged sialic acid residues. The detection of soluble PSA-carrying NCAM fragments with apparent molecular weights ranging from 120 to 180 kDa is remarkable because these fragments, in particular the 180-kDa soluble NCAM fragment, have been thought to derive from NCAM glycoprotein backbone without PSA (Takamatsu et al. 1994). The present study makes it likely that the soluble fragments with molecular weights ranging from 120 to 180 kDa represent, at least in part, the PSA-carrying soluble extracellular domain of NCAM.

Regulation of proteolysis of NCAM by calmodulin

Like other cell surface receptors, in particular adhesion molecules such as TrkA (Diaz-Rodriguez et al. 2000; Llovera et al. 2004), L-selectin (Kahn et al. 1998; Diaz-Rodriguez et al. 2000), amyloid precursor protein (Diaz-Rodriguez et al. 2000) and L1 (Kalus et al. 2003), the proteolytic processing of NCAM is regulated by the calcium-binding protein calmodulin. As for L1, a specific calmodulin inhibitor increases processing of the transmembrane isoforms of NCAM. This regulation of proteolytic processing is involved in NCAM-mediated homophilic neurite outgrowth, because inhibition of such processing by a metalloprotease inhibitor leads to a reduction in neurite outgrowth, as we have shown in the present study. The dependence of neurite outgrowth on ectodomain shedding has been observed for L1-induced neurite outgrowth (Kalus et al. 2003) and netrin-induced deleted colon cancer-mediated neurite outgrowth (Galko and Tessier-Lavigne 2000b). The potential involvement of calmodulin, a calcium-binding protein and modulator of calcium-dependent mechanisms, is interesting taken that calcium is an eminent regulator of activity-triggered events, such as neurite outgrowth, synaptogenesis and synaptic plasticity. This is also interesting in view of the observation that NCAM activation by homophilic binding mechanisms modulates calcium influx into neuronal cells (Schuch et al. 1989). Furthermore, release of a soluble form of NCAM in microdialysates of long-term potentiated hippocampus underscores the importance of activity on the release of proteolytic NCAM fragments (Fazeli et al. 1994; Luthi et al. 1994). The involvement of calmodulin as a modulator of calcium-dependent activities provides a biologically meaningful link between generation of a diffuseable fragment of the extracellular domain of NCAM and neuronal activity.

Involvement of the actin cytoskeleton in release of the 110-kDa proteolytic NCAM fragment

The involvement of the cytoskeleton in regulation of proteolytic processing and, thus, neurite outgrowth, was investigated with agents that stabilize or destabilize the actin or tubulin cytoskeleton. Increased levels of the soluble 110-kDa fragment were seen in the presence of the actin-destabilizing agents cytochalasin D and latrunculin B. The microtubule-destabilizing agent colchicine and the microtubule-stabilizing agent paclitaxel, however, did not influence the levels of the released 110-kDa fragment. Because the actin cytoskeleton is predominantly present in growth cones, whereas microtubules are mainly present in neurites and not in growth cones, these results point to a specific role of NCAM in regulating growth cone movement. The fact that actin-destabilizing agents increase the release of the 110-kDa NCAM fragment into the culture supernatant is noteworthy, although the underlying mechanism has yet to be elucidated. However, the present findings indicate that the dynamics of the actin cytoskeleton are important for proteolytic processing and ultimately for neurite outgrowth. The fact that both calcium-dependent mechanisms, mediated by calmodulin and the actin cytoskeleton of the growth cone, influence the generation of a soluble NCAM fragment suggests that the release of such a fragment has particular biological significance.

Biological significance of the soluble extracellular domain of NCAM isoforms

Our observation that the release of the extracellular domain fragment of NCAM from the cell surface appears to be linked to calcium-mediated activity and that the actin cytoskeleton is also implicated points to a significant function of the soluble extracellular domain of NCAM in important morphogenetic events, such as neurite outgrowth and activity-regulated effects on synaptic activity and plasticity. The fact that the NCAM fragment may diffuse into the extracellular space and interact with extracellular matrix components (Probstmeier et al. 1989) suggests that the soluble NCAM fragment modulates the cellular environment to yield feedback mechanisms to NCAM at the cell surface in homophilic signal transduction pathways (Schuch et al. 1989). The soluble NCAM fragment may also act heterophilically with receptors at the cell surface (Kadmon et al. 1990; Williams et al. 1994; Saffell et al. 1997). It is thus conceivable that NCAM-synthesizing cells build concentration gradients of either soluble or matrix-embedded NCAM that modulate growth cone guidance and cell migration by conditioning the cellular environment. It is also possible that proteolytic processing of NCAM uncovers binding sites of the residual transmembrane stump that serve as newly uncovered receptors for ligands that would not have interacted with this part of the extracellular domain of NCAM if it had not been cleaved. It is therefore likely that proteolytically processed NCAM may subserve at least two functions: modification of the extracellular milieu and transmembrane signalling via the residual transmembrane NCAM receptor stumps.

However, it is also possible that NCAM shed from the cell surface may perturb homophilic and heterophilic interactions between neural cells in a dominant-negative fashion, leading to altered morphogenesis and synaptic function. It is interesting in this context that increased levels of extracellular domains of adhesion molecules have been found in the CSF of schizophrenic patients and in other psychoses (Ignatova et al. 2004; Poltorak et al. 1995; Vawter et al. 2001). NCAM dysregulation has been linked to other diseases, including bipolar disorders (Arai et al. 2004), Alzheimer's disease (Todaro et al. 2004) and other dementias (Strekalova et al. 2005). A recent study has shown that expression of the soluble extracellular domain of NCAM in transgenic mice leads to abnormalities in GABAergic interneurones and alteration in behaviour reminiscent of many features characteristic of some traits in schizophrenia or other neuropsychiatric disorders (Pillai-Nair et al. 2005). It remains to be seen whether the actions of the soluble form of the extracellular domain of NCAM evoke these features by cell surface-mediated homophilic interactions or by triggering NCAM-mediated signal transduction pathways as a trans-interacting ligand for homophilic or heterophilic interaction partners at the cell surface.

Acknowledgements

We are grateful to Patricia Maness and Elisabeth Bock for NCAM cDNA clones, Atanasio Pandiella for the vector encoding ADAM17/TACE, Rita Gerardy-Schahn for the antibody against PSA and for Endo N, Stuart J. Frank for the TACE antibody, and Amgen for providing ADAM17/TACE-deficient cells.

Ancillary