Ectodomain Shedding of the Neural Recognition Molecule CHL1 by the Metalloprotease-disintegrin ADAM8 Promotes Neurite Outgrowth and Suppresses Neuronal Cell Death*

The neural cell adhesion molecule “close homologue of L1,” termed CHL1, has functional importance in the nervous system. CHL1 is expressed as a transmembrane protein of 185 kDa, and ectodomain shedding releases soluble fragments relevant for its physiological function. Here we describe that ADAM8, a member of the family of metalloprotease disintegrins cleaves a CHL1-Fc fusion protein in vitro at two sites corresponding to release of the extracellular domain of CHL1 in fibronectin (FN) domains II (125 kDa) and V (165 kDa), inhibited by batimastat (BB-94). Cleavage of CHL1-Fc in the 125-kDa fragment was not detectable under non-reducing conditions arguing that cleavage resulting in the 165-kDa fragment is more relevant in releasing soluble CHL1 in vivo . In cells transfected with full-length ADAM8, membrane proximal cleavage of CHL1 was similar and not stimulated by phorbol ester 12- O -tetradec-anoylphorbol-13-acetate and pervanadate. No cleavage of CHL1

causes qualitative and irreversible changes in the function of these molecules. As demonstrated with genetic and biochemical means, enzymes capable of these functions are ADAMs 1 (for "a disintegrin and metalloprotease domain"), proteins that constitute a family of transmembrane glycoproteins with essential physiological roles in fertilization, myogenesis, and neurogenesis (2). These functions are due to distinct protein domains involved in either cell-cell fusion or cell-cell interaction, or to their zinc-coordinating metalloprotease domain. To date, the family of ADAM proteinases comprises more than 30 members in different species (3,4). Fourteen of the murine ADAMs contain the catalytic consensus sequence HEXXHXXGXXHD in their metalloprotease domains and are therefore predictably proteolytically active (3,4). The cleavage of myelin basic protein (MBP) by ADAM10/MADM (mammalian disintegrin-metalloprotease) was the first demonstration of proteolysis by AD-AMs (5). The TNF-␣ convertase (TACE/ADAM17) was purified on the basis of its ability to cleave TNF-␣ (6, 7) and a number of other peptide and protein substrates in vitro (1,8). TACE is also implicated in the shedding of factors in the EGF ligand family, such as heparin-binding epidermal growth factor (HB-EGF) and amphiregulin (reviewed in Ref. 2), and the physiological importance of this cleavage was demonstrated by similar defects in TACE and HB-EGF-deficient mice (9,10).
In the CNS, proteolytic processing of the amyloid precursor protein (APP (11)) is an essential event in the pathology of Alzheimer's disease. As ␣-secretases, a number of ADAMs have been described to cleave APP at the respective ␣-cleavage site, ADAM9, ADAM10, and ADAM17, suggesting that ADAM family members serve essential roles in the CNS.
Recently, ADAM8, an ADAM originally cloned from monocytic cells (12), has been shown to be expressed in the CNS in neurons, reactive glia cells (astrocytes and microglia), and oligodendrocytes (13) and was described as a sheddase of the low affinity IgE receptor CD23 (14). In addition to substrates of ADAM8 in the immune response, a specific substrate of ADAM8 in the CNS has not yet been described. From these examples it is clear, that for each substrate and tissue, the physiological relevance of processing has to be proven by a * Work was supported in part by the Deutsche Forschungsgemeinschaft, SFB 549 (A4) (to J. W. B. and H. Jockusch), SCHA 185/38 -1.2 (to M. S.), and by the Ministry of Education and Science Northrhine-Westphalia, Graduate School for Bioinformatics and Genome Research (grant to S. N.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Both authors contributed equally to this work. ʈ Present address: Pasquale Lab, The Burnham Institute, 10901 North Torrey Pines Rd., La Jolla, CA 92037.
In the CNS, neural recognition molecules play important roles in specifying cell interactions during development, regeneration, and modification of synaptic activity. One subfamily of recognition molecules is the immunoglobulin (Ig) superfamily. In their extracellular parts these molecules contain IgG-like domains with essential cysteine residues and four to five fibronectin repeats (FN). Within these FN domains, cell adhesion molecules can be proteolytically processed and released from the cell membrane, thereby exerting important biological functions. The first identified member of the Ig family was L1 (15). L1 is a 200-to 220-kDa type I transmembrane protein. In the mouse brain, L1 protein is detectable in four distinct molecular forms with sizes of 200, 180, 140, and 80 kDa (16). It was recently shown that L1 is processed by ADAM10 into 85and 32-kDa fragments (17), but also the proprotein convertase PC5A plays an essential role in L1 processing leading to processed fragments of 180-and 140-kDa size (18).
Based on cross-reactivity of L1 antibodies, the close homologue of L1, termed CHL1, was cloned (19). In brain extracts, CHL1 is detectable in three distinct fragments of 185, 165, and 125 kDa. The 185-kDa fragment was only weakly detectable in non-detergent soluble brain extracts indicating that this might be the membrane-spanning form of CHL1, whereas the 165and 125-kDa fragments are considered as proteolytically released fragments (20). The soluble forms of L1 and CHL1, in which the extracellular portion was fused to an Fc tail, promote neuronal survival and control neurite outgrowth (21). As demonstrated for processed L1, soluble forms can interact effectively with the extracellular matrix to regulate migration by interactions with the integrin ␣ v ␤ 5 (17,22) or by unknown ligands. Similar observations have been made for CHL1 (23). Expression of CHL1 in injured and regenerating central and peripheral neurons is strongly up-regulated (24,25), but its functional role in regeneration remains to be defined. Because specific proteolytic processing of CHL1 may be of functional importance for neuronal survival, neurite outgrowth, and cell migration, we screened for the ability of several members of the family of ADAM metalloproteases to process CHL1 into its naturally occurring fragments.

MATERIALS AND METHODS
Animals-Breeding and experimental use of mice were performed in agreement with the German law on the protection of animals, with a permit by the local authorities. The wr mutation was maintained on a C57BL/6 background. 30-day-old mutant mice with a manifest WR phenotype were used for biochemical analysis. Mice deficient in ADAM8 were received from Dr. A. J. P. Docherty (Celltech, Slough, UK) and were maintained on a C57BL/6 background. Genotype analysis was done by PCR diagnosis on tail tip DNA. A more detailed description of these mice will be given elsewhere.
Cloning of Expression Constructs-The prodomain and the catalytic domain of mouse ADAM8 were amplified by PCR with the primers mXhoFLA8MET (5Ј-CTC GAG ATC AGA TCT CCT GAG GCT TAA ACT GAG GGA AGG ACA CGA ACC GGT TGA CAT-3Ј) and mEcoREnA8Sta (5Ј-GAA TTC GAC GAC GAC GAC AAG ATG CTT GGC CTC TGG CTG CTC-3Ј). The resulting PCR product was ligated into the respective cloning sites (XhoI at the 5Ј end; EcoRI at the 3Ј end) of pSecTag2B vector (Invitrogen) to generate pA8MPSecTag. Fulllength mouse ADAMs10 and 17 cDNAs were generated by reverse transcription-PCR with mRNA from mouse brain and spleen, respectively. Amplification of cDNAs by PCR was performed with Pfu polymerase with proofreading activity and the following primers: ADAM10s, 5Ј-ATG GTG TTG CCG ACA GTG TTA ATT CT-3Ј; ADAM10as, 5Ј-TTA GCG TCG CAT GTG TCC CAT TTG A-3Ј. For ADAM17 containing a Bi-Pro tag (26), we used the following primers: ADAM17s, 5Ј-ATG AGG CAG TCT CTC CTA TTC CTG-3Ј; ADAM17as (with Bi-Pro tag sequence), 5Ј-TCA GAT CTC CTG AGG CTT AAA CTG AGG GAA GGA GCA CTC TGT TTC TTT GCT GTC A-3Ј. The resulting PCR products were cloned in the mammalian expression vector pTARGET (Promega, Heidelberg) containing a cytomegalovirus promoter. Insertion of correct sequences was verified by sequencing.
Cell Culture-COS-7 cells were grown in Dulbecco's modified Eagle's medium in the presence of 10% fetal calf serum and 1% glutamine. Transient transfections were performed with LipofectAMINE (Invitrogen, Groningen, Netherlands). Stable transfected CHOK1 cells for the production of Fc fusion proteins were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplied with 5% ultra low IgG fetal calf serum, 1% glutamine, 1% non-essential amino acids, and HT-Supplement. From these cells, Fc fusion proteins were prepared as described previously (21).
Purification of Soluble ADAM8 Catalytic Domain-COS-7 cells were transiently transfected with the construct pA8MPSecTag (Invitrogen, Heidelberg, Germany) containing the pro-and the catalytic domain with a C-terminal His tag and an N-terminal secretion signal (Sec tag). For purification of ADAM8 protease, COS-7 supernatants were collected 36 -48 h after addition of serum-free Dulbecco's modified Eagle's medium. Samples were concentrated with Amicon Ultra Centrifugal Filter Devices 10K (Millipore) and further purified by affinity purification with nickel-nitrilotriacetic acid chromatography (ProBond TM Purification System, Invitrogen). After elution, the supernatants were concentrated a second time by ultrafiltration ϳ100-fold compared with harvested supernatants. The purity of soluble protease was confirmed by SDS-PAGE, concentrations were determined using the BCA reagent (Perbio Science, Bonn, Germany), and proteolytic activity was monitored by a fluorescent cleavage assay in which ADAM8 cleaves the quenched fluorescent peptide Suc-H-(Mcp)-GSLPQKSH-K(Dpa)-R-amide (27).
Analysis of Protease Activities in Cell Lysates-24 h after transfection of Chinese hamster ovary or COS-7 cells with the constructs encoding full-length ADAM proteases, the cells were washed 1ϫ with phosphatebuffered saline and lysed with a buffer containing 20 mM HEPES, pH 7.4, 1% Nonidet P-40, and 0.5% deoxycholate. After centrifugation at 15,000 ϫ g for 2 min, the resulting cell pellet containing cell-bound proteases was directly assayed for protease activity by adding 150 l of mixture consisting of 50 M TNF-␣ substrate (Dnp-SPLAQAVRSSSR-  (26)) and 50 l of assay buffer (20 mM HEPES, pH 7.4, and 0.0015% Brij 35). Activity was measured at 340-nm excitation and 405-nm emission.
N-terminal Sequence Analysis-For preparation of samples for sequencing, ϳ10 g of CHL1-Fc protein was incubated overnight with 10 g of soluble, recombinant ADAM8. The samples were run on an SDS-PAGE gel and washed in borate buffer (50 mM H 3 BO 3 , pH 9.0, 20% ethanol, 1 mM dithiothreitol) after electrophoresis. Samples were electroblotted onto polyvinylidene difluoride membranes (Porablot, Macherey and Nagel, Dü ren, Germany) and stained with 0.2% Coomassie Brilliant Blue in 45.5% EtOH, 9% acetic acid, and 1 mM dithiothreitol. For N-terminal sequencing, the respective bands were excised and applied to a protein sequencer (Knauer, Germany). Five cycles of automated Edman degradation were run to obtain sufficient sequence information.
Neurite Outgrowth Assays-Cerebellar granule neurons were prepared from 6-to 8-day-old CD1 mice as described previously (29). Cells were seeded onto poly-L-lysine (PLL)-coated coverslips and were maintained in serum-free medium (30). 1-2 h after seeding the neurons, CHL1-Fc protein was added to the medium as a positive control. To determine neurite outgrowth in the presence of supernatants from transfected COS-7 cells, serum-free medium was applied 12 h after co-transfection of COS-7 cells with CHL1/ADAM constructs and was left on the cells for 24 h. The COS-7 supernatants were applied to cerebellar neurons, and 18 h after addition of either CHL1-Fc protein or supernatants of transfected COS-7 cells, neurite outgrowth was determined. To visualize neurite outgrowth, cells were fixed with glutaraldehyde (4%) for 5 min and stained with either Richardson's Blue or with a polyclonal anti-NF200 antibody (Sigma, Germany). For each experimental setup, neurites of at least 50 cells with processes longer than the cell body diameter were measured using a Zeiss Axiophot microscope with digital equipment.
Neurite outgrowth assays in co-cultures of cerebellar neurons with transfected COS-7 cells were performed by plating 5 ϫ 10 4 COS-7 cells in 30-mm plates on PLL-coated coverslips. One day after transfection of COS-7 cells, freshly prepared cerebellar neurons were added at densities of 10 5 cells/plate, and the co-culture was maintained in serum-free medium. After 24 h, the number of attached neurons per visual field and neurite lengths were determined as described above.
Each neurite outgrowth experiment was performed with cerebella from 3 mice, and each condition was determined in triplicates. The experiments were repeated at least 3 (n ϭ 9) or 4 (n ϭ 12) times as indicated. Data are given as means Ϯ S.E.
Cell Survival Assay-Cerebellar granule neurons were seeded in 96-well plates coated with PLL at densities of 10 5 /well. After cell attachment, either serum-free medium or supernatants from transfected COS-7 cells were applied. After days 1, 3, and 5, the numbers of surviving cells were counted by Trypan blue exclusion staining. Data are from three independent experiments in triplicates and as means Ϯ S.E. A Student t test was used to determine the significance, and a p value of Ͻ0.01 was considered as highly significant.

Cleavage of Recombinant CHL1 by ADAM8 -Recombinant
ADAM8 was expressed as a secreted soluble form from COS-7 cells containing the pro-and metalloprotease domain (A8-MP). To monitor for catalytic activity, a parallel assay with a fluorogenic substrate was performed (27). Extracellular domains of NCAM, L1, and CHL1 (Fig. 1) fused to an Fc tag (21) were incubated with purified ADAM8 (Fig. 2A). Whereas NCAM and L1 were not cleaved by ADAM8, CHL1-Fc was cleaved into two products: a prominent 165-kDa fragment corresponding to the entire extracellular portion of CHL1 and a less prominent 125-kDa fragment. Whereas the 165-kDa fragment was detectable under reducing and non-reducing conditions, the 125-kDa fragment disappeared under non-reducing conditions ( Fig. 2A,  right panel). Specificity of the cleavage was confirmed by immunostaining using an antibody against the extracellular portion of CHL1 (Fig. 2B), and equal loading as well as processing of pro-ADAM8 into active protease by autocatalytic prodomain removal was confirmed by Western blot analysis (Fig. 2C). Usually, about 50% of he ADAM8 protein is processed into the active form. In Western blot experiments with anti-CHL1, the specific cleavage of CHL1 was confirmed by detecting the noncleaved CHL1-Fc protein of about 200-kDa size and besides the cleaved CHL1 fragments, bands of ϳ80 kDa and ϳ40 kDa, corresponding to the fragments of CHL1-Fc containing the residual Fc portion (Fig. 2B), which were subjected to N-termi-nal sequencing. For the 165-kDa fragment, the cleavage site was determined at LVP2GAEHIV located within the fifth partial FN III-like domain at amino acid position 1040, for the 125-kDa fragment, the cleavage site within the second FN III-type domain was determined at amino acid position 753 in the sequence WKP2QGAPE (Fig. 1). Extensive cleavage of CHL1 by soluble ADAM8 was inhibited by broad-range inhibitors of metalloproteases, BB-94 and OPT, but not by 500 nM TIMPs 1-4, as demonstrated by a silver-stained gel and by quantitative analysis (Fig. 2D). The observed inhibition profile is in accordance with previous cleavage studies on ADAM8 (27,31).
Determination of Protease Activities of ADAM8, ADAM10, and ADAM17-As a prerequisite for the functional analyses performed with ADAM proteins in co-transfection experiments, we performed fluorescence activity assays with cells transiently transfected with full-length cDNAs for ADAM8, EQ-ADAM8, ADAM10, and ADAM17 (Fig. 3A). In cell lysates from these cells, we determined protease activities for the MBP peptide substrate with ADAM8 and ADAM10, but not with lysates from mock transfected cells, EQ-ADAM8, and ADAM17. To monitor ADAM17 activity, we used the TNF-␣ substrate, which is significantly cleaved only by ADAM17. These data indicate that the mouse cDNAs used to express ADAM8, ADAM10, and ADAM17 encode functional proteases in COS-7 cells.
Cleavage of CHL1 by ADAM8 in Cell Culture-To detect ectodomain shedding of CHL1 by ADAM8 in cell culture, fulllength CHL1 and ADAM8 were co-transfected into COS-7 cells. Cell lysates and supernatants were analyzed. For comparison, wild type ADAM8 and the catalytically inactive ADAM8, in which Glu 330 was exchanged by Gln (EQ-ADAM8 (27)) were also co-transfected. The full-length 185-kDa CHL1 transmembrane protein was detected, slightly more in cells transfected with EQ-ADAM8 than in cells transfected with ADAM8. Cotransfection of ADAM8 significantly increased the amount of soluble 165-and 125-kDa fragments of CHL1 in supernatants of cells transfected with ADAM8, but not with EQ-ADAM8 (Fig. 3B).
We conclude from this experiment that ADAM8 expression in CHL1-transfected COS-7 cells results in significantly higher amounts of soluble CHL1 compared to the basal CHL1 shedding. The specificity of the cleaving protease is demonstrated by co-transfection experiments with cDNAs for CHL1 and ADAM10 or CHL1 and ADAM17, in which no significant cleavage of CHL1 was observed. Using high speed centrifugation of the supernatant, the ϳ60-kDa band representing the membrane stub and the corresponding 125-kDa band were removed; also under non-reducing conditions, the 125-kDa band was not detectable (Fig. 3B). To analyze the dependence of ADAM8mediated shedding on activators of protein kinase C, well described for ADAM17/TACE, we performed co-transfection experiments in the presence of phorbol ester TPA and pervanadate, a phosphatase inhibitor. No significant effect on CHL1 shedding was observed, even when TPA and pervanadate were applied in high concentrations (Fig. 3C).
CHL1 Shedding in ADAM8-deficient Mice-The correlation between ADAM8 expression and CHL1 shedding was further analyzed in situ. Proteolytic processing of CHL1 in brain extracts from 30-day Wobbler (WR) mutant mice and from mice deficient in ADAM8 was compared with wild type mice (Fig. 4). In cerebellum/brain stem from wild type mice, ADAM8 is present in its catalytically active form, and CHL1 processing into the 125-and 165-kDa fragments was observed. In cerebellum/ brain stem from Wobbler mice, an up-regulation of ADAM8 has been demonstrated as a consequence of neurodegeneration and astrogliosis (13) (Fig. 4A). Consistently, an increased proteolytic processing of 185-kDa CHL1 into both soluble fragments was detected (Fig. 4A). In homozygous ADAM8-deficient mice, the ADAM8 protein was not detectable. Processing of CHL1 in brain extracts of these mice was dramatically reduced emphasizing a significant contribution of ADAM8 to the proteolytic processing of CHL1 in this particular brain region in situ.
Effect of CHL1 Shedding on Neurite Outgrowth in Neuronal Cell Cultures-To analyze the physiological function of soluble CHL1 generated by ADAM8-dependent ectodomain shedding, we performed neurite outgrowth and cell survival assays with mouse cerebellar neurons in the presence of conditioned media from CHL1 and ADAM8 co-transfected COS-7 cells (Fig. 5). As described previously, recombinant L1-Fc or CHL1-Fc proteins are able to stimulate neurite outgrowth from cerebellar or hippocampal neurons (21). Supernatants from COS-7 cells transfected with ADAM8, CHL1, or both were derived by 24-h maintenance of cells in serum-free medium. Supernatants from ADAM8-or CHL1-transfected cells showed a basal level of neurite outgrowth (20 Ϯ 1.6 m for ADAM8, 22 Ϯ 1.9 m for CHL1). In contrast, supernatants from cells co-transfected with ADAM8 and CHL1 caused significantly increased neurite outgrowth up to 43 Ϯ 4 m compared with only CHL1 transfected cells (22 Ϯ 2 m). A comparable rate to the basal level of neurite outgrowth was achieved with EQ-ADAM8/CHL1 (21 Ϯ 2 m) or ADAM10/CHL1 (19.5 Ϯ 2.5 m) or ADAM17/CHL1 (21 Ϯ 3 m). In the same experiments, recombinant CHL1-Fc protein was used as a standard. In a dose-response curve, 1 g/ml CHL-Fc protein resulted in a maximal neurite outgrowth (45 Ϯ 5 m). By a dot blot assay, we determined the concentration of the soluble CHL1 165 fragment by comparison with known CHL1-Fc concentrations and determined a concentration of ϳ100 ng/ml soluble CHL1 in the supernatant. Based on this calculation we assume that the concentration of soluble CHL1 required for the maximal neurite outgrowth is one order of magnitude lower than the concentration of CHL1-Fc fusion protein. We can not exclude that soluble CHL1 may attach to the substrate and exerts its effect not only as a soluble molecule, but this also holds for CHL1-Fc, so the effects may be comparable.
Enhancement of Neurite Outgrowth by Substrate-coated Soluble CHL1-We used a co-culture of cerebellar granule neurons with transfected COS-7 cells to determine if CHL1 proteolysis enhances neurite outgrowth when offered as a substrate (Fig. 6). The numbers of substrate-attached cells as well as the neurite lengths were determined. In accordance with the previous experiment, the number of substrate-attached cells was significantly larger when granule neurons were plated together with COS-7 cells transfected with ADAM8 and CHL1 (375 Ϯ 20). A lower number of substrate-attached cells and significantly lower levels of neurite outgrowth were observed when either ADAM8 or CHL1 were used alone (100 Ϯ 5 and 180 Ϯ 7, respectively), or when ADAM10 (160 Ϯ 10) and ADAM17 (168 Ϯ 3) were used. Instead of EQ-ADAM8, we used batimastat (BB-94) for inhibition of CHL1 cleavage in CHL1/ADAM8 co-transfected cells (170 Ϯ 15). When shedding of CHL1 was inhibited by BB-94, substrate attachment was reduced below the basal level obtained with CHL1 alone. This difference suggests that, in addition to ADAM8, another unknown protease that is inhibited by BB-94 could be involved in ectodomain shedding of CHL1, although at a much lower level.
Reduction of Cell Death by Soluble CHL1-The time course of cell death of cerebellar granule neurons was determined by using serum-free medium from COS-7 cells transfected either with CHL1 alone or with CHL1/ADAM8 (Fig. 7). As estimated by the trypan blue exclusion method, the number of surviving cells declined to 19% Ϯ 1.5% after 5 days in culture when the conditioned medium was derived from either only CHL1 or CHL1/EQ-A8 transfected cells. Similar values were obtained when neurons were treated with supernatants from CHL1/ ADAM10 and CHL1/ADAM17 co-transfected cells (data not shown). In contrast, a significantly higher proportion (36% Ϯ 1.8%) of cells survived when conditioned medium from COS-7 cells co-transfected with ADAM8 and CHL1 was applied to neuronal cultures. Thus, survival of neurons was increased by ϳ50% due to the presence of soluble CHL1 derived by proteolytic processing. DISCUSSION We present the first evidence that ectodomain shedding of the neural cell adhesion molecule CHL1 by the metalloprotease-disintegrin ADAM8 has a physiological role by providing soluble CHL1, which is able to promote neurite outgrowth and reduce cell death of cerebellar neurons. CHL1 cleavage by ADAM8 is specific, because neither ADAM10 nor ADAM17, both expressed in brain in significantly higher levels than ADAM8, were able to cleave CHL1. In addition, L1-Fc and NCAM-Fc fusion proteins were not cleaved by ADAM8 suggesting a defined substrate-protease relationship. Given the fact that the metalloprotease domain of ADAM8 was used for our studies, we conclude that other domains of ADAM8 are not essential for the observed substrate specificity. It is interesting to note that, in addition to cleavage in the fibronectin type III homologous domain (FN III) 2 (position 753, WKP2QGAPE), the main cleavage site in the CHL1 protein is within the fifth FN domain of CHL1 in the juxtamembranous region (position 1040, LVP2GAEHIV). Under non-reducing conditions, the 125-kDa fragment of CHL1 was hardly detectable, suggesting that the cleavage site generating this fragment is located between two cysteine residues forming a disulfide bond under native conditions. This argues for the membrane proximal region to contain the physiologically relevant cleavage site of CHL1 for the release of the extracellular domain rather than the membrane distal one. The fifth FN III domain is not present in other related immunoglobulin superfamily members and has the lowest homology between L1 and CHL1, which could explain the clear-cut substrate specificity. It may be interesting to note that a weak similarity of the 165-kDa CHL1 cleavage site was found for the autocatalytic cleavage site of ADAM8 (40% within a 10-amino acid stretch from P5 to P5Ј: ADAM8, NDLGP2RALEI; CHL1, EVLVP2GAEHI). Experiments to introduce point mutations or a 10-amino acid deletion into the 165-kDa cleavage site of CHL1 did not provide sufficient information about critical amino acid residues, because the observed deficit in proteolytic cleavage is most likely due to the finding that mutated CHL1 proteins did not reach the cell surface. Thus the physiological consequences of the ADAM8mediated proteolytic cleavage of CHL1 could not be investigated.
In cell-based assays, release of the extracellular domain of CHL1 by ADAM8 was not enhanced by activators of ectodomain shedding such as protein kinase C activators, for instance the phorbol ester TPA or by the phosphatase inhibitor pervanadate. This is in contrast to numerous reports on activation of TACE/ADAM17-induced ectodomain shedding by these compounds, suggesting that the mechanisms of intracellular regulation of proteolytic activity for ADAM8 is different from that of ADAM17 involving protein kinase C epsilon (32).
In neurite outgrowth stimulation experiments, soluble CHL1 was ϳ10 times more potent than the CHL1-Fc protein. This observation supports the notion that proteolytic release of CHL1 is of high functional relevance in situ. Release of CHL1 into the extracellular milieu would allow the molecule to diffuse away from the cell surface and to bind to receptor(s) remote from the original site of CHL1 synthesis. The one or more receptors have so far remained elusive.
A role for CHL1 in cell migration has recently been demonstrated (23). In this study, ␣ 1 ␤ 1 and ␣ 2 ␤ 1 integrin-mediated cell migration toward collagen I was enhanced by CHL1. Similar to L1, integrin-dependent cell migration could be potentiated by fragments released from membrane-bound CHL1 (22). Thus, ADAM8 may have multiple effects on cell adhesion, either by cleavage of CHL1 or by direct integrin binding via its disinteence of commonly used activators TPA (50 and 500 ng/ml) and pervanadate (1ϫ, 77 M; 10ϫ, 770 M); "ctrl," co-transfection of ADAM8 and CHL1 without activators. One day after transfection, medium was changed to serum-free conditions in the absence or presence of the activators. 6 h later supernatants were harvested.

FIG. 3. Cleavage of CHL1 by ADAM8 in COS-7 cells.
A, cell-bound protease activity in COS-7 after transient transfections, given as relative mean fluorescence Ϯ S.E. after a 1-h reaction. mock, pTarget without insert. For ADAM8, EQ-ADAM8, and ADAM10, the fluorescent MBP peptide was used; for ADAM17 the TNF-␣ peptide was used. B, cotransfection of CHL1/wild-type ADAM8 and CHL1/mutant EQ-ADAM8 in COS-7 cells. In the cell lysate (upper panel), the unprocessed 185-kDa form of CHL1 is detectable with CHL1 antibody. In serum-free media of cells co-transfected with A8 and CHL1, 165-kDa and 125-kDa fragments and the membrane stub (ϳ60 kDa) were detected by CHL1 antibody (middle panel). Co-transfection of CHL1 and EQ-A8 (Glu 330 to Gln exchange in the catalytic site) did not result in CHL1 shedding (right lane). In the right panel, co-transfections with ADAM10 and ADAM17 did not result in significant cleavage of CHL1. In addition, the supernatants of A8/CHL1-transfected cells were analyzed after a high speed spin ("spin") and under non-reducing ("nr") conditions without mercaptoethanol. Note that under both conditions, the 125-kDa band as well as the membrane stub fragment are absent. A8 and EQ-A8 and ADAM17 were detected by anti-Bi-Pro antibody 4A6 recognizing a tag in the C terminus of the proteins. For ADAM10, we used a polyclonal serum against the cytoplasmic tail. White arrowheads indicate proforms; filled arrowheads indicate processed forms of ADAM proteins. C, shedding experiments in the pres-grin domain (27) thus activating intracellular signal transduction pathways in coordination with CHL1.
In the central nervous system, ADAM8 and CHL1 are expressed in neurons and oligodendrocytes (19,20) so that processing of CHL1 by ADAM8 in these cells could occur in cis interaction on the same cell surface. Whether ADAM8 interacts directly of indirectly with CHL1 remains to be analyzed. An interesting possibility is the interaction of ADAM8/integrins FIG. 4. Cleavage of CHL1 in WR mutant and ADAM8-deficient mice. Brain homogenates from 30-day-old wild type C57BL/6 mice, age-matched homozygous Wobbler mutant mice (A) and homozygous ADAM8-deficient mice (B) were prepared from cerebellum/brain stem, and 20 g of extracts or ConA eluates was separated on 5% gels. Blots were probed with anti-CHL1 antibody and with anti-glyceraldehyde-3phosphate dehydrogenase as loading control. For ADAM8, ConA preparations were used and stained with A8 antibody directed against the cytoplasmic domain of ADAM8. In A, upper panel, a short exposure of the blot showing intensities of 185-and 165-kDa bands, respectively. In the longer exposure (below), cleavage of the125-kDa band was enhanced in WR mice. In B, brain extracts from wild type mice (ϩ/ϩ) and ADAM8-deficient mice (Ϫ/Ϫ) were probed for CHL1 (upper panel), ADAM8 (middle panel), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). with CD9 (tetraspanin), which would recruit CHL1 into a complex, in a mechanism similar to that discussed for the regulation of shedding of heparin-binding epidermal growth factor HB-EGF (33). Interestingly, under pathological conditions, ADAM8 is additionally expressed by astrocytes and microglia (13), and, for reactive astrocytes in the optic nerve, enhanced CHL1 expression was also reported (25). Thus, cleavage of CHL1 under pathological conditions could result in enhanced remodelling of reactive astrocytes extending their processes, a hallmark of many neuropathological states.
Based on the enhanced expression of ADAM8 in Wobbler mice, it is likely that increased CHL1 shedding under pathological conditions is an important event in the reactive activation of glial cells in the central and possibly in the peripheral nervous system. In addition, the proinflammatory cytokine TNF-␣ is an important signaling molecule in diseases of the central nervous system (34), where TNF-␣ induces expression of ADAM8 (13), which in turn may lead to enhanced proteolytic release of CHL1, as observed in the Wobbler mouse. In the cerebellum and brain stem of Wobbler mice, enhanced cleavage but not transcriptional up-regulation of CHL1 mRNA expression was observed. Thus, it is likely that TNF-␣ regulates CHL1 shedding via transcriptional activation of ADAM8. It remains to be determined whether enhanced CHL1 shedding in situ has similar effects as in cell culture particularly regarding support of axonal regeneration and suppression of neuronal cell death after trauma.