ADAM9 Inhibition Increases Membrane Activity of ADAM10 and Controls α-Secretase Processing of Amyloid Precursor Protein*

Background: Raising ADAM10 α-secretase activity has been considered as an attractive therapeutic option in Alzheimer disease. Results: Administration of the prodomain of ADAM9 (proA9) prevents ADAM10 shedding and increases membrane α-secretase activity in a neuronal cell line. Conclusion: Use of proA9 is a means to modulate cellular ADAM10 activity. Significance: proA9 can be used in vivo to increase ADAM10 α-secretase activity. Prodomains of A disintegrin and metalloproteinase (ADAM) metallopeptidases can act as highly specific intra- and intermolecular inhibitors of ADAM catalytic activity. The mouse ADAM9 prodomain (proA9; amino acids 24–204), expressed and characterized from Escherichia coli, is a competitive inhibitor of human ADAM9 catalytic/disintegrin domain with an overall inhibition constant of 280 ± 34 nm and high specificity toward ADAM9. In SY5Y neuroblastoma cells overexpressing amyloid precursor protein, proA9 treatment reduces the amount of endogenous ADAM10 enzyme in the medium while increasing membrane-bound ADAM10, as shown both by Western and activity assays with selective fluorescent peptide substrates using proteolytic activity matrix analysis. An increase in membrane-bound ADAM10 generates higher levels of soluble amyloid precursor protein α in the medium, whereas soluble amyloid precursor protein β levels are decreased, demonstrating that inhibition of ADAM9 increases α-secretase activity on the cell membrane. Quantification of physiological ADAM10 substrates by a proteomic approach revealed that substrates, such as epidermal growth factor (EGF), HER2, osteoactivin, and CD40-ligand, are increased in the medium of BT474 breast tumor cells that were incubated with proA9, demonstrating that the regulation of ADAM10 by ADAM9 applies for many ADAM10 substrates. Taken together, our results demonstrate that ADAM10 activity is regulated by inhibition of ADAM9, and this regulation may be used to control shedding of amyloid precursor protein by enhancing α-secretase activity, a key regulatory step in the etiology of Alzheimer disease.


Prodomains of A disintegrin and metalloproteinase (ADAM)
metallopeptidases can act as highly specific intra-and intermolecular inhibitors of ADAM catalytic activity. The mouse ADAM9 prodomain (proA9; amino acids 24 -204), expressed and characterized from Escherichia coli, is a competitive inhibitor of human ADAM9 catalytic/disintegrin domain with an overall inhibition constant of 280 ؎ 34 nM and high specificity toward ADAM9. In SY5Y neuroblastoma cells overexpressing amyloid precursor protein, proA9 treatment reduces the amount of endogenous ADAM10 enzyme in the medium while increasing membrane-bound ADAM10, as shown both by Western and activity assays with selective fluorescent peptide substrates using proteolytic activity matrix analysis. An increase in membrane-bound ADAM10 generates higher levels of soluble amyloid precursor protein ␣ in the medium, whereas soluble amyloid precursor protein ␤ levels are decreased, demonstrating that inhibition of ADAM9 increases ␣-secretase activity on the cell membrane. Quantification of physiological ADAM10 substrates by a proteomic approach revealed that substrates, such as epidermal growth factor (EGF), HER2, osteoactivin, and CD40-ligand, are increased in the medium of BT474 breast tumor cells that were incubated with proA9, demonstrating that the regulation of ADAM10 by ADAM9 applies for many ADAM10 substrates. Taken together, our results demonstrate that ADAM10 activity is regulated by inhibition of ADAM9, and this regulation may be used to control shedding of amyloid precursor protein by enhancing ␣-secretase activity, a key regulatory step in the etiology of Alzheimer disease.
Members of the A disintegrin and metalloproteinase (ADAM) 2 family of proteinases function as sheddases by cleaving type I and type II integral single membrane proteins to generate soluble forms of these proteins (1,2). In addition to their catalytic and disintegrin domains, ADAM family members contain a prodomain that is required for expression, folding, and efficient transport of the proteinase (3). ADAM prodomains are efficient and highly specific inhibitors for their cognate ADAM proteases, and even when proteolytically cleaved from their nascent polypeptides, they can still serve as intermolecular inhibitors of their respective ADAM proteins (4,5). ADAM proteinases have functions in the etiologies of a variety of diseases and conditions. ADAM17 and -10, for example, are known sheddases for over 40 different type I and II integral membrane proteins (6,7), and improper regulation of their catalytic activities is implicated in neurological diseases (8,9), carcinogenesis (10 -12), and inflammatory conditions (13)(14)(15)(16)(17)(18). In particular, the cleavage of amyloid precursor protein at the ␣-secretase site is considered as instrumental in prevention of Alzheimer disease by favoring the non-amyloidogenic pathway (19). In neurons, ADAM10 has been demonstrated to be essential for the ␣-secretase step of amyloid precursor protein (APP) processing (20). Therefore, the means to control ␣-secretase levels are of therapeutic relevance in Alzheimer disease.
Understanding the hierarchy of ADAM proteinases could aid in determining broader biological mechanisms and reveal an additional level of ADAM regulation. For example, Cissé et al. (21) demonstrated, using transfection experiments, that ADAM10 ␣-secretase activity in mouse embryonic fibroblasts and a neuronal cell line could be increased by ADAM9 overexpression, possibly by affecting the shedding or release of ADAM10 from the membrane of cells. Recently, Tousseyn et al. (22) provided direct evidence that ADAM9 and ADAM15 process ADAM10, leaving behind an intracellular cytoplasmic domain with a potential function in regulating transcription of genes in the cell nucleus. This type of cleavage event is known as regulated intramembrane proteolysis (23). However, the physiological consequences of this cleavage step have remained elusive. A specific inhibitor of ADAM9 could be used to reveal the in vivo effect of ADAM10 shedding (i.e. to investigate a possible regulation of ADAM10-dependent shedding events).
Typically, forced expression of ADAM family members in mouse embryonic fibroblasts derived from knock-out mice and siRNA-mediated silencing have been used as tools to validate the role of a particular disintegrin metalloproteinase in shedding events. We have chosen to use specific inhibitors in order to understand how modulation of only the enzyme's catalytic activity affects cellular processing because with pharmaceutical agents, activities oftentimes are regulated, whereas the gene product remains intact. To date, the only available specific inhibitors of ADAM family members are small molecules described by Incyte (24,25), and protein therapeutics using modified tissue inhibitor of metalloproteinases (26), the prodomains of ADAM17 and -10 (4, 5), and an antibody to ADAM17 (27). Therefore, studies were undertaken to express, refold, purify, and examine a prodomain construct based on ADAM9 to easily achieve the highest degree of specificity for ADAM9 inhibition. A number of parameters were varied to obtain prodomain in milligram quantities that had refolded properly as assessed by inhibition studies with ADAM9. We demonstrate that the prodomain is a specific inhibitor of ADAM9 and show that ADAM9 regulates the cellular activity of ADAM10. Furthermore, proA9 was also used as a tool to demonstrate that specific inhibition of endogenous ADAM9 catalysis increases shedding of ADAM10 substrates in cellular assays.

Methods
Cloning of ADAM9 cDNA-A DNA fragment containing the ADAM9 prodomain (residues 24 -204) was cloned into a modified PET vector at the NdeI, BamHI restriction sites. The modified PET vector encodes His 6 between NdeI and BamHI sites to produce a protein with a N-terminal His tag. DNA primers were as follows: N-His(24 -204), 5Ј-primer, GGA GCC CAT ATG CCA GTC CTC GAG GCC GGG CGA; 3Ј-primer, GGA GCC GGA TCC TTA TCT GCG CAG CTG AGT GAC.
Expression and Purification of Soluble Prodomain-The construct was transformed into E. coli strain BL21(DE3)STAR (Invitrogen). For a typical sample preparation, bacteria were grown in 4 ϫ 1 liter of Luria broth (LB) at 37°C until the A 600 reached 0.4. The culture was incubated at 20°C for 30 min before adding isopropyl-␤-D-thiogalactopyranoside (1 mM) to induce protein expression. Cells were harvested after 16 h by spinning at 4°C for 30 min at 4000 rpm in a Sorvall JA 10 rotor. The supernatant was removed, and pellets were either stored at Ϫ70°C or used directly.
Cell pellets were lysed in 30 ml of buffer containing 50 mM phosphate, pH 8.0, 10 mM imidazole, 0.05% ␤-mercaptoethanol, and 300 mM NaCl at 4°C containing cell lytic (Sigma-Aldrich) (3 ml of a 10ϫ concentrated solution). Lysed bacteria were sonicated to shear the DNA and RNA, and then polyethyleneimine was added to 0.1% to precipitate the DNA. Samples were centrifuged at 13,000 rpm for 30 min. The cleared supernatant was applied to 1 ml of TALON resin (Clontech, Mountain View, CA) pre-equilibrated with lysis buffer without cell lysate. After two 5-ml washes with lysis buffer and 50 mM phosphate, pH 8.0, containing 20 mM imidazole, 0.05% ␤-mercaptoethanol, and 300 mM NaCl, the protein was eluted with a solution containing 50 mM phosphate, pH 8.0, 150 mM imidazole, 0.05% ␤-mercaptoethanol, and 300 mM NaCl. The eluted protein was concentrated using an Amicon Ultra filtration device (molecular mass cut-off 10 kDa) from Millipore (Billerica, MA) and further purified with a Sephacryl-S200 column (320 ml) on an Akta FPLC system at a flow rate of 1 ml/min. FPLC buffers contained 25 mM Tris, pH 8.0, 100 mM NaCl, and 0.05% ␤-mercaptoethanol. Fractions containing protein were concentrated and passed through an Endotrap blue column from Profos to remove endotoxin and then stored as 10% glycerol stocks at Ϫ80°C.
Expression, Refolding, and Purification of Insoluble Prodomain-Colonies of freshly transformed ArcticExpress TM (Agilent Technologies, Santa Clara, CA) or BL21(DE3) (Invitrogen) were used in all refolding and expression experiments. Cells from a 1-liter overnight culture grown in LB containing ampicillin and gentamycin were centrifuged at 5500 ϫ g and resuspended in 50 ml of LB broth. Twenty-five milliliters of this suspension was used to inoculate 1 liter of LB containing ampicillin. For the ArcticExpress conditions, cultures were incubated at 10°C with shaking for 2 h, induced by adding isopropyl-␤-D-thiogalactopyranoside to 0.2 mM, and grown for an additional 20 h. Cells were harvested by centrifugation for 15 min at 5500 ϫ g at 4°C.
Inclusion bodies containing proA9 were isolated from cells lysed in 5 volumes of Bug Buster Master Mix (Novagen), 0.5 mg/ml lysozyme (Sigma-Aldrich), 5 mM MgCl 2 , and 5 mM NaATP, containing Complete TM EDTA-free proteinase inhibitors (Roche Applied Science), per gram of cell paste. The lysis suspension was incubated for 30 min at room temperature with gentle agitation and centrifuged for 30 min at 16,000 ϫ g at 4°C to collect the inclusion bodies. Purification of inclusion bodies was accomplished by washing twice versus 5 volumes of 0.1ϫ Bug Buster Master Mix and 2 times versus 5 volumes of water. The resulting pellets were resuspended in water or 50 mM Tris-Cl, pH 8.0, and stored frozen at Ϫ80°C. Refolding conditions were established using the HiPER-FOLD TM starter kit from Barofold.
Using the best refolding conditions determined above, inclusion bodies were added to buffer containing 50 mM CHES, pH 9, and 5 mM TCEP and then placed under pressure in a Barofold apparatus for 24 h at room temperature. Soluble protein after the pressure was released, was purified further by passage through 10 ml of Ni 2ϩ -NTA resin (Qiagen, Valencia, CA) followed by washes of 10 and 20 mM imidazole and elution with 250 mM imidazole in buffers that contained 4 mM TCEP, 50 mM NaP i , pH 8, and 300 mM NaCl. The eluted protein was concentrated to less than 5 ml using an Amicon Ultra filtration device (molecular mass cut-off 10 kDa) from Millipore (Billerica, MA) and further purified with a Sephadex-16/60 column (120 ml) on an Akta FPLC system at a flow rate of 1 ml/min. FPLC buffers contained 25 mM Tris, pH 8.0, 100 mM NaCl and 4 mM TCEP. Fractions containing protein were concentrated, passed through an Endotrap Blue column from Hyglos/Profos, and stored as 10% glycerol stocks at Ϫ80°C. In some experiments, proA9 was refolded from inclusion bodies from BL21(DE3) purified via a nickel column, and then dialyzed against 25 mM CHES, pH 9, 100 mM NaCl, and 0.035% ␤-mercaptoethanol or passed through a Sephadex-16/60 column using the same buffer conditions. In addition, proA9 was sometimes refolded from urea after purification with Ni 2ϩ -NTA-agarose and then further purified via a Sephadex 16/60 column.
Inhibition Assays-Fluorescence intensities were measured every 2 min at excitation and emission wavelengths of 485 and 530 nm, respectively, at room temperature in black-coated 96-well plates using a top to bottom Fluoroskan II fluorometer. Activity of TNF-␣-converting enzyme catalytic disintegrin, ADAM8, ADAM12, and ADAM10 catalytic/disintegrin recombinant proteins was monitored as described previously using the fluorescent substrate PEPDab005 (28). Inhibition assays were done using up to 8.4 M proA9. ADAM9 activity was assayed using the fluorescent substrate PEPDab010 (29) in 70 l of buffer with 5 l of proA9 (75-8000 nM final) or buffer control. For each proA9 concentration, a well containing only proA9 and substrate was used to correct for fluorescence fluctuations due to the prodomain itself. ADAM9 was provided as a 0.1 g/l stock solution, which was diluted 1:50, from which 8 l was added to start the reaction. The functional purity of ADAM9 was confirmed by using hydroxamate inhibitors and the determination of IC 50 values, which were compared with literature values (29).
Measurement of ADAM10 Activity in the Medium and in Membrane Fractions-SY5Y neuroblastoma cells were cultured in duplicate using a 6-well plate for 24 h in serum-free medium with proA9 (5 M) or a buffer control. Medium was removed, spun to remove cell fragments, and set aside for assaying. Cells were washed with PBS and then scraped from the plate and resuspended in a 1.5-ml tube in a cold solution of 0.25 M sucrose, 50 mM Tris, pH 8, and a protease inhibitor mixture from Roche Applied Science. Cells were broken via pipetting up and down, and the suspension was spun at 13,000 ϫ g to pellet the membranes, which were resuspended and washed with sucrose buffer. After pelleting, membranes were resuspended in 200 l of sucrose buffer/well of cells. Protein concentrations were determined using the Bio-Rad BPA assay.
The medium and membrane suspension were tested for ADAM10 activity by using the proteolytic activity matrix analysis (PrAMA) technique developed by Miller et al. (30) using substrates PEPDab005, PEPDab010, PEPDab008, PEPDab013, and PEPDab014, which varied in their specificities toward different ADAM family members and MMPs. Briefly, 12.5 M substrate concentrations in ADAM buffer (60 l) were incubated with either 20 l of medium or 10 l of resuspended membranes. Fluorescence units versus time were monitored with a Fluostar BMG Optima using excitation and emission wavelengths of 485 and 530 nM respectively. Data were analyzed by taking the initial velocities and fitting to a matrix developed previously using the PrAMA technique.
Calculation of K i Values-The fractional inhibitor activity (I f ) was calculated by dividing the initial velocity from the FU versus time graph obtained with inhibitor (v i ) by the initial velocity without inhibitor (v o ). All inhibition assays were performed at room temperature, and the data were fit to the following equation using Sigma Plot software, where I f is fractional inhibition, I is the inhibitor concentration, and K i(app) is the inhibitor concentration that gives 50% inhibition.
Mechanism of Inhibition-ADAM9 was diluted 1:500 in buffer (25 mM Tris, pH 8, 6 ϫ 10 Ϫ4 % Brij-35), and 35 l was added to a 96-well plate. Various concentrations of prodomain (2-l volume) in 25 mM Tris-Cl, pH 8, 100 mM NaCl, 10% glycerol, 4 mM TCEP, and 10 mM CaCl 2 were added to the enzyme, and preincubation was carried out for 15 min. Then 33 l of substrate (5-100 M) of PEPDab010 was added, and fluorescence intensities were measured with excitation at 485 nm and emission at 530 nm. The final concentration of prodomain ranged from 75 to 4000 nM. Fluorescence versus time was plotted, and slopes were derived from straight line fits of initial velocities. The reciprocals of the velocities were plotted versus the reciprocals of substrate squared because the normal Lineweaver-Burk plot gave curved lines. The initial velocities versus substrate concentrations were fit as a family of curves to several allosteric models. The data fit best to a pure competitive model as described in Equation 2, where binding of inhibitor prevents substrate from binding to both sites.
In this equation, v is the velocity obtained from the fluorescence units versus time plot, V is the maximum velocity, S is the substrate concentration, I is the inhibitor concentration, K s is the dissociation constant for substrate binding to enzyme, and K i is the inhibition constant.
Cell Culture-SH-SY5Y-APP swe neuroblastoma cells expressing APP with the Swedish mutation (APP swe ; APP K595N , APP M596L ) were cultured in DMEM supplemented with 10% FBS, 2 mM glutamine, and 400 g/ml G418. For ADAM10 and APP shedding analyses, cells were maintained in serum-free medium for 6 and 12 h. Cell culture supernatants were analyzed by Western blotting and ELISA after concentrating using Vivaspin columns (Generon, UK) with an exclusion range of 20 kDa.
Western Blotting-For Western blot sample preparations, cell supernatants were concentrated and mixed with 5ϫ SDS sample buffer (50% glycerol, 10% SDS, 100 mM ␤-mercaptoethanol, 0.05% bromphenol blue in 250 mM Tris-HCl, pH 6.8), boiled for 10 min, and loaded onto 7.5% SDS-PAGE for soluble APP detection. After gel runs, the proteins were blotted onto nitrocellulose transfer membranes (Protran BA79, 0.1-mm pore size, Schleicher & Schuell) by tank blotting. After control staining with 0.1% Ponceau S solution and blocking for 2 h with 5% skim milk in TBS containing 0.1% Tween 20, proteins were analyzed by immunostaining using antibodies described above in the given concentrations.
ELISAs-Concentrations of soluble APP␤ (sAPP␤) were determined using the Meso Scale Discovery (MSD) sAPP␤ assay and MSD antibody triplex assay as described by the manufacturer (Meso Scale Discovery, Gaithersburg, MD). Concentrations of sAPP␣ were measured using a biotinylated antibody (clone 14D6) with the same specificity as the 4B4 (20). The streptavidin-precoated 96-well plates (MSD, L15SA-1/ L11SA-1) were blocked with blocking solution for 30 min, washed three times with Tris Wash Buffer, and coated with the biotinylated antibody 14D6 for 1 h. After washing three times, the wells were incubated with 25 ml of standards or samples and 25 ml of the anti-rabbit 5313 antibody for 2 h. The wells were washed three times and incubated with 25 ml of MSD Sulfo-Tag goat anti-rabbit antibody (MSD, R32AB-5) as detection antibody. The wells were washed three times and incubated with the MSD Read Buffer T, and the MSD plates were measured on the MDS Sector Image 600 plate reader. The raw data were measured as electrochemiluminescence signal detected by photodetectors.
Statistical Analysis-Data were presented as mean values Ϯ S.D. Student's t test was applied for data analysis. p values of Ͻ0.05 were considered significant.
Peptide Array Experiments-BT474 breast tumor cells were obtained from ATCC and were grown in RPMI medium containing 10% fetal bovine serum. The cells were seeded at 1 ϫ 10 6 cells/ml and plated in either 96-or 6-well plates. After overnight incubation, cells were placed in minimum essential medium without serum. Prodomain of ADAM9, ADAM10, or a buffer control was added to the medium after filter sterilization with a 0.22-m filter. The final concentration of glycerol was adjusted to 1%. After 24 h, medium was collected and frozen. The frozen medium was analyzed by RayBiotech in an L series 507 microarray. Data were normalized to both positive and negative controls that were run alongside samples. Each run represents averages of triplicate values after normalization. Ratios of control versus treated were determined for two separate runs. Data were presented as mean values. A value between 0.8 and 1.4 for an average of two runs indicates no change in factor levels, whereas a value greater than 1.4 indicates an increase and a value below 0.8 means a decrease of factor in the medium of treated cells.

Generation and Characterization of ADAM9 Prodomain-A
bacterial expression construct for the prodomain of ADAM9 (proA9) was generated, excluding the N-terminal hydrophobic signal sequence region and terminated at the furin cleavage site (amino acids 24 -204 of mADAM9). A construct encoding a C-terminal His-tagged protein rendered a completely insoluble protein (data not shown), so that the N-terminal construct was used in all studies.
Because the yield of recombinant protein was low (Ͻ50 g/liter), refolding of proA9 was done from inclusion bodies. Once refolded, samples were analyzed for protein content by SDS-PAGE and for inhibitory potency against ADAM9 to assess if the prodomain had refolded properly (Table 1 and Fig.  1A). The optimal refolding conditions proved to be 50 mM CHES, pH 9, with 5 mM TCEP without arginine.
proA9 purified from the insoluble fraction after refolding had identical properties to the proA9 obtained from the soluble fraction (Fig. 1, B and C). Using this proA9 preparation, inhibition studies against several recombinant ADAM proteases were carried out to evaluate specificity. Recombinant proA9 is an

Corresponding inhibitory activity of ADAM9 using a fixed volume of buffer containing proA9 from different refolding conditions
Refolding conditions for proA9 varied in pH values, presence of arginine (R), and/or glutathione oxidized and reduced mixture (Redox). Note that identical sample numbers 1-13 are shown in the lanes of Fig. 1A.  Fig. 2A and Table 2). ADAM8 and -12 are weakly inhibited by proA9 with approximate apparent K i values in the micromolar range (Ͼ1 and Ͼ6 M, respectively). Recombinant ADAM10 and ADAM17/TNF-␣-converting enzyme proteases were not inhibited by proA9 even at concentrations of Ն8 M (Table 1). Given these parameters, a low micromolar concentration of proA9 can be used in cell-based assays to inhibit ADAM9 activity. Mechanism of Inhibition-proA9 was tested to determine the mode of inhibition for human ADAM9. A typical Lineweaver-Burk plot, where the substrate was varied as a function of inhibitor concentration, gave curved lines for the reciprocal plot. Thus, a 1/v versus 1/S 2 plot was used and gave a good fit with varying slopes but with the intercepts remaining the same (Fig.  2B). This indicated that proA9 is a competitive inhibitor of ADAM9. This analysis also indicated that ADAM9 has at least two binding sites for the fluorescent substrate. The velocities versus substrate at varying inhibitor values were fit as a family of curves to the pure competitive inhibitor model. The K i value calculated with this method was 240 Ϯ 110 nM, and the binding constant, K s , for substrate was 37 Ϯ 9 M. The K i is in close agreement with what was found for the apparent inhibition constant, K i(app) , calculated from the inhibition data (Equation 1).

Lane number Sample ID Inhibition
proA9 Inhibits ADAM10 Shedding and Increases ␣-Secretase Activity on the Cell Membrane-In an earlier report, shedding of ADAM10 by ADAM9 and -15 was described, but the physiological consequences of this process remain elusive (21,22). We used the proA9 as a specific inhibitor to address this question. To provide direct evidence for ADAM9 regulating ADAM10 activity, SH-SY5Y neuroblastoma cells overexpressing APP were treated with proA9 at concentrations of 1 and 5 M, and the extent of soluble ADAM10 was determined by Western blotting using an antibody directed against the MP

TABLE 2 Inhibition of protease activity by proA9 for different recombinant ADAM proteins
ProA9 was incubated with the human recombinant ADAM proteins indicated, and the amount of inhibition was determined by fluorometric assay as described under "Experimental Procedures."

Recombinant protease
ProA9 ADAM9 Affects Cellular ADAM10 Activity NOVEMBER 25, 2011 • VOLUME 286 • NUMBER 47 domain of ADAM10 (Fig. 3A). ADAM9 inhibition by proA9 resulted in reduced release of soluble ADAM10 (30 Ϯ 4% for 1 M, 15 Ϯ 5% for 5 M) in the medium. Furthermore, a method developed previously, termed PrAMA (30), was used to quantify the amount of proteolytically active ADAM10 in the medium. PrAMA employs a panel of synthetic FRET-based polypeptide protease substrates to record a quantitative signature of cleavage rates for a given biological sample. Prior knowledge of individual MMP/ADAM cleavage specificity profiles measured with purified enzymes allows PrAMA to decipher specific enzyme (e.g. ADAM10) activities from observed cleav-age signatures. In this work, we used five different proteinase substrates, as described under "Experimental Procedures," that vary in their specificity for ADAMs and MMPs to infer the activities of MMP9, ADAM10, and ADAM17. PrAMA analysis confirmed a decrease in the amount of ADAM10 activity in medium from SY5Y cells with proA9 treatment (Fig. 3B), whereas no significant change in MMP9 or ADAM17 activity was observed (data not shown). These results demonstrate that inhibition of ADAM9 leads to a significant reduction of  ADAM10 release and confirm previous findings that ADAM9 is a processing enzyme for ADAM10 (21,22,31). We next tested whether this reduction caused an increase in the membrane-associated activity of ADAM10. Western analysis and FRET assays of the membrane fraction were accomplished and analyzed by PrAMA. The amount of ADAM10 as measured by both of these techniques indicated an increase in the cellular fraction after proA9 treatment of the SY5Y cells (Fig. 3C). As with analysis of the medium, PrAMA did not detect significant changes in ADAM17 activity upon proA9 treatment. Because ADAM10 activity was increased on the membrane of cells given a specific inhibitor of ADAM9, the release of APP, a physiological substrate for ADAM10, was quantified by ELISA from SY5Y cells to determine if soluble APP levels increased in medium from proA9-treated cells (Fig.  4, A-C).
Western analysis of whole cell lysate and supernatant from SY5Y cells after proA9 treatment indicated similar amounts of APP (Fig. 4A). However, when medium was collected after 6and 12-h treatments with proA9 or a buffer control and the amounts of sAPP␣ and sAPP␤ were analyzed via a multiplex ELISA, a dose-dependent increase in sAPP␣ and corresponding decrease in sAPP␤ were detected, indicating that the ␣-secretase activity of ADAM10 can be regulated by specific inhibition of ADAM9 (Fig. 4, B and C).
Peptide Array Screening-Having shown that ADAM9 regulates ADAM-shedding activity, a protein array analysis was performed to obtain a more general screen on the regulation of ADAM9 and ADAM10 activity in cancer cells. The Ray Biotech L series protein array was chosen because it had been used successfully in other published assays to quantify protein levels in samples from cell medium or biological fluids (32). In the RayBiotech patented approach, medium samples, either treated or buffer control, were biotin-labeled and incubated on a glass chip printed with 507 different antibodies. Briefly, BT474 cells were treated with 5 M proA9. As a comparison with proA9 inhibition, proA10 was used as a specific inhibitor of ADAM10-dependent shedding events.
Using this method, three different groups of proteins can be classified (Table 3): an increase in the medium of physiological substrates for ADAM10 after proA9 treatment (proA10 treatment resulted in a reduction of the same substrates), a decrease in the medium of factors after proA9 but not proA10 inhibition, and an attenuation of factors found in the medium after either proA9 or proA10 treatment.
In the first group, endogenous shedding of known ADAM10 substrates, such as Her2, osteoactivin, CD40-ligand, and EGF, was down-regulated by proA10 and up-regulated by proA9 addition to the BT474 cells. In contrast, novel factors possibly shed by ADAM9 and not by ADAM10 were also found, such as MAC-1, glucocorticoid-induced tumor necrosis factor receptor, platelet-derived growth factor receptor ␣, and transforming growth factor-␣ (TGF-␣). Several proteins decreased in the medium from cells treated with either inhibitor, such as neuropilin 2, lipopolysaccharide-binding protein, and OX40-ligand. In addition, TNF receptor 1 and TNF-␣/amphiregulin (AR), physiological substrates for ADAM8 (33) and ADAM17 (34 -36), respectively, were not affected by proA9 or proA10 treatments, indicating that the inhibitors are quite selective in their actions (data not shown).

DISCUSSION
With respect to the selectivity of proA9, the observed IC 50 for inhibition of ADAM8 is slightly higher than 1 M (see Table 2), so the K i difference is about 5-fold. However, it is likely that some inhibition of ADAM8 may occur at 5 M proA9 concentration, although in BT474 cells, the presence of both L-selectin and TNF receptor 1 in the medium, two substrates for ADAM8, is not affected by proA9 treatment. This could be due to very low ADAM8 levels in BT474 cells that was confirmed by RNA TABLE 3 Ratios of Type I and II integral membrane proteins in medium after proA9 and proA10 treatment of BT474 cells proA9 (5 M) or proA10 (23-213 wild type, 10 M) and corresponding buffer controls were incubated alongside BT474 cells for 24 h in serum-free conditions. Media were collected and analyzed by the RayBiotech 507 L series peptide array. Data represent the average ratio from two independent runs done in triplicate. analysis and by a PrAMA experiment where the substrate specific for ADAM8 (PEPDab013) is only slightly processed in both membrane and medium samples, although other less selective substrates with the same k cat /K m for ADAM8 are processed very well (data not shown). In contrast, no inhibition of ADAM17 occurred even in enzymatic assays using 8 M proA9, indicating that the inhibitor has a 13-130-fold selectivity for ADAM9 over these other family members. Because no inhibition was observed, an upper limit can only be estimated. Because ADAM10 has been described as the essential ␣-secretase in neurons, the control of neuronal ADAM10 activity has crucial importance in the etiology of Alzheimer disease (20), and attempts to raise cellular ADAM10 levels have been considered as potential therapeutic strategies in Alzheimer disease (19). We found that specific proA9-mediated inhibition of ADAM9 catalytic activity in SY5Y neuroblastoma cells regulates ADAM10 activity on the cell membrane by preventing its proteolytic release. Because the amount of membrane-bound ADAM10 increases when its shedding is reduced, we postulate that the enzyme's cleavage activity would correspondingly be increased. In accordance with these findings, proA9 inhibition of ADAM9, a principal sheddase for ADAM10, leads to an increase in the ␣-secretase activity for APP in cell-based assays, increasing the extracellular levels of sAPP␣, arguing for an inhibition of ADAM9 to favor a non-amyloidogenic pathway. Because ADAM9 is dispensable in vivo, as shown by ADAM9 deficient mice (37), this therapeutic strategy seems to be feasible.

Factor
In previous studies, loss and gain of function assays were performed in transfected cells and demonstrated that ADAM9 indirectly increases the amount of sAPP␣, probably through cleavage of ADAM10. Our findings demonstrate that an immediate reduction in soluble ADAM10 by inhibition of ADAM9 catalysis increases ADAM10 processing of APP at the ␣ cleavage site. This complements more long term experiments that require longer periods to constitute a complex interplay of ADAM proteins after knockdown (21) or overexpression of ADAM9 in the presence of siRNA for ADAM10 (31). In addition, we demonstrate the effect of endogenous inhibition of ADAM9 rather than a reduction in its activity after transfection.
Further proof that ADAM9 inhibition leads to an increase in membrane ADAM10 activity arises from results with the BT474 tumor cells, where increased levels of proteins for proven substrates of ADAM10 were found in the medium from proA9-treated cells. Of 507 different factors that can be detected in medium samples by the peptide array approach, only a handful are secreted at reduced levels, and several are type I or II membrane proteins capable of serving as substrates for ADAM9. These factors range from receptors to growth factors. The mechanism by which proA9 specifically reduces type I and II integral membrane protein levels in the medium is probably due to its inhibition of ADAM9 catalytic activity because it is a competitive inhibitor. In contrast, proA10 treatment affects ADAM10-dependent shedding events (over 100 factors are reduced in the medium from BT474-treated cells; data not shown) but has no effect in parallel assays on the newly identified potential substrates for ADAM9. These results indi-cate that generally specific inhibition of ADAM9 controls membrane ADAM10 activity, which could be of general in vivo relevance.