Autolytic Processing at Glu586-Ser587within the Cysteine-rich Domain of Human Adamalysin 19/Disintegrin-Metalloproteinase 19 Is Necessary for Its Proteolytic Activity*

We investigated the regulation of the proteolytic activity of human adamalysin 19 (a disintegrin and metalloproteinase 19, hADAM19). It was processed at Glu586(P1)-Ser587(P1′) site in the cysteine-rich domain as shown by protein N-terminal sequencing. This truncation was autolytic as illustrated by its R199A/R200A or E346A mutation that prevented the zymogen activation by furin or abolished the catalytic activity. Reagents that block furin-mediated activation of pro-hADAM19, decRVKR-CMK, A23187, and brefeldin A abrogated this processing. The sizes of the side chains of the P1 and P1′ residues are critical for the processing of hADAM19. The amount of processing product in the E586Q or S587A mutant with a side chain almost the same size as that in the wild type was almost equal. Conversely, very little processing was observed when the size of the side chain was changed significantly, such as in the E586A, E586G, or S587F mutants. Two mutants with presumably subtle structural distinctions from wild type hADAM19, E586D and S587T, displayed rare or little processing and had very low capacities to cleave α2-macroglobulin and a peptide substrate. Therefore, this processing is necessary for hADAM19 to exert its proteolytic activities. Moreover, a new peptide substrate, Ac-RPLE-SNAV, which is identical to the processing site sequence, was cleaved at the E-S bond by soluble hADAM19 containing the catalytic and disintegrin domains. This enzyme cleaved the substrate with K m , k cat, andk cat/K m of 2.0 mm, 2.4/min, and 1200 m −1min−1, respectively, using a fluorescamine assay. Preliminary studies showed that a protein kinase C activator, phorbol 12-myristate 13-acetate, promoted the cellular processing of hADAM19; however, three calmodulin antagonists, trifluoperazine, W7, and calmidazolium, impaired this cleavage, indicating complex signal pathways may be involved in the processing.

Adamalysin 19/ADAM19/MDC ␤, cloned from mice (42,43) and humans (44,45), is a type I membrane-bound protein containing the basic domains of ADAMs, such as the prodomain, metalloprotease and disintegrin domain, cysteine-rich domain, transmembrane domain, and cytoplasmic domain (3). In addition to the processing of neuregulin (NRG) (36), human adamalysin 19 (hADAM19) is believed to play a role in osteoblast differentiation, in the distinction between macrophages and dendritic cells, and as a marker for the differentiation and characterization of dendritic cells (44). Recently, we demonstrated that hADAM19 is activated by furin in the secretory pathway and that the intracellular removal of the prodomain is required for its proteolytic activity, which was assessed with an ␣2-macroglobulin (␣2-M) trapping assay in vitro (46). Emerging evidence indicates that metalloproteinase activity is regulated by the process of shedding or truncation, as in the cases of MT1-MMP, MT5-MMP, ADAM13, and ADAMTS4 (40,(47)(48)(49). Here we show that autolytic processing at Glu 586 2 Ser 587 of hADAM19 within its cysteine-rich domain is required for its endopeptidase activity and that efficient processing of hADAM19 is mainly dependent on the sizes of both Glu 586 and Ser 587 . We also show that PKC, CaM, and calcium signals may regulate the hADAM19 processing, which is not sensitive to GM6001 or TIMP-3. Moreover, we present a peptide substrate that mimics the processing site and may be used to determine the activity of soluble hADAM19 by a fluorescamine assay.
PCR Primers, Mutagenesis, and Expression Constructs-All inserts tagged with FLAG at their C terminus were cloned into pCR3.1uni, including wild type hADAM19 (F46), soluble hADAM19 (D52), 199 RA-D (described in Ref. 46), and all mutants used in this study. The primer sequences for full-length (E346A-F) and soluble (E346A-D) Glu 346 3 Ala mutants were: forward primer, 5Ј-C ATG GCC CAC GCG ATG GGC CAC-3Ј; reverse primer, 5Ј-GTG GCC CAT CGC GTG GGC CAT G-3Ј. For deletion from the cysteine-rich domain to the end of the C terminus (D-CR): forward primer, 5Ј-ACC ATG CCA GGG GGC GCA GGC GCC-3Ј; reverse primer, 5Ј-GGT ACC ATC CAT CTG GTA GAA G-3Ј. For the soluble Glu 586 3 Ala mutant (E586A-D): forward primer, 5Ј-CGG CCC CTG GCG TCC AAC GCG-3Ј; reverse primer, 5Ј-CGC GTT GGA CGC CAG GGG CCG-3Ј. For the soluble Glu 586 3 Gly mutant (E586G-D): forward primer, 5Ј-CGG CCC CTG GGG TCC AAC  GCG-3Ј; reverse primer, 5Ј-CGC GTT GGA CCC CAG GGG CCG-3Ј.  For the soluble Glu 586 3 Asp mutant (E586D-D): forward primer,  5Ј-CGG CCC CTG GAC TCC AAC GCG-3Ј; reverse primer, 5Ј-CGC  GTT GGA GTC CAG GGG CCG-3Ј. For the soluble Glu 586 3 Gln  mutant (E586Q-D DNA Transfection and Generation of Stable Cell Lines-COS1 cells were seeded into 24-well plates for 16 -24 h at 80% confluence and transfected with the indicated plasmids using LipofectAMINE2000 according to the instructions provided by Invitrogen (Gaithersburg, MD). After 6 -10 h, serum-free Dulbecco's modified Eagle's medium and the indicated reagents were added, and the mixture was incubated for another 24 h. The conditioned media and cell lysates were then analyzed by Western blotting (46). The same transfection procedure was performed to generate stable MDCK cell lines, and the selection for hADAM19 was begun in the presence of G418 (400 g/ml) after transfection for 24 h. The conditioned media and/or cell lysates of the clones were subjected to Western blotting to confirm the expression of hADAM19 (46).
Western Blotting-The experiments were carried out as described previously (46). Briefly, cells were grown to 80% confluence and were treated as indicated. After centrifugation for 15 min at 14,000 ϫ g and 4°C to clear any debris, the serum-free media were collected and prepared for SDS-PAGE. The cells were lysed with RIPA buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.25% sodium deoxycholate, 0.1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 2.5 M GM6001, 10 g/ml aprotinin, 10 g/ml E64, and 10 g/ml pepstatin A) for 15 min on ice. The supernatant was collected after centrifugation for 20 min at 14,000 ϫ g and 4°C. After electrophoresis, the proteins were transferred onto nitrocellulose membranes, probed with anti-FLAG-M2 or anti-hADAM19, and developed as before (46).
Purification of Soluble hADAM19 and N-terminal Sequencing-All proteins were purified on anti-FLAG-M2 affinity columns as described previously (46); however, HEPES buffer (50 mM HEPES, pH 7.5, 200 mM NaCl, 10 mM CaCl 2 , 25 M ZnCl 2 , 0.05% Brij-35) was used instead of TBS buffer for the purpose of determining the activity of hADAM19 by a fluorescamine assay. Briefly, cells from stable lines expressing soluble hADAM19, D52-5, E586D-D, S587T-D, and D-CR, were grown to 100% confluence, washed twice with phosphate-buffered saline, and incubated for 48 h. in serum-free media. The conditioned media were collected, centrifuged to clear any debris, and loaded onto an anti-M2 immunoaffinity column (1 ml of resuspended agarose) that had been prewashed with HEPES buffer. The bound materials were extensively washed with HEPES buffer, eluted with FLAG peptides, and collected in 500-l fractions. The fractions were analyzed by Western blot using anti-hADAM19 antibodies or anti-FLAG-M2, and the protein was quantified by its UV absorbance at 280 nm. The fractions containing the most wild type or mutant hADAM19 proteins were used for the ␣2-M trapping assay and fluorescamine assay. In the case of D52-5, the most concentrated fraction was also prepared for protein N-terminal sequencing to determine the shedding site. After separation by SDS-PAGE, the samples were transferred to a PVDF membrane and stained with Coomassie Blue R-250. After destaining, the hADAM19 bands were excised and sent to Margaret Seavy at the Bioanalytical Core Facility at the Florida State University for N-terminal amino acid sequencing.
␣2-M Trapping Assay-The detailed experimental procedure was previously reported (46,50). Briefly, equal amounts of purified wild type and mutated soluble hADAM19 were mixed with 24 l of ␣2-M (0.2 unit/ml), respectively, adjusted to a total volume of 100 l with HEPES buffer, and incubated at 37°C for the indicated times. A 20-l aliquot of the mixture was removed at the indicated times, put in 2ϫ SDS-PAGE sample buffer, and boiled. Following SDS-PAGE, the protein bands in the gels were visualized by silver staining.
Determination of Kinetic Parameters of hADAM19 Using a New Peptide Assay-The N-terminal acetylated peptide (Ac-RPLESNAV) was synthesized by Dr. Umesh Goli at the Biochemical Analysis, Synthesis, and Sequence Service Laboratory at Florida State University. The stock solutions of the peptide substrates were prepared in 0.05 M HEPES, pH 7.5, 0.2 M NaCl, 0.01 M CaCl 2 , and 0.01% Brij-35. The peptide concentrations ranged from 0.2 to 4.0 mM, and the enzyme concentration was 160 nM. N-terminal sequencing for this released peptide was performed after overnight hydrolysis by the enzyme. Hydrolysis of the peptide by the enzyme was monitored by measuring fluorescence intensity (51). After incubation for the indicated times, the hydrolytic reaction was quenched by adding 20 l of 100 mM 1,10-phenanthroline into 100 l of reaction mixture. For the control sets, the 1,10-phenanthroline solution was added at the start of incubation. After the reaction was quenched, 100 l of 0.1% fluorescamine in 10% Me 2 SO/90% assay buffer was added into each reaction mixture. Relative fluorescence of the product coupled with fluorescamine was determined on a PerkinElmer Life Sciences LS-50B spectrofluorometer using an excitation wavelength of 386 nm and an emission wavelength of 477 nm. Excitation and emission slit widths were both 10 nm. Relative fluorescence was converted to nanomoles of product using a standard curve obtained with arginine. The final solution was diluted further with the assay buffer, if it was necessary. The values of k cat and K m were determined by fitting the data to the Michaelis-Menten equation. A molar extinction coefficient of 33,120 M Ϫ1 cm Ϫ1 , which was calculated by using Genetic Computer Group (GCG) software, was used to determine the concentration of soluble hADAM19.

RESULTS
hADAM19 Is Processed within the Cysteine-rich Domain-Shedding or truncation is increasingly being recognized as a key regulator of the activity of some metalloproteinases (40,(47)(48)(49). It was intriguing to investigate the possibility of shedding or truncation in hADAM19. As shown in Fig. 1A, in addition to the pro and active forms of hADAM19, we detected a doublet of protein of about 58 kDa by anti-FLAG-M2 in the lysate of the transfected COS1 cells expressing full-length hADAM19 with C-terminal FLAG tags (F46). We did not detect the doublet using antibodies against the catalytic and disintegrin domains of hADAM19, indicating that the 58-kDa protein is the processed C-terminal fragment lacking the metalloproteinase and disintegrin domains of hADAM19. On the other hand, another doublet of around 72 kDa was detected in the same lysate by these two hADAM19 antibodies but not anti-FLAG-M2 (Fig. 1A), suggesting that the 72-kDa protein is the processed N-terminal fragment containing the metalloproteinase and disintegrin domains of hADAM19. Taken together, a cellular processing occurs in hADAM19 after its disintegrin domain in the COS1-transfected cells. Notably, the doublets result from different forms of glycosylation as verified by glycosidase F treatment (Ref. 46 and data not shown). To probe the domain in which the processing occurs, we made two deletion forms of hADAM19 with C-terminal FLAG tags, soluble hADAM19 (D52), which lacked the transmembrane domain and cytoplasmic domain, and a form that lacked the cysteinerich domain through the cytoplasmic domain (D-CR) (see Fig.  4). When D52 and D-CR were transfected into COS1 cells, we were able to detect several proteins in the conditioned media using both anti-FLAG-M2 and the antibody against the disintegrin domain, including the soluble pro and active forms of hADAM19. However, a 26-kDa protein in the conditioned medium from D52-transfected cells was clearly detectable using anti-FLAG-M2, but not the disintegrin domain antibody, suggesting that the 26-kDa protein is the processed C-terminal fragment of soluble hADAM19. In contrast, a doublet at 60 -65 kDa was detected in the medium by the anti-disintegrin domain antibody, but not anti-FLAG-M2, indicating that the 60to 65-kDa protein is the processed N-terminal fragment of soluble hADAM19 (Fig. 1B). There were no additional proteins detectable in the conditioned medium from the cells transfected with D-CR (data not shown); therefore, the processing of hADAM19 might be accounted for by shedding within its cysteine-rich domain.
PMA Up-regulates Processing of Full-length but Not Soluble hADAM19 -PMA is a potent inducer of ectodomain shedding of many transmembrane proteins (5, 8 -12, 14, 16, 19, 20, 34). To examine the effect of PMA on the processing of hADAM19, the COS1 cells transfected with F46 were treated with PMA overnight. As shown in Fig. 2A, PMA obviously enhanced hADAM19 processing in the COS1-transfected cells as detected under reducing conditions. Interestingly, we failed to detect the processed N-terminal fragments in the conditioned media from the F46-transfected COS1 cells, even after PMA treatment (data not shown), suggesting that the processed N-terminal fragment may be still associated with the remaining C-terminal fragment or full-length of hADAM19 via one or more disulfide bonds. Therefore, according to the definition of ectodomain shedding, membrane-bound proteins release their soluble forms and the processing of hADAM19 within its cysteine-rich domain is not technically a form of shedding. To test if PMAenhanced processing relies on the transmembrane and cytoplasmic domains, we treated the COS1 cells expressing D52 with PMA. As shown in Fig. 2B, PMA had a negligible effect on the processing of soluble hADAM19, suggesting that PMA might enhance the processing of hADAM19 via interaction with the cytoplasmic domain, transmembrane domain, or both. As a note, we used the processed C-terminal fragment at 26 kDa as a marker for the processing of soluble hADAM19 throughout this study.
The Processing at Glu 586 2 Ser 587 within the Cysteine-rich Domain Occurs by an Autolytic Mechanism-A recent report showed that ADAM13 shedding is dependent on its own metalloproteinase activity (48). We therefore hypothesized that the metalloproteinase activity of hADAM19 is also involved in its processing. To test this hypothesis, we used several independent approaches. As shown in Fig. 3A, rare processing was detected in the conditioned media from D52-transfected COS1 cells that were treated with decRVKR-CMK, which blocks the activation of hADAM19 (46). This indicates that prodomain removal is necessary for the processing at its cysteine-rich domain of hADAM19. Furthermore, 199 RA-D, an inactive mutant of soluble hADAM19 resistant to furin-mediated removal of its prodomain (46), displayed no processing, confirming that the soluble pro-form of hADAM19 lacks the capacity to process at its cysteine-rich domain (Fig. 3A). (The total protein level in the media was comparable to that in the media of the D52 cells.) We also generated another soluble inactive form of hADAM19 (E346A-D), in which the active residue Glu at 346 in the metalloproteinase domain was mutated to Ala, and transfected it into COS1 cells. Obviously, no processing at its cysteine-rich domain was observed in the media of these cells. Once again, the total protein level in the media was almost equal to that in the media of the D52 cells (Fig. 3A). These results strongly argue that the processing at its cysteine-rich domain of soluble hADAM19 depends on its own metalloproteinase activity. However, neither GM6001 nor TIMP-3, inhibitors that typically block sheddase activity (26 -30), inhibited the autolytic processing of hADAM19 at its cysteine-rich domain (data not shown).
To further confirm that autolytic processing at its cysteinerich domain occurs in soluble hADAM19, we generated stable MDCK cell lines called D52-5, 199 RA-D-6, and E346A-D-17, which expressed soluble D52, 199 RA, and E346A, respectively. As we expected, the 26-kDa-processed fragment was clearly detectable in the conditioned media from D52-5; the processing was dramatically inhibited by decRVKR-CMK (Fig. 3B). Furthermore, there was no processing at the cysteine-rich domain in 199 RA-D-6 and E346A-D-17 (Fig. 3B). (The total amount of soluble protein was comparable among the conditioned media of these cell lines). Once again, PMA, GM6001, and TIMP-3 failed to affect the processing of D52-5 (Figs. 2B and 3B, data not shown).
N-terminal sequencing revealed that the starting sequence of the purified 26-kDa protein was SNAVPIDT, which is identical to 596 SNAVPIDT 594 within the cysteine-rich domain of hADAM19. This suggests that the processing of hADAM19 occurs at Glu 586 2 Ser 587 within its cysteine-rich domain (Fig.  3B).
The Sizes of Glu 586 and Ser 587 Are Critical for the Processing at Glu 586 2 Ser 587 of Soluble hADAM19 -To examine the importance of the Glu 586 (P1) and Ser 587 (P1Ј) sites in the processing at Glu 586 2 Ser 587 of soluble hADAM19, we changed the size, charge, and polarity of these two residues by mutagenesis. The constructs are shown in Fig. 4. Shown in Fig.  5A, there was no or little detectable processing in the COS1 cells transfected with either E586D-D or S587T-D, suggesting that subtle changes in the sizes of residues at the P1 and P1Ј positions can dramatically impair the processing. Furthermore, rare or little processing was observed when significant changes were made to the side chains of the residues at these sites, as in the cases of the Glu 586 to Ala or Gly, and Ser 587 to Phe mutants (Fig. 5A). On the other hand, the amount of processing product in the E586Q-D or S587A-D mutant, in which the side chain of the amino acid residue was almost the same size as that in wild type soluble hADAM19, was almost equal (Fig. 5A).
To further confirm the fate of processing at Glu 586 2 Ser 587 upon changes at the P1 and P1Ј sites, we chose mutants with subtle changes and, presumably, structural similarities, Glu 586 to Asp and Ser 587 to Thr, to generate stable transfectants in MDCK cells. Shown in Fig. 5B, rare or little processing was detectable in the conditioned media from these stable MDCK transfectants, confirming that both soluble E586D-D and S587T-D have undetectable or little ability to process, even if they are cleaved by furin in the MDCK transfectants. (There were no significant differences in protein levels among the transfectants.) The Signals of CaM and Calcium May Be Involved in the Processing at Glu 586 2 Ser 587 of hADAM19 -In addition to the PKC pathway, there are several other signal pathways, involving tyrosine kinase, MAPKs, phosphatase, phosphatidylinositol 3-kinase, calcium, and CaM, which regulate the ectodomain-shedding process (10,(13)(14)(15)(16)(17)(18)(19)(21)(22)(23). Specific inhibitors were used to determine whether or not these signal pathways regulate the processing at Glu 586 2 Ser 587 of hADAM19. As shown in Fig. 6, only CaM inhibitors impaired the processing of soluble hADAM19. The other inhibitors, including genistein, PD98059, pervanadate, wortmannin, and LY290042, had no significant effects on the processing of soluble or full-length hADAM19 (data not shown). In addition, both A23187 and brefeldin A block the activation of both soluble and full-length hADAM19 (46). There was no processing at Glu 586 2 Ser 587 detectable under the treatment of either A23187 or brefeldin A (data not shown), indicating that hADAM19 is activated and processed in the secretory pathway. Taken together, our results suggest that PKC, CaM, and calcium signal pathways may be related to the processing at Glu 586 2 Ser 587 of hADAM19.
The Processing at Glu 586 2 Ser 587 Is Necessary for hADAM19 to Exert Its Proteolytic Activity against ␣2-M-To assess the significance of the processing at Glu 586 2 Ser 587 of hADAM19, we purified the proteins from D52-5, E586D-D, and S587T-D, respectively. Intriguingly, the processed N-terminal fragments, containing the metalloproteinase, disintegrin, and parts of the cysteine-rich domain, were detected as mature forms using anti-disintegrin antibody (data not shown). This suggests that the processed N-terminal segments bind with unprocessed soluble forms or processed C-terminal-soluble fragments by one or more disulfide bonds, consistent with the results obtained early from the full-length hADAM19 (data not shown). When we probed the purified proteins with anti-FLAG-M2, as shown in Fig. 7A, S587T-D displayed relatively more processing than E586D-D, in which a very low level of processing was detected. D52-5 showed much more processing than E586D-D and S587T-D. As shown in Fig. 7B, D52-5 protein had a much greater ability to form a complex with ␣2-M and generate two products compared with both E586D-D and S587T-D proteins, which displayed much lower activities. Obviously, S587T-D had a relatively higher activity than E586D-D, which showed very low activity. These results perfectly coincide with the Western blot shown in Fig. 7A, indicating that the more that hADAM19 processed at Glu 586 2 Ser 587 , the more proteolytic activity it exerted against ␣2-M in vitro.
Ac-RPLE-SNAV Is Cleaved by Active hADAM19 -To test hADAM19 activity, we synthesized a new peptide, Ac-RPLE-SNAV, which encompasses the processing site of hADAM19 and is conserved among humans and mice (42,45). Because the processing at Glu 586 2 Ser 587 is mediated by its own metalloproteinase activity (Fig. 3), we were able to use this peptide to assay soluble hADAM19 activity and obtained the similar results as from the ␣2-M assay (Fig. 7B). Once again, the highest activity was seen in D52-5 with 1% cleavage after incubation overnight. E586D-D and S587T-D only showed 30 and 10% the activity exerted by D52-5, respectively (data not shown). As we knew, the activity at 1% was not enough for a fluorescence assay. However, given that D52-5 was not fully processed (maybe 25% of the mature hADAM19 got processed according to the Western blotting result as shown in Fig. 7A), we decided to generate another stable line in MDCK cells expressing D-CR (Fig. 4). The reason we deleted the whole cysteine-rich domain is that we surmised that this domain has a potential interaction with the metalloproteinase domain, disintegrin domain, or both that might decrease the activity. Fortunately, purified D-CR from the MDCK transfectants had a higher activity for cleaving our new peptide substrate, and the cleavage product detected by N-terminal sequencing was peptide SNAV, confirming that the peptide substrate was cleaved at the E-S bond, which is identical to the processing site in the hADAM19 protein. Furthermore, as shown in Fig. 8, the k cat , K m , and k cat /K m values for 150-min incubation were 2.4 min Ϫ1 , 2.0 mM, and 1200 M Ϫ1 min Ϫ1 , respectively. Thus, we developed a peptide substrate for determining soluble hADAM19 activity by a fluorescamine assay.

DISCUSSION
In the current report, we have demonstrated that processing of hADAM19 occurs at Glu 586 2 Ser 587 within the cysteine-rich domain by its own metalloproteinase activity and is a necessary step to display its proteolytic activity against both a peptide substrate and ␣2-M in vitro. We have also revealed that the processing at Glu 586 2 Ser 587 of hADAM19 is regulated by a unique pathway, distinguishable from those shown for other ADAMs, MT-MMPs, and other membrane-bound proteins.
Shedding or Processing of Metalloproteinases-Growing evidence suggests that shedding is of vital importance for the regulation of metalloproteinase activity. For MT-MMPs, Pei's group reported that MT5-MMP is shed by furin, down-regulating its activity, and that interleukin-8 triggers the signal for both release and activation of MT6-MMP by an unknown mechanism (47,52). The activity of MT1-MMP can be autolytically terminated directly on the cell surface or via production of a soluble functional fragment, consequently down-regulating enzyme activity on the cell surface (40). Among ADAMs, ADAM13 is the only one that has been shown to shed its ectodomain intracellularly by an autolytic mechanism, producing an active enzyme able to bind with ␣2-M and integrins (48). In addition, truncation of mature ADAMTS4 at its C terminus is required for its aggrecanase activity (49). In this report, we demonstrate that hADAM19 carries out processes within its cysteine-rich domain, resulting in an active enzyme shown by both ␣2-M and peptide substrate assays in vitro (Figs. 3, 7, and 8). It might, therefore, be a general regulatory mechanism that MT-MMPs, such as MT1-MMP and MT5-MMP, are down-regulated by shedding to release active forms from the cell surface, whereas ADAMs must shed, carry out process, or become truncated at the C terminus to exert their functions, such as acting as a sheddase, binding with integrins on the cell surface, or digesting components of the extracellular matrix.
Regulation of the Processing at Glu 586 2 Ser 587 of hADAM19 -The signal pathways involved in shedding or truncation processes, especially of ADAMs, are poorly understood. In the present report, we provide unique characteristics of the regulation of the processing at Glu 586 2 Ser 587 of hADAM19. We found that PMA, a common inducer of shedding, also enhances the processing at Glu 586 2 Ser 587 of hADAM19 ( 2A). The mechanism probably involves the cytoplasmic domain, transmembrane domain, or both, because PMA did not alter the processing of soluble hADAM19 (Fig. 2B). This is consistent with reports showing that the cytoplasmic domain of ADAM9 is required for PMA-induced shedding (34) and that the membrane anchor of TACE is necessary for its processing of TNF-␣ (53,54). In addition, there are a few reports that have demonstrated the requirement of the cytoplasmic tail for shedding of pro-NRG, APP, pro-TGF, and L1 adhesion molecules (17,55,56). However, in most cases, endogenous and/or inducer-mediated shedding is independent of the cytoplasmic domain ( Fig.  2B) (14,17,19,22,24). Calcium ionophore, A23187, is another potent inducer of most protein-shedding processes (13)(14)(15)(16). Nevertheless, we found that A23187 and brefeldin A block the activation of both full-length and soluble hADAM19. Subsequently, no processing at Glu 586 2 Ser 587 was detectable, suggesting that hADAM19 is activated and processed in the secretory pathway (46, data not shown). Inhibitors of CaM have been shown to stimulate the shedding of several proteins, including MT1-MMP, pro-TGF␣, pro-NRG, and APP, by a mechanism independent of both PKC and calcium (17)(18)(19)(20). However, in our study, a reverse result was obtained; the processing at Glu 586 2 Ser 587 was inhibited by the CaM inhibitors (Fig. 6). Finally, we found that tyrosine kinase, MAPK, phosphatase, or phosphatidylinositol 3-kinase do not seem to play roles in the processing at Glu 586 2 Ser 587 of hADAM19 (data not shown), although they have been previously shown to participate in some shedding processes (9 -12, 21-23, 29).
In addition, the proteinases responsible for the shedding of many cell surface molecules seem to have broad sequence specificity as revealed by mutational analysis of residues around the cleavage site of pro-TGF␣, APP, IL-6 receptor, L-selectin, and pro-TNF␣ (57)(58)(59)(60)(61). In this report, we used mutagenesis to show that the residue sizes of the side chains at both the Glu 586 and Ser 587 sites are extremely important for normal processing at Glu 586 2 Ser 587 of hADAM19. Even delicate changes, such as Glu 586 3 Asp and Ser 587 3 Thr, caused dramatic decreases in the processing. Notably, many studies have shown that some potent synthetic inhibitors of metalloproteinases, such as TAPI, BB94, and GM6001, can block most, if not all, shedding processes and that many shedding processes are sensitive to TIMP-3, a matrix-associated TIMP that preferably inhibits ADAMs (5-8, 19, 24 -30). However, neither GM6001 nor TIMP-3 inhibits the autolytic processing at Glu 586 2 Ser 587 of hADAM19 (Fig. 3B), which is consistent with some reports showing that the shedding of MT5-MMP, MT6-MMP, and IL-6 receptor is not affected by metalloproteinase inhibitors (47,52,62). One possibility, we speculate, is that shedding, in most cases, occurs at membrane-proximal regions on the cell surface, which are easily accessible to hydroxamate-based inhibitors and TIMP-3 (5,25,60,63). Human ADAM19 processing takes place at a region distal from the transmembrane domain in the secretory pathway, which is less accessible to GM6001 and TIMP-3 (Fig. 3B). How hADAM19 initiates the processing at Glu 586 2 Ser 587 and what the roles for the disintegrin-and cysteine-rich domains are during the processing at Glu 586 2 Ser 587 remain to be uncovered. Perhaps dimerization through the disintegrin-and/or cysteine-rich domains is the key step for the processing as proposed for other ADAMs (3). This might also be an explanation for the fact that the processed N-terminal fragments, containing the metalloproteinase, disintegrin, and parts of the cysteine-rich domain, were not detected in the conditioned medium from the transfected COS1 cells with full-  length hADAM19 and were present among the purified soluble hADAM19 proteins (data not shown).
A New Peptide Substrate Based on the Processing Site Sequence for hADAM19 -The peptide substrates currently used to measure MMP activity are synthesized based on the cleavage sites of protein substrates. Because hADAM19 processes at Glu 586 2 Ser 587 by its own metalloproteinase activity, we surmised that a peptide encompassing the processing site would be an ideal substrate. Indeed, this peptide, Ac-RPLE-SNAV, was suitable for a fluorescamine assay of enzyme activity and was cleaved at the E-S site as determined by peptide N-terminal sequencing. Although the wild type hADAM19 showed a minimally detectable activity by this method, D-CR, containing the metalloproteinase and disintegrin domains, displayed a higher activity to determine kinetic parameters after incubation for 150 min using our fluorescamine assay (Fig. 8). The kinetic parameters, k cat , K m , and k cat /K m , were 2.4 min Ϫ1 , 2.0 mM, and 1200 M Ϫ1 min Ϫ1 , respectively, demonstrating that this is a poor substrate and will have a limited usage. It is necessary that the peptide substrates will be optimized by creating analogs that are hydrolyzed more efficiently by hADAM19 in the near future. Our preliminary data show that TIMP-3 can inhibit hADAM19 activity in this peptide-based assay, 2 which is similar to the results for ADAM10, -12, and -17 and ADAMTS4 and -5 in vitro (64 -67), and confirm that the TIMP-3 we used was active. However, TIMP-3 had no effect on the cellular processing at Glu 586 2 Ser 587 of soluble hADAM19 (Fig. 3B), indicating that this processing occurred intracellularly, to where TIMP-3 molecules might not be accessible. Interestingly, ADAMTS4 is truncated at Glu 373 2 Ala 374 , the site for the cleavage of ADAMTS4 and ADAMTS5, not MMPs. In contrast, the truncation site at Asn 341 2 Phe 342 is mediated by MMPs, not aggrecanases (49,68). Our mutational data showed that the Glu 586 2 Ala 587 processing site was also optimal for autolytic processing of hADAM19 (Fig. 5B), indicating that our peptide substrate might be useful to determine the activity of other ADAMs, such as aggrecanases. This peptide substrate may also be used for testing metalloproteinase inhibitors against hADAM19 and other ADAMs.