Intracellular activation of human adamalysin 19/disintegrin and metalloproteinase 19 by furin occurs via one of the two consecutive recognition sites.

Adamalysin 19 (a disintegrin and metalloproteinase 19, ADAM19, or meltrin beta) is a plasma membrane metalloproteinase. Human ADAM19 zymogen contains two potential furin recognition sites (RX(K/R)R), (196)KRPR(200)R and (199)RRMK(203)R, between its pro- and catalytic domains. Protein N-terminal sequencing revealed that the cellular mature forms of hADAM19 started at (204)EDLNSMK, demonstrating that the preferred furin cleavage site was the (200)RMK(203)R downward arrow(204)EDLN. Those mature forms were catalytically active. Both Pittsburgh mutant of alpha(1)-proteinase inhibitor and dec-Arg-Val-Lys-Arg-chloromethyl ketone, two specific furin inhibitors, blocked the activation of hADAM19. Activation of hADAM19 was also blocked by brefeldin A, which inhibits protein trafficking from the endoplasmic reticulum to the Golgi, or, a calcium ionophore known to inhibit the autoactivation of furin. When (202)KR were mutated to AA, the proenzyme was also activated, suggesting that (197)RPRR is an alternative activation site. Furthermore, only pro-forms of hADAM19 were detected in the (199)RR to AA mutant, which abolished both furin recognition sites. Moreover, the zymogens were not converted into their active forms in two furin-deficient mammalian cell lines; co-expression of hADAM19 and furin in these two cell lines restored zymogen activation. Finally, co-localization between furin and hADAM19 was identified in the endoplasmic reticulum-Golgi complex and/or the trans-Golgi network. This report is the first thorough investigation of the intracellular activation of adamalysin 19, demonstrating that furin activated pro-hADAM19 in the secretory pathway via one of the two consecutive furin recognition sites.

The proprotein convertases (PCs) are a large family of serine proteinases that recognize dibasic or RX(K/R)R motifs and cleave the peptide bond on the carboxyl side (30 -32). As a major proprotein convertase, furin is concentrated in the trans-Golgi network (TGN) and cycles between this compartment and the cell surface through the endocytic pathway. The autoactivation and intracellular trafficking of furin are well characterized. Numerous studies have shown that furin activates a large number of proproteins in multiple compartments (30 -32). For instance, furin has been demonstrated to mediate the activation of proenzymes, such as ␤-amyloid-converting enzyme (BACE), some matrix metalloproteinases (MMPs), including MMP-11, -14, -16, and -24, and some ADAMs, including ADAM1, -9, -12, -15, and -17 and ADAMTS1, -4, and -12 (16 -23, 30, 31, 33-40). However, the molecular mechanism and pathway by which the cells regulate the potentially important interactions between these proenzymes and the proprotein convertase in cells are not fully understood. In this report, we present evidence that furin is responsible for the activation of hADAM19 and this activation can occur via one of the two consecutive recognition sites and that furin is co-localized with the substrate in the ER-Golgi complex and/or TGN.
DNA Transfection and Generation of Stable hADAM19 Expression Cell Lines-LipofectAMINE 2000-mediated DNA transfections into MDCK cells were performed following the instructions provided by Invitrogen. Stable lines were selected in the presence of G418 (400 g/ml) and screened by Western blotting as described (40 -42).
Western Blotting-The experiments were carried out as described previously (40,41). Briefly, cells were grown to 80% confluence and were treated as indicated. After centrifugation at 14,000 ϫ g for 15 min at 4°C to clear any debris, the serum-free media were prepared for SDS-PAGE. The cells were lysed with RIPA (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 in ice. The supernatant was collected after centrifugation at 14,000 ϫ g for 20 min at 4°C. After electrophoresis, the proteins were transferred onto nitrocellulose membranes and probed with anti-FLAG-M2 or anti-hADAM19 and developed as described (40,41).
Purification of Soluble hADAM19 and Protein N-terminal Sequencing-All proteins were purified on anti-FLAG-M2 affinity columns as described (42,43). Briefly, cells stably expressing wild type soluble hADAM19 (D52-5) or 199 RR to AA mutant ( 199 RA-D-6) were grown to 100% confluence, then washed with PBS twice and incubated for 48 h in serum-free medium containing GM6001 (to prevent the degradation of hADAM19). The conditioned media were collected, centrifuged to clear debris, and loaded onto an anti-M2 immuno-affinity column (1 ml of resuspended agarose) prewashed with Tris-buffered saline. The bound materials were extensively washed with Tris-buffered saline, eluted with FLAG peptides, and collected in 200-l fractions. The fractions were analyzed by Western blot using anti-hADAM19 antibodies or anti-FLAG-M2. The fraction containing the highest hADAM19 protein concentration was prepared for protein N-terminal sequencing. After separation by SDS-PAGE, the samples were transferred to a polyvinylidene difluoride membrane and stained with Coomassie Blue R-250. After destaining, the hADAM19 bands were excised and sent to the Bioanalytical Core Facility at the Florida State University for N-terminal amino acid sequencing.
␣ 2 -M Trapping Assay to Determine Endopeptidase Activity of hADAM19 Species-The detailed experimental procedure was previously reported (27,28). Briefly, 10 l of the fraction containing purified soluble hADAM19 was mixed with 24 l of ␣ 2 -M (0.2 unit/ml), adjusted to a total volume of 100 l by adding HEPES buffer (50 mM HEPES, pH 7.5, 200 mM NaCl, 10 mM CaCl 2 , 25 M ZnCl 2 , 0.05% Brij- 35), and incubated at 37°C for 1-5 days. A 20-l aliquot of the mixture was removed daily, put into 2ϫ SDS-PAGE sample buffer, and boiled. Following SDS-PAGE, the protein bands in the gels were visualized by silver staining.
Transient Transfection into COS1, RPE.40, or 7.P15 Cells-COS1, RPE.40 or 7.P15 cells were seeded in 24-or 6-well plates for 16 -24 h at 80% confluence prior to transfection. The cells were then transfected with the indicated plasmids using LipofectAMINE 2000. After 6 -10 h, serum-free or 5% FBS DMEM containing 2.5 M GM6001, with or without CMK, BFA, or A23187 at indicated concentrations, were added for another 24 h. The conditioned media and cell lysates were analyzed by Western blotting. For co-transfection experiments in these cells, the indicated plasmid was transfected alone as a control and then cotransfected with expression plasmids directing the production of furin, PACE4, pAT, pATp, furin and pAT, or furin and pATp. After 6 -10 h, serum-free or 5% FBS DMEM containing GM6001 (2.5 M) was added to the transfected cells for 16 -24 h. Then, the conditioned media and cell lysates were analyzed by Western blotting as described above.
Glycosylation Analysis-N-Glycosylation in vitro was investigated by endoglycosidase F treatment as described previously (40,41). Briefly, transfected cells were grown to 80% confluence and incubated in serumfree medium for 24 h. The conditioned media were then collected, and the cells were lysed with RIPA. After centrifugation, the conditioned media or the supernatant from RIPA were treated with glycosidase F (5 units, Roche Molecular Biochemicals) for 20 h at 37°C and analyzed by Western blotting.
Confocal Microscopy-The procedures have been described in detail previously (38,40). Briefly, MDCK cells expressing hADAM19 wild type or 199 RA mutant were grown on coverslips in six-well plates with or without treatment with CMK, BFA, or A23187. After fixing with Lina's fixation buffer for 30 min, the cells were permeabilized with buffer A (0.3% Triton X-100, 1% neutral detergent solution, 1% bovine serum albumin, and 0.01% NaN 3 in PBS) for 1 h and incubated for 3 h with anti-furin and anti-FLAG-M2 (1:100 dilution in buffer A) for double staining. After washing with PBS three times, secondary antibodies conjugated with either fluorescein isothiocyanate or rhodamine red were added to the cells for 1 h, followed by four washes with PBS. Confocal microscopy experiments were performed at the Biological Science Imaging Resource Facility at Florida State University.

Removal of the hADAM19 Prodomain Is Dependent on Furin
Activity-The sequence of hADAM19 contains two potential furin recognition sites (RX(K/R)R), 196 KRPRRMK 203 R, between its pro-and catalytic domains (Fig. 1A). To ascertain the role of furin in the cleaving of the hADAM19 prodomain, wild type hADAM19 (F46) with a C-terminal FLAG tag was transfected into COS1 cells alone or co-transfected with furin, PACE4, ␣ 1 -proteinase inhibitor (pAT), both furin and pAT, Pittsburgh mutant of ␣ 1 -proteinase inhibitor (pATp, a specific inhibitor of furin) (37,45), or both furin and pATp. As shown in Fig. 1B, active hADAM19 forms were increased by the introduction of furin, but not PACE4. Furthermore, pATp blocked the processing of hADAM19; furin could not restore this processing when cells were co-transfected with furin and pATp (lane 8). pAT had little effect on endogenous or furin-induced processing of hADAM19 (lanes 2, 3, 5, and 6). Because pATp did not completely block the hADAM19 processing, the low levels of endogenous processing may also be meditated by other proprotein convertases in addition to furin (lanes 7 and 8).
Interestingly, both the pro-and active forms of hADAM19 were doublets. These doublets may be differentially glycosylated forms. According to protein sequence analyses, hADAM19 has five potential glycosylation sites (27). Indeed, endoglycosidase F converted the doublets into a single pro-or active form, respectively (data not shown). To verify that the active hADAM19 lacks a prodomain, hADAM19 antibodies against the pro-, catalytic, or disintegrin domains (28) were used to probe the proteins in the cell lysates. The results in Fig. 1C clearly showed that the processed 80-kDa hADAM19 came from the removal of its prodomain because it was not recognized by the antibody against the prodomain peptide; however, it was detected with antibodies against its catalytic and disintegrin domains, respectively. These data showed that furin activity played a major role for the intracellular removal of hADAM19 prodomain.
To further demonstrate a direct role for furin in the activation of the hADAM19 zymogen, COS1 cells transfected with wild type hADAM19 (F46) were incubated with dec-Arg-Val-Lys-Arg-CMK (decRVKR-CMK), a widely used inhibitor of furin (17,38,40,46,47). As shown in Fig. 2A, decRVKR-CMK blocked the activation of hADAM19 in a dose-dependent manner. Because furin is mainly localized in TGN and the autoactivation of furin is calcium-dependent (48), we investigated whether hADAM19 activation occurred in the trans-Golgi network and required calcium. COS1 cells transfected with wild type hADAM19 were treated with BFA, which blocks protein trafficking from the ER to the Golgi apparatus (48,49), or A23187, a calcium ionophore known to inhibit the maturation of furin (48). As shown in Fig. 2B, only the pro-forms of hADAM19 were detected upon treatment with either BFA or A23187. These results are consistent with furin-mediated activation of hADAM19.
Deletion of the Transmembrane Domain and the Cytoplasmic Tail of hADAM19 Does Not Alter the Processing of the Prodomain by Furin-To isolate soluble hADAM19 protein for enzyme activity assays, a construct encoding the extracellular domain (ectodomain) of hADAM19 containing a C-terminal FLAG tag was generated; this construct was called D52 and lacked the transmembrane domain and cytoplasmic domain  lanes 1 and 2), anti-catalytic domain (Cat) (lanes 3 and 4), or antidisintegrin domain (Dis) (lanes 5 and 6).
( Fig. 4A). This hADAM19 ectodomain construct was transfected into COS1 cells. As shown in Fig. 3A, only the active forms of soluble hADAM19 were detected in the media from COS1 cells co-expressing the hADAM19 ectodomain and furin. Both pro-and active forms were detected in the media from the cells co-transfected with PACE4 or transfected with the ectodomain construct alone. Additionally, the active forms were detected in the cell lysates only when cells were co-transfected with furin (Fig. 3A, lanes 5-8).
A dose-dependent inhibition of soluble hADAM19 activation by decRVKR-CMK was observed (Fig. 3B). However, there was no significant effect of decRVKR-CMK on the intracellular levels of hADAM19. Furthermore, there was no secretion of soluble hADAM19 in the transfected cells treated with either BFA or A23187 (Fig. 3C). Also, pATp dramatically decreased the amount of active forms in the medium when it was expressed in COS1 cells, but pAT failed to do so (Fig. 3C). Once again, no significant differences were seen in response to these treatments in the cell lysates (Fig. 3C). These results show that the soluble forms of hADAM19 were processed in the same manner as the full-length form and are consistent with furinmediated activation of hADAM19.
There Are Two Alternative Furin Recognition Sites between the Pro-and Catalytic Domain of hADAM19 -Upon the examination of the hADAM19 protein sequence, two consecutive furin recognition sites (RX(K/R)R), 196 KRPR 200 R and 199 RRMK 203 R, were found (Fig. 1A). We hypothesized that these two furin recognition sites are alternatively used for the intracellular activation of pro-ADAM19 by furin. To test this hypothesis, three mutants were generated in full-length and ectodomain hADAM19, which converted the 196 KR, 199 RR, and 202 KR into AA, respectively. These were named as 196 RA-F, 196  All of the plasmids were transfected into COS1 cells to compare the levels of activated hADAM19 for the wild type and RA mutants in both the full-length (RA-F) and ectodomain forms (RA-D). As shown in Fig. 4 (B and C), no active forms of 199 RA mutants were detected as a result of the absence of a furin cleavage motif, whereas almost equivalent amounts of the active forms of the 196 RA and 202 RA mutants were detected. The wild type, full-length hADAM19 (Fig. 4B) and ectodomain form (Fig. 4C) were studied in parallel with the mutants. In addition, the protein levels of hADAM19 were almost equal among the cell lysates from these transfectants (Fig. 4, B and C). These results strongly supported the hypothesis that processing of the prodomain of hADAM19 was dependent on the presence of either one of the two consecutive furin recognition sites between the pro-and catalytic domains.
To further confirm that furin processed hADAM19 via one of the two alternative furin recognition sites, samples of the media from the four cell lines in Fig. 4C were incubated with the medium containing soluble furin, which was obtained from the  6 and 12). The next day, cells were treated with either 10 g/ml BFA (lanes 3 and 9) or 0.5 M A23187 (lanes 4 and 10) in serum-free medium for 24 h. The conditioned media (lanes 1-6) or cell lysates (lanes 7-12) were analyzed as in A.
furin-transfected COS1 cell culture. The medium from COS1 cells transfected with a blank vector was used as a negative control. As shown in Fig. 4D, soluble furin did not process the 199 RA mutant of the soluble hADAM19. However, the wild type and the mutant soluble proteins containing a furin recognition motif were cleaved by furin. This furin-mediated processing was sensitive to decRVKR-CMK inhibition, consistent with the intracellular processing results obtained earlier (Figs. 3A and  4C). These results demonstrated that furin could activate hADAM19 at both furin cleavage sites between the prodomain and the catalytic domain of the zymogen.
Removal of the Prodomain Was Required for hADAM19 to Exert Its Proteolytic Activity-In our previous reports, an in vitro assay was established using ␣ 2 -M to test the activity of hADAM19 (27,28). To assess the importance of zymogen activation to the proteolytic activity of hADAM19, stable lines of wild type hADAM19 and its 199 RA mutants were generated in MDCK cells, in which the endogenous furin activity is high (38,40,50). One stable line was chosen from each group as a representative to be treated with CMK, BFA, or A23187 and to examine whether hADAM19 would display the same process-ing as it did in COS1 cells. As predicted, CMK, BFA, or A23187 blocked the activation of wild type hADAM19 in the stably transfected MDCK cells called F46-4 (Fig. 5A). MDCK cells stably expressing the full-length 199 RA ( 199 RA-F-9) showed no conversion of the pro-hADAM19 to its active form (Fig. 5A). Furthermore, as shown in Fig. 5B, the active forms were only detected in the medium from MDCK cells stably expressing soluble hADAM19 (D52-5). When D52-5 cells were treated with decRVKR-CMK for 24 h, the pro-forms of soluble hADAM19 were predominantly detected from the cell culture medium. There were no active forms detected in the medium from the 199 RA-D6 mutant cells, which were MDCK cells stably expressing soluble hADAM19 with the 199 RA mutation (Fig. 5B).
Soluble hADAM19 proteins were purified from conditioned media of D52-5 and 199 RA-D6 cells. The endopeptidase activity of the purified metalloproteinases was tested using an ␣ 2 -M trapping and cleaving assay. As shown in Fig. 5C, only the wild type proteins could complex with ␣ 2 -M and generate two cleaved products. This activity was completely blocked by EDTA. The 199 RA mutant proteins were inactive, likely because the prodomain containing the cysteine switch residue  4 and 9), or 202 RA-D (lanes 5 and 10) overnight, followed by incubation in serum-free medium for 24 h. The samples were analyzed by Western blotting with anti-FLAG-M2. D, activation of the soluble hADAM19 by exogenous soluble furin. The condition media from lanes 2-4 in C were mixed with equal volumes of conditioned serum-free media from COS1 cells transfected with the blank vector (lanes 1, 4, 7, and 10) or soluble furin (lanes 2, 3, 5, 6, 8, 9, 11, and 12). After being incubated with (lanes 3, 6, 9, and 12) or without 50 M CMK (lanes 1, 2, 4, 5, 7, 8, 10, and 11) at 37°C for 24 h, the samples were analyzed by Western blotting as in C.
was not removed by furin (Fig. 5B). These results suggest that hADAM19 activation also obeys the cysteine-switch mechanism for zymogen latency and activation. Furthermore, Nterminal sequences of the purified hADAM19 proteins from the media revealed that the processed doublets (Fig. 5B, lane 2) had the identical N-terminal sequences of 204 EDLNSMK, suggesting that furin prefers to cleave hADAM19 using the recognition site of 200 RMK 203 R rather than 197 RPR 200 R. The doublets have different glycosylation patterns as verified by endoglycosidase F treatment experiments (data not shown). These results confirmed the prediction that hADAM19 was activated by furin through cleavage of the 203 R-204 E peptide bond at the sequence 199 RRMKR2 204 EDLNSMK.
Two Furin-deficient Cell Strains (RPE.40 and 7.P15) Do Not Activate Pro-hADAM19, and the Introduction of Furin into These Cells Restores Zymogen Activation-To further confirm that furin activity was required for the intracellular activation of pro-hADAM19, the wild type of full-length and soluble forms of hADAM19 (F46 and D52) were transfected into RPE.40 cells and 7.P15 cells, two furin-deficient cell strains (44,51). Processing of hADAM19 to its mature forms was negligible in these cell lines. However, the active forms were clearly detectable when cells co-expressed furin (data not shown). As shown in Fig. 6, there were barely detectable levels of the active forms of soluble hADAM19 in the media of the two D52-transfected cell lines. High levels of active forms were only detected in the media when the cells co-expressed furin, although PACE4 also increased the amount of active forms when it was co-expressed with D52 (Fig. 6, A and B). Curiously, the active forms were only detected in the lysates of cells co-expressing D52 and furin. These data further confirm that furin was responsible for the intracellular activation of hADAM19.
Furin Was Co-localized with hADAM19 in the ER-Golgi Complex and/or TGN-To verify that hADAM19 was a physiologically relevant substrate of furin, the cellular localization of furin and hADAM19 was examined by confocal microscopy using MDCK cells stably expressing hADAM19 (Fig. 7). Untreated cells are marked as F46-4 (control) (Fig. 7, top panels). Co-localization of hADAM19 and furin was clearly observed, and was consistent with ER-Golgi complex and/or TGN localizations (top, middle panel). Furthermore, hADAM19 was also seen at the edges of the plasma membrane (right lane, top panel) where furin was rare (left lane, top panel). Interestingly, a similar pattern of co-localization between the 199 RA mutant and furin was also observed (Fig. 7, bottom panels), suggesting that the co-localization was independent of the recognition sites for furin in hADAM19. To test whether the agents that block the activation of hADAM19 could prevent co-localization between furin and hADAM19 (Fig. 5A), the cells were treated with CMK, BFA, or A23187. None of these treatments interfered with the co-localization pattern of furin and hADAM19 (Fig. 7), suggesting that furin may be co-localized with hADAM19 in perinuclear ER-Golgi complex and/or TGN independent of the furin catalytic activity. DISCUSSION Proteolysis of the extracellular matrix and cell surface proteins mediated by metalloproteases, including MMPs and ADAMs, is of vital importance for tissue-remodeling processes during normal and pathological conditions, such as tissue mor-phogenesis, wound healing, inflammation, and tumor cell invasion and metastasis (3-7, 52, 53). Metalloproteases are synthesized as inactive proenzymes or zymogens, and their latency is maintained by a cysteine-switch residue in the propeptide domain in which the thiol group is coordinated to the active site zinc (II) (2, 9 -12). To display any proteolytic activities, the prodomain located N-terminal to the catalytic domain must be removed from the zymogen in most cases. Recently, PCs, such as furin and or furin-like serine peptidases, have been recognized as very important enzymes for the zymogen activation, although various mechanisms have been proposed for the activation of pro-MMPs and pro-ADAMs. Furin or furin-like PCs mediate zymogen activation by recognizing a conserved RX(K/ R)R motif in the boundary between pro-and catalytic domains. This motif is present in almost all ADAMs and nine MMPs (2,13,53). By analyzing the intracellular activation of hADAM19, we have demonstrated that both furin activity and one of the two consecutive sites in 197 RPRRMK 203 R in ADAM19 are required for activation, which is dependent on calcium and proper secretory pathway trafficking. Furthermore, we have provided direct evidence that furin is co-localized with hADAM19 in ER-Golgi complex and/or TGN. This colocalization between furin and hADAM19 is independent of the furin recognition site and is resistant to a variety of treatments, such as CMK, BFA, and A23187, that inhibit furin activity, vesicular trafficking, and calcium signal, respectively. These findings are consistent with the report published recently showing that furin was co-localized with MMP16 independent of their apparent enzyme-substrate relationship (40).
Latency and Activation of ADAMs-The classic cysteine switch mechanism for pro-MMP latency and activation was originally proposed for MMPs (10) and may be applied for many MMPs discovered with the exception of MMP-3, MMP-23, and MMP-26 (41, 54 -58). The activation of pro-MMP-3 by a mercurial compound was triggered by a perturbation of the conformation of the precursor rather than a direct disruption of the Cys-zinc interaction (54). A salt bridge in pro-MMP-3 might also contribute to the latency of the proenzyme (55). Organomercurial treatment failed to activate pro-MMP-26 with a unique cysteine-switch motif, PH 81 CGXXD, and when the conserved cysteine-switch sequence, PR 81 CGXXD, in the prodomain of pro-MMP-26 was restored by mutagenesis, the cysteine-switch activation mechanism was not induced (58).
Regarding the ADAM family members, the active ADAMs, such as ADAM1, -9, -10, -12, -15, -17, -19, -28, and ADAMTS1, -4, and -12, contain a catalytic site consensus sequence (HEXXH) in their metalloprotease domains (2, 11-14, 16 -21, 27-29, 53, 59 -61). They may also have a putative cysteineswitch residue in their prodomain to keep them inactive (9). For example, the investigation by Leochel et al. (11) demonstrated that the latency and activation mechanism of ADAM12 was similar to the cysteine switch model proposed for MMPs. ADAM9, -15, and -17 showed catalytic activity against their substrates only after their prodomains were removed (12,21,22). However, for many ADAMs, including ADAM19, no direct evidence has been provided to support the hypothesis that the Cys-zinc coordination is required for latency. For ADAM17/ tumor necrosis factor-␣ convertase, the prodomain was not only an inhibitor of the catalytic domain, but also appeared to act like a chaperone, facilitating secretion, folding, or both of the ADAM protein (12). In this report, we have demonstrated that, after the removal of the prodomain of hADAM19 by furin, the enzyme has endopeptidase activity against ␣ 2 -M. However, furin is unable to cleave the prodomain of the 199 RR to AA hADAM19 mutant lacking a furin recognition site in the boundary of the pro-and catalytic domains. This mutant has no FIG. 7. Co-localization of hADAM19 and furin. MDCK cells stably expressing wild type hADAM19 (F46-4) or 199 RA mutant ( 199 RA-F-9) grown on coverslips in six-well plates were treated with nothing (control), 100 M CMK, 10 g/ml BFA, or 0.5 M A23187 for 24 h. The fixed slides were stained for both hADAM19 with anti-FLAG-M2 (right panels) and furin with anti-furin (left panels). The merged pictures for both hADAM19 and furin are presented in the middle column. Note that hADAM19 is co-localized with furin, independent on furin motif, with neither CMK, BFA, nor A23187 altering the staining pattern. proteolytic activity using an ␣ 2 -M trapping assay. These results demonstrate that at least one of the furin recognition sites is required for the removal of the propeptide domain by furin to activate pro-hADAM19. The detailed mechanism of pro-ADAM19 latency and activation and the role of the cysteineswitch sequence remain to be further investigated.
For the activation of ADAM zymogens, two mechanisms have been reported. One is the removal of the prodomain by autolysis, but it was shown only in ADAM28 (60). The predominant mechanism for the activation of ADAMs is mediated by furin or furin-like PCs in the secretory pathway. This mechanism has been shown in many ADAMs, including ADAM1, -9, -12, -15, -17, and -19, and ADAMTS1, -4, and -12, using N-terminal sequencing, specific inhibitors of furin, blockers of protein trafficking from ER to Golgi, exogenous soluble furin in vitro, furin-deficient cell lines, and mutagenesis at the furin recognition site(s) (RX(K/R)R) between the pro-and catalytic domain (Refs. 16 -23; this report). In the present report, we provide a thorough investigation of ADAM zymogen activation mediated by furin (Figs. 1-6) and evidence that furin is co-localized with ADAMs in the ER-Golgi complex and/or TGN (Fig. 7), showing that ADAMs are similar to MMPs in these respects (38,40).
There Are Two Consecutive Furin Recognition Sites in the Boundary of the Pro-and Catalytic Domains of hADAM19 -The minimal furin recognition sequence requires basic residues at P 1 and P 4 (RXXR) and in some cases, at the P 1 position, an amino acid with a hydrophobic aliphatic side chain is not suitable (31). Typically, there is only one furin recognition site between the pro-and catalytic domain of the substrates of furin as found in most members of the ADAM family, seven MMPs, pro-BACE, and Notch1 receptor (2, 13-15, 30, 31, 33-40, 53, 62). In this report, we present evidence for the first time that there are two consecutive furin recognition sites, 197 RPR 200 R and 200 RMK 203 R, between the pro-and catalytic domain in hADAM19, which adhere to the rules for efficient cleavage by furin (31). Only pro-forms were detectable in the 199 RA mutant, which lacked a furin recognition site between its pro-and catalytic domain, whereas the mutants of both 196 RA and 202 RA, which possessed recognition sites, were converted into the active forms. Thus, the Arg residue at the P 4 site is required for the intracellular hADAM19 maturation mediated by furin (Figs. 4 and 5B). Interestingly, N-terminal sequencing of wild type mature forms (Fig. 5) confirmed that the preferred intracellular cleavage site for hADAM19 activation is the one nearer to the catalytic domain, 200 RMK 203 R, as predicted before (26,27). This motif is conserved in mice as 201 RMK 204 R (24). The distal motif, 197 RPR 200 R in humans, however, is replaced with 198 QPR 201 R in mice, which is not efficiently cleaved by furin.
A notion that pro-hADAM19 activation by furin may be sequential, i.e. 200 R 201 M is cleaved first followed by 203 R 204 E, seems to be consistent with the partially activated soluble species seen for 196 RA-D compared with 202 RA-D data in Fig.  4C; however, it does not agree with the data shown in Fig. 4B, where it is seen that the presence of furin with the full-length 196 RA-F leads to more activated species than 202 RA-F. The delicate changes in the interactions between furin and the different mutants that have subtle structural and conformational differences might be partially responsible for the different activation levels observed. Moreover, among all the protein N-terminal sequence data of wild type hADAM19 activated species, only 204 EDLNSMK was found; the alternative cleavage site product of 201 MKRED was not detected. Most importantly, the minimal furin recognition sequence requires basic residues at P 1 and P 4 (RXXR) and the Arg residue at the P 4 site is required for the intracellular hADAM19 maturation mediated by furin (Figs. 4 and 5B). It may not be possible for furin, an endopeptidase, to effectively cleave the product of the 200 R-201 M cleavage because the 201 MK 203 R-204 ED sequence lacks the required Arg at the P 4 site. Thus, our data suggest that the 203 R-204 E site is the predominant cleavage site and 200 R-201 M is an alternative cleavage site by furin when the predominant site is missing. This is consistent with the model proposed for wild type MT1-MMP, in which the pro-MT1-MMP is processed primarily at the 108 RRKR site to generate the active proteinase and the secondary site within 86 KXXRRXR is cleaved only when the primary 108 RRKR motif was mutated (39).
Notably, there are two potential consecutive furin recognition sites in other metalloproteinase zymogens, including ADAM11 (AB009675, 292 RLRRK 297 R), ADAM22 (AF155382, 219 RPKRSK 225 R), ADAMTS4 (AF148213, 206 RPRRAK 212 R), MT2-MMP (NM_002428, 126 RRRRK 131 R), and MT5-MMP (AJ010262, 118 RRRRNK 224 R). The ones nearer to the catalytic domains are conserved in different species, whereas the distal ones might be acquired later during evolution. Although the significance of the two alternative recognition sites in these precursors remains poorly understood, we may speculate that the processing of these zymogens are crucial for some biological events; the zymogens may be activated by furin at a different cleavage site even if the primary site is abolished by mutation.
Significance of Furin and Its Related PC Pathways in the Processing of Precursors-Furin and its related PCs have been demonstrated as the major enzymes responsible for the maturation of many precursors, such as some ADAMs and MMPs (Refs. 16 -23 and 37-40; this report). Furthermore, zymogens of BACE, a major enzyme related to Alzheimer's disease, and some growth factors and cell surface receptors, such as transforming growth factor ␤, insulin-like growth factor, hepatocyte growth factor receptor, and Notch1 receptor, are converted into their active forms by these PC pathways (30,31,(33)(34)(35)(36)62). Thus, this activation mechanism by a PC may play key roles in many physiological and pathological events. In fact, furin knockout mice are embryonic lethal (63), and inhibition of furin results in absent or decreased invasion and tumorigenicity of human cancer cells (64,65). The inability to activate many types of proproteins, including some pro-ADAMs and pro-MMPs, in furin null mice may contribute to the abnormal phenotypes during early development and morphogenesis in those mice. On the other hand, the design and synthesis of furin specific inhibitors may lead to a new strategy in the treatment of cancer and other diseases, such as Alzheimer's disease, in the human adult.