Shedding of Membrane Type Matrix Metalloproteinase 5 by a Furin-type Convertase

The shedding of membrane-associated proteins has been recognized as a regulatory mechanism to either up-regulate or down-regulate cellular functions by releasing membrane-bound growth factors or removing ectodomains of adhesion molecules and receptors. We have reported previously that the ectoenzyme of membrane type matrix metalloproteinase 5 (MT5-MMP) is shed into extracellular milieu (Pei, D. (1999) J. Biol. Chem. 274, 8925–8932). Here we present evidence that MT5-MMP is shed by a furin-type convertase activity in the trans-Golgi network. Among proteinase inhibitors screened, only decanoyl-Arg-Val-Lys-Arg-chloromethylketone, a known inhibitor for furin-type convertases, blocked the shedding of MT5-MMP in a dose-dependent manner. As expected, decanoyl-Arg-Val-Lys-Arg-chloromethylketone also prevented the activation of MT5-MMP, raising the possibility that the observed shedding could be autolytic. However, an active site mutant devoid of any catalytic activity, is also shed efficiently, thus ruling out the autolytic pathway. The shedding cleavage was subsequently mapped to the stem region immediately upstream of the transmembrane domain, where a cryptic furin recognition site, 545RRKERR, was recognized. Indeed, MT5-MMP and furin are co-localized in thetrans-Golgi network and the shed species could be detected inside the cells. Furthermore, deletion mutations removing this cryptic site prevented MT5-MMP from shedding. The resulting mutants express a gain-of-function phenotype by mediating more robust activation of proMMP-2 than the wild type molecule. Thus, shedding provides a potential mechanism to regulate proteolytic activity of membrane-bound MMPs.

Shedding or proteolytic release of membrane-bound molecules has been established as an important regulatory mechanism to down-regulate cell adhesive receptors such as L-selectin (1)(2)(3), generate soluble ligands such as tumor necrosis factor ␣ (4), heparin-binding epidermal growth factor (5), and Delta ligand for Notch (6), or release dormant transcriptional factors (7,8). Given the diverse molecules shed from membranes, specific mechanisms must have been evolved to handle the specific shedding needs in various biological processes. For proteins bound to plasma membrane, two members of the a disintegrin and metalloproteinase (ADAM) 1 family, tumor necrosis factor ␣-converting enzyme/ADAM17 and Kuz/ADAM10, have been identified as efficient sheddases (see review in Ref. 9). The availability of inhibitors against the ADAMs, e.g. hydroxamate-based synthetic compounds or tissue inhibitor of matrix metalloproteinase-3 (9,10), should allow rapid determination if a shedding process is ADAM-or metalloproteinasedependent, thus facilitating the identification and characterization of alternative shedding pathways.
Matrix metalloproteinases are a family of zinc-dependent and neutral pH optimal endopeptidases believed to play a critical role in the remodeling of extracellular matrix under both physiological as well as pathological conditions (11,12). To date, ϳ25 MMPs have been reported and confirmed by cDNA cloning and chromosomal localizations (for review, see Refs. 11 and 12). Although the majority of the MMPs are secretory in nature, a growing list of newly identified MMPs appear to be membrane-bound by at least three distinct anchoring mechanisms: 1) type I transmembrane domains for MT1, -2, -3, and -5-MMPs (13)(14)(15)(16); 2) glycosyl phosphatidylinositol linkage for MT4 and 6-MMPs (17,18); and 3) type II transmembrane domain for MMP-23/ cysteine-array MMP (19). MT1-MMP, the archetypal membrane-bound MMP, mediates proMMP-2 activation, cell invasion, migration, fibrinolysis, collagenolysis, and angiogenesis when anchored on plasma membrane (13, 20 -24). Truncation of the transmembrane (TM) domain renders MT1-MMP incapable of activating proMMP-2 in transfected cells (23), whereas a similarly TM-truncated MT1-MMP is capable of processing proMMP-2 when purified and assayed in vitro (25). Thus, the transmembrane domain along with its cytosolic domain may confer unique cellular localization required for the proper function of these membrane-bound MMPs (21,23,24).
MT5-MMP is a brain-specific MT-MMP closely related to MT1, -2, and -3-MMPs both structurally and functionally (16,26). For example, it activates proMMP-2 when co-transfected in various cells (16,26). Like MT1-MMP, recombinant MT5-MMP expresses proteolytic activities against extracellular matrix components such as proteoglycans (25,27). On the other hand, MT5-MMP appears to have several unique features. It has a short half-life of ϳ30 min at 37°C (27). In fact, a synthetic inhibitor, BB-94, has to be included in conditioned media to keep the enzyme from autocatalytic decay, thus ensuring its integrity throughout the purification process at 4°C (27). Furthermore, it is shed readily from cell surface (16). The shed species behaves like a secretory MMP and can be detected by gelatin zymography (16). Interestingly, BB-94, an inhibitor known to block both MMPs and ADAMs, did not inhibit the shedding process, suggesting that MT5-MMP be shed by a novel mechanism independent of metalloproteinase activity. In this report, we demonstrate that MT5-MMP is shed by a furintype convertase activity cleaving a cryptic furin recognition motif 545 RRKERR within its stem region. We propose that shedding provides a potential mechanism of down-regulation for MT5-MMP activity.

MATERIALS AND METHODS
Cell Lines, Chemicals, and Immunological Reagents-Cell lines including MDCK and its derivatives were obtained and maintained as described (16,28). The following stable lines were used: a stable cell line expressing full-length mouse MT5-MMP (F591) and a cell line expressing MT51-570F (16,28). Stable cell lines, EA20 and EA24, were generated by stable transfection of pCR3.1MT5-MMPE252A into MDCK cells and characterized as described (28). Laboratory chemicals and proteinase inhibitors were from Sigma or Calbiochem (San Diego, CA). The furin inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone (CMK) was purchased from Bachem (Philadelphia, PA). Cell culture reagents were from Life Technologies, Inc. Anti-MT5-MMP antibody was described previously (16). Anti-furin antibody was purchased from Affinity BioReagents, Inc. (Golden, CO). Secondary antibodies were from Jackson Immunoresearch Laboratories, Inc. (West Grove, PA). BB-94 was a gift from British Biotech (Oxford, United Kingdom).
Homogenization and Fractionation of Mouse Cerebellum-Fresh dissected cerebellum tissues were homogenized in 25 mM HEPES (pH 7.4) containing 0.32 M sucrose 10 mM EDTA, 5 M BB-94, 10 M aprotinin, 10 M E64, 10 M pepstatin, 1 mM phenylmethylsulfonyl fluoride, and centrifuged at 1,000 ϫ g for 10 min at 4°C. The resulting supernatant was further centrifuged at 100,000 ϫ g for 60 min at 4°C to separate the soluble protein supernatant from the membrane pellet, which was then resuspended in the same solution. Equivalent amount of proteins from pellet and supernatant were analyzed by SDS-PAGE and probed with anti-MT5-MMP antibody as described under "Cell Lines, Chemicals, and Immunological Reagents." Effects of Proteinase Inhibitors and Other Chemical Agents on MT5 Shedding-Cells transfected with the control vector or various forms of MT5-MMP were plated in six-well plates and allowed to grow to confluence. The cells were then washed with phosphate-buffered saline (PBS) three times and replenished with serum-free Dulbecco's modified Eagle's medium alone or supplemented with E64 (5 or 50 M), aprotinin (10 or 100 g/ml), pepstatin (5 or 50 M), BB-94 (5 or 50 M), CMK (50 M), A23187 (500 nM), or brefeldin A (BFA, 5 g/ml). The supernatants were collected 48 h later and analyzed by zymography and Western blotting as described (16). Cells were lysed in RIPA buffer and analyzed by Western blotting as described (16,28). Cells were also extracted with saponin as described and analyzed by Western blotting (19).
Generation and Characterization of MT5-MMP::GFP and Mutations in the Stem Region-The entire open reading frame of mouse MT5-MMP was isolated by Pfu-based PCR and cloned into a modified GFP construct as described (19). The MT5⌬22::GFP and MT5-MMP⌬6 mutants carrying deletions of 22 residues Gln 540 -Thr 561 or the 545 RRKERR motif from the stem region of MT5-MMP were constructed by duplex PCR with the 5Ј and 3ЈMT5-MMP primers (16), and two mutagenic pairs of primers: TGG ATG GGC TGC AAG ATC GAT GAC GTG CCA GG, CTT GCA GCC CAT CCA GTC; and CAG AAG GAG GTA GAG CTG CCC CAG GAT GAT GTG, CTC TAC CTC CTT CTG CTT to loop-out the 22 or 6 residues (29). The resulting fragments were cloned into pCR3.1GFP or pCR3.1, respectively, and subsequently confirmed by double stranded DNA sequencing. Error-free mutants were then selected to generate stable cell lines from MDCK cells as described (16). Immunofluorescent Staining and Confocal Microscopy-The cells grown on glass coverslips were fixed in 4% paraformaldehyde solution with Triton X-100 (1%). The coverslips were then blocked with 1% normal donkey serum and stained with anti-furin or anti-MT5-MMP antibodies, followed by Rhodamine Red X-or fluorescein isothiocyanate-labeled secondary antibody (1 h each). The slides were then scanned in a Bio-Rad confocal system at the University of Minnesota Bioimaging Laboratory. The images were then further processed using Adobe Photoshop version 6.

Profile of MT5-MMP Protein Products in Vivo-
We have demonstrated previously that MDCK cells expressing the brain-specific MT5-MMP shed a gelatinolytic species into conditioned media at ϳ27 kDa on zymography and a major species at ϳ34 kDa plus several minor ones on Western blots in the absence of any proteinase inhibitor, a pattern similar to that generated by autocatalytic fragmentation of the 56-kDa fulllength recombinant ectoenzyme (27). To test the hypothesis that MT5-MMP is also shed in natural settings, we profiled the expression of MT5-MMP in various regions of mouse brain and revealed by reverse transcription-PCR that it is expressed highly in cerebellum, modestly in cerebrum, but minimally in heart (Fig. 1A, middle panel of lanes 1-3). Consistently, Western blot analysis with anti-MT5-MMP antisera detected a major species at approximately 34 kDa in supernatants of tissue homogenates from cerebellum and cerebrum, but not heart (Fig. 1A, upper panels, lanes 1 and 2 versus lane 3). This soluble species from tissue homogenates is almost identical to the shed fragment of MT5-MMP from recombinant cells (16), suggesting that MT5-MMP is shed in vivo. When equivalent amounts of supernatants and pellets freshly prepared from cerebellum were analyzed simultaneously, the same 34 kDa was detected in the supernatants while a ϳ65-kDa species along with several smaller and minor ones was detected in the membrane , and heart (lane 3) tissues were divided into two portions. Total RNAs were extracted from the first set and analyzed by reverse transcription-PCR for the housekeeping gene, glyceraldehyde-3-phosphate dehydrogenase (lower panel, ϳ1.2 kilobases) and MT5-MMP (middle panel, ϳ0.4 kilobases) as described (16). The other sets were homogenized and the resulting supernatants were analyzed by Western blotting with anti-MT5-MMP antibody (top panel). Note the MT5-MMP-specific species at 34 kDa. P, protein; R, RNA. B, cerebellum tissues were homogenized and then separated into supernatants and membrane fractions as described under "Materials and Methods." Both supernatants (S) and the membrane pellets (P) were analyzed by Western blotting as described in A. Note the 65-kDa full-length species and the 34-kDa shed species from the membrane pellet and the supernatant, respectively, as marked on the right. pellet (Fig. 1B, lanes 2 versus 1, arrow). These products were also detected in recombinant cells expressing MT5-MMP (16,26), thus supporting the idea that MT5-MMP is shed in vivo in a similar fashion as observed in vitro. Furthermore, it is estimated that the 34-kDa species amounts to ϳ40% of the total MT5-MMP present in the pellets and supernatants. These data argue that MT5-MMP is shed in natural settings as well.
Shedding of MT5-MMP Is Resistant to a Broad Spectrum of Proteinase Inhibitors-Since the MT5-MMP protein products detected in recombinant cells appear to recapitulate those detected in cerebellum as demonstrated in Fig. 1, we decided to focus on this established in vitro system to dissect the shedding process (16). In fact, this strategy remains the only viable alternative since the primary cells isolated from mouse cerebellum apparently lost the expression of MT5-MMP at the mRNA and protein levels in culture (data not shown).
To probe the proteolytic mechanism of shedding, we attempted to block shedding with various proteinase inhibitors. Control as well as MT5-MMP-transfected MDCK cells (F591) constitutively expressed MMP-9 migrating at approximately 92 kDa on zymography in Fig. 2A (lanes 1-14). A 27-kDa gelatinolytic species was secreted into conditioned media by MT5-MMP-transfected cells as reported previously ( Fig. 2A, lane 6, arrowhead) (16). Proteinase inhibitors including aprotinin for serine proteinases (10 or 100 g/ml), E64 for cysteine proteinases (5 or 50 M), BB-94 for metalloproteinases (5 or 50 M), and pepstatin A for aspartyl proteinases (5 or 50 M) (30) in serum-free culture media failed to block the shedding of the ϳ27-kDa gelatinolytic species ( Fig. 2A, lanes 6 -14). BB-94 did not inhibit the shedding process as reported previously, even at concentrations as high as 50 M, but apparently converted the smaller fragments into a ϳ52-kDa species as detected by both zymography and Western blotting ( Fig. 2A, lanes 7, 8, 17, and 18 marked by arrows). Thus, this 52-kDa species should be considered as the primary product of shedding, which was autocatalytically fragmented into the smaller ones at 27 or 34 kDa, in agreement with our previous report that BB-94 stabilizes active MT5-MMP (27). Consequently, BB-94 was always included in culture media to prevent autocatalytic fragmentation when analyzing MT5-MMP shedding. Since BB-94 blocks the activities of ADAMs, these data would rule out the ADAMs as the MT5-MMP sheddase, thus suggesting that MT5-MMP is shed by an unknown mechanism.
A Chloromethylketone-based Inhibitor of Furin-type Convertases Blocks the Activation as Well as the Shedding of MT5-MMP-Like other MT-MMPs, MT5-MMP contains a furin site and may be activated by furin in the TGN as demonstrated previously (29,31). Indeed, MT5-MMP-transfected cells are capable of activating proMMP-2 (16,26), indicating that MT5-MMP must have been processed and activated. We analyzed cell-associated MT5-MMP products by Western blotting. In the absence of BB-94, only the 65-kDa pro species was detected (Fig. 2B, lane 1). The active species at ϳ58 kDa became detectable with the addition of BB-94, which inhibited autocatalytic decay as described (27) (Fig. 2B, lane 2). The minor and smaller ones are nonspecific and present in MDCK cells as well (data not shown). To implicate furin as the activator, a furin inhibitor, CMK, was included in the culture media as indicated in Fig. 2B, a dose-dependent inhibition of MT5-MMP processing was observed (lanes 2-6), suggesting that MT5-MMP is processed by a furin-type convertase (32). In the supernatants, we observed not only the expected conversion of the smaller species into the 52-kDa one (Fig. 2B, lanes 7 and 8) but also a dose-dependent decrease of the shed species (Fig. 2B, lanes  8 -12) by Western blotting, indicating for the first time that a furin-type convertase activity is required for the observed shed-ding. Although CMK is expected to block the activation of MT5-MMP due to the presence of a consensus furin recognition site between its pro and catalytic domains (16), its efficient blocking of shedding is quite unexpected.
Shedding of Catalytically Inactive MT5E252A-Given the fact that active MT5-MMP autocatalytically fragments itself into smaller species (27), it is possible that CMK inhibited the shedding by blocking furin-mediated activation of MT5-MMP, thus preventing autocatalytic shedding. To rule out this possibility, we analyzed the shedding profile of a catalytically inactive mutant of MT5-MMP, MT5E252A. As shown in Fig. 3A, this mutant carries a single point mutation converting Glu to Ala at the active site, rendering MT5-MMP inactive, as demonstrated by its inability to activate proMMP-2 (16). Should shedding be autocatalytic, MT5E252A should not be able to shed its ectodomain. However, the ectodomain of MT5E252A was shed very efficiently into conditioned media when two independently derived cell lines, EA24 and EA20, were analyzed (arrows, Fig. 3B, lanes 3, 4, 7, and 8). In fact, the mutant protein was shed as a single species at ϳ52 kDa in contrast to the smaller ones from wild type MT5-MMP in the absence of BB-94 (Fig. 3B, lanes 3 and 4 versus lane 2). Furthermore, BB-94 did not cause any upshift in molecular weight for MT5E252A as it did for the wild type molecule (Fig. 3B, lanes  7 and 8 versus lane 6), reinforcing the notion that the smaller molecular mass species are derived autocatalytically from the ϳ52-kDa species. Without any proteolytic activity, MT5E252A is more stable than the wild type molecule both inside and outside the cells, as indicated by the detection of the cellassociated 58-kDa processed product without BB-94 (arrowhead, Fig. 3C, lanes 1 and 7 versus Fig. 2B, lanes 1 and 7). In addition, we observed an intracellular species co-migrating with the shed 52-kDa species (Fig. 3C, asterisk between lanes 1  and 2), suggesting that it may be generated intracellularly prior to secretion. We then estimated the percentage of shedding for MT5E252A at ϳ30% at the steady state level (Fig. 3C,  lane 7 versus lane 1). Consistent with data for wild type MT5-MMP in Fig. 2, CMK blocked both the processing and the shedding of MT5E252A in a dose-dependent manner (Fig. 3C,  lanes 2-6 and 8 -12), reinforcing the idea that MT5-MMP is shed not autocatalytically, but by a furin-type convertase in trans. The estimated IC 50 for CMK to block shedding is ϳ7 M. These data demonstrate for the first time that furin or its related proprotein convertases could serve as a sheddase for membrane-bound molecules.
Co-localization of MT5-MMP and Furin in the trans-Golgi Network-As described earlier, MT5-MMP contains a furin motif, RRRNKR 124 and, thus, may be processed by furin for zymogen activation (32,33). Indeed, CMK inhibited the processing of MT5-MMP precursor as well as its shedding into media as shown in Figs. 2B and 3C. These data raise the possibility that MT5-MMP and furin should co-localize in the TGN. To test this possibility, we performed double immunofluorescence staining with anti-MT5-MMP and anti-furin antibodies. Surprisingly, most of MT5-MMP signals are localized intracellularly within TGN (Fig. 4A, panel a), in contrast to the report that the archetypal MT1-MMP is primarily localized on plasma membrane (13). The staining pattern for furin, on the other hand, is localized in the TGN as reported previously (Fig.  4A, panel c) (34). When overlaid, it is apparent that both furin and MT5-MMP are co-localized (Fig. 4A, panel b). Similar staining patterns were observed between MT5-MMPE252A and furin (data not shown). These co-localization data argue that both the activation and shedding of MT5-MMP may occur in the TGN simultaneously. Fig. 4A suggests that most of the MT5-MMP molecules might have been shed prior to be presented onto plasma membrane with little MT5-MMP as membrane-bound forms beyond the TGN. The fact that MT5-MMP is co-localized with its sheddase, furin, in the TGN (Fig. 4A) raises the possibility that shedding takes place, just like activation, in the TGN. Indeed, a cell-associated 52-kDa species, detected in Fig. 3C (marked by an asterisk), closely resembles the shed species (also 52 kDa), and is sensitive to CMK treatment. Should this 52-kDa protein be the shed species prior to being secreted, it must have lost its transmembrane domain intracellularly. To prove that this 52-kDa species lost its transmembrane domain by shedding prior to secretion, we performed saponin extractions, a technique that can differentially extract soluble proteins from intracellular compartments (19). As shown in Fig To further confirm the intracellular nature of MT5-MMP shedding, we treated the cells with pharmacological agents known to disrupt vesicular trafficking or furin maturation. As shown in Fig. 4C, A23187, a calcium ionophore known to inhibit furin maturation in the ER (35), blocked the shedding by depleting the calcium in ER (lane 8 versus lane 6). Similarly, BFA inhibited the shedding process by interfering with ER to Golgi transport (Fig. 4C, lane 9 versus lane 6) (36). As controls, BB-94 did not and CMK did block the shedding process as described previously (Fig. 4C, lanes 7 and 10; see also Figs. 2B and 3C). Together, these data argue that the ectodomain of MT5-MMP is shed in the TGN.

Detection of Shed MT5-MMP in Intracellular Compartment and Inhibition of Shedding by Brefeldin A and A23187-The near absence of MT5-MMP on cell surface as shown in
Mapping of the Shedding Cleavage within the Stem Region-To estimate the approximate location of the shedding cleavage, we compared the mobility of the shed species with that of MT5 1-570 F, a secretory form generated by deleting its transmembrane domain (27). As shown in Fig. 5A, the shed species is predicted to be approximately 3 kDa smaller than MT5 1-570 F (marked by two small horizontal bars between lanes 1 and 2). The estimated molecular mass for the entire stem region, 538 CKQKE VERRK ERRLP QDDVD IMVTI DDVPG SVN 570 plus the FLAG epitope, DYKDDDDK, is 4.8 kDa (Fig.  5B). A differential of ϳ3 kDa would suggest that the shedding cleavage occurs around R 550 L (Fig. 5B), where a cryptic furin recognition motif and cleavage site, 545 RRKERR, was identified (16,32).
The 545 RRKERR Motif Is Required for Shedding-The localization of shedding cleavage around the cryptic 545 RRKERR motif within the stem region suggests that it may play a critical role in the observed shedding process. To ascertain the contribution of the stem region in regulating the shedding process, we constructed two deletion mutants. We first constructed the MT5⌬22 mutant, which contains a deletion of 22 residues from the stem region, including the 545 RRKERR site (Fig. 6A). To track the cellular localization of MT5-MMP, we linked a GFP molecule at the C terminus of MT5-MMP to generate MT5⌬22::GFP (Fig. 6A). As shown in Fig. 6B, the fusion of a GFP molecule at the C terminus did not affect the shedding process, as shed species were detected in media conditioned by the wild type MT5::GFP fusion in the absence or presence of BB-94 (lanes 7 and 8). However, removal of the 22 residues from the stem region in MT5⌬22::GFP abolished the shedding process completely (Fig. 6B, lanes 9 and 10), despite more robust expression for MT5⌬22::GFP than the wild type mole-cule as demonstrated by Western analysis of the cell lysates (Fig. 6B, lanes 4 and 5 versus lanes 2 and 3), arguing that the stem region regulates MT5-MMP shedding. To test if the 545 RRKERR motif is required for the shedding process, we constructed a deletion removing only the 545 RRKERR motif in MT5-MMP and named it MT5⌬6 as shown in Fig. 6A. Stable cell lines transfected with MT5⌬6 did not shed any appreciable amount of MT5-MMP ectodomain, while the wild type MT5-MMP did as demonstrated in Fig. 6C (lanes 5 and 6 versus  lanes 3 and 4). Consequently, the cell-associated MT5⌬6 was higher than its wild type counterpart (Fig. 6C, lanes 11 and 12  versus lanes 9 and 10). Together, we conclude that the cryptic furin recognition site, 545 RRKERR, is the cis-acting signal in the stem region required for the observed shedding.
Functional Consequence of MT5-MMP Shedding-Initial characterization for MT5-MMP activity was demonstrated when both MT5-MMP and MMP-2 constructs were co-transfected into MDCK cells (16), because proMMP-2 added to the transfected cells was not activated efficiently (data not shown). In light of the observed shedding of MT5-MMP, this inefficiency could be explained by the near absence of cell surfaceassociated MT5-MMP due to shedding (Fig. 4A). To monitor the consequence of shedding, we analyzed the localization pattern of GFP-tagged wild type MT5::GFP or the MT5⌬22::GFP mutant (see Fig. 6A for details). As shown in Fig. 7A, deletion of the stem region containing the cryptic 545 RRKERR signal significantly enhanced the cell surface expression of MT5- MMP  (panel b versus panel a). Prominent signals were observed on plasma membrane for MT5⌬22::GFP in almost every cells (Fig.  6A, panel b), whereas the wild type MT5::GFP is sequestered in  1-6) were washed three times with PBS before being lysed (lanes 1 and 4) or extracted with saponin as supernatants (lanes 2 and 3) or remaining pellets (lanes 5 and 6) as described (19). The collected samples were analyzed by Western blotting as described in Fig. 1B Fig. 4A (panel a)), suggesting that the attachment of GFP did not alter the trafficking of MT5-MMP. In a time-course study of proMMP-2 activation by MT5::GFP and MT5⌬22::GFP mutant, we observed more robust activation of proMMP-2 by MT5⌬22::GFP than the wild type protein (Fig.  7B, lanes 2, 5, 8, and 11 versus lanes 3, 6, 9, and 12). The difference was most dramatic in some of early time points such as the 4-h mark when MT5⌬22::GFP activated a significant portion of proMMP-2 while wild type MT5-MMP::GFP did not (Fig. 7B, lane 2 versus lane 3). Taken together, these data demonstrate that shedding down-regulates the activity of MT5-MMP and the deletion mutants, MT5⌬22::GFP and MT5⌬6, could be considered as gain-of-function mutants. DISCUSSION The plasma membrane of a cell is populated with various surface molecules that can (i) receive specific extracellular signals, (ii) instruct neighboring cells to proliferate or differentiate, or (iii) remodel or modify the neighboring microenvironment. To properly execute their physiological functions, cells must regulate their plasma membrane contents with precision and efficiency. In general, cell surface molecules are synthesized in the ER, processed and packaged in the TGN and then delivered via secretory vesicles to the plasma membrane (37). For cell surface receptors, ligand binding usually triggers rapid internalization (38). The internalized receptors may be recycled back to the cell surface or delivered to lysosomes for degradation (38). For molecules with no obvious ligands, their fates at cell surface are less well understood. With increasing frequency, cell surface molecules are found to be shed into extracellular milieu to either down-regulate their function such as the shedding of L-selectin from leukocytes or generate functional forms, as exemplified by the release of soluble tumor necrosis factor ␣ (1, 4). Thus, shedding has been recognized as an important regulatory mechanism for cells to control its microenvironment. The MT-MMPs qualify as cell surface molecules that can modify or remodel the extracellular environment. Destructive in nature, these enzymes like all other MMPs are regulated at multiple levels, including transcriptional and translational controls, zymogen activation, and inhibitions by endogenous inhibitors such as tissue inhibitor of matrix metalloproteinases (11,12). Being membrane-anchored, MT-MMPs are subject to additional regulations such as vesicular trafficking. At the present, little is known about how MT-MMPs are regulated on the plasma membrane. The shedding of MT5-MMP presented in this paper offers a concrete mechanism how membrane-bound MMP activity at the cell surface could be regulated.
Shedding of MT5-MMP Is Obligatory and Evolutionarily Conserved-MT5-MMP distinguishes itself being readily shed into extracellular milieu (Figs. 2B and 3C) (16). In this report, we present evidence that the ectoenzyme of MT5-MMP is shed through an obligatory mechanism by a furin-type convertase recognizing a specific motif, 545 RRKERR, within its stem region. Since furin-type convertases are ubiquitous, this shedding mechanism should be operational in almost all cell types characterized so far (32,34,35). Indeed, we have also transfected MT5-MMP into various cell lines from human, rat, hamster, and canine and observed shedding in all cell lines examined (data not shown). Furthermore, extracts from mouse cerebellum contains a 34-kDa soluble species, almost identical to the shed species identified in transfected cells, suggesting that MT5-MM is shed in vivo. It is of interest to point out that MT5-MMPs from mouse (16), rat (AB023659), and human (26) all contain an identical RRKERR motif within the stem region, arguing that the shedding process in evolutionarily conserved.
Modulation of MT5-MMP Level on Cell Surface by Shedding-In addition to the 545 RRKERR motif in the stem region, MT5-MMP contains a bona fide furin recognition site R Ϫ6 RR Ϫ4 NK Ϫ2 R Ϫ1 sandwiched between its pro and catalytic domains (16,26), presumably for zymogen activation as described for similar MMPs (25,29,31). We envision two scenarios by which furin or related convertases may regulate MT5-MMP activity on cell surface. First, there is a distinct convertase for activation and another one for shedding. This is likely given the fact that the R Ϫ6 RR Ϫ4 NK Ϫ2 R Ϫ1 site between pro and catalytic domains is a perfect consensus for furin, the archetypal proprotein convertase (32), whereas the 545 RRK-ERR in the stem region is sub optimal for furin recognition due to the presence of a Lys at Ϫ4 position instead of the preferred Arg (32,34). In fact, the motif in the stem region could be viewed as two tandem dibasic motifs, 545 RR and 549 RR, which can be recognized by those proprotein convertases expressed in the regulated secretory pathway in neuro-endocrine cells (32,34). Thus, furin could be the activating convertase while the  (27), were analyzed side-by-side by Western blotting with anti-MT5-MMP antibody. The molecular sizes were estimated according to relative mobility with Stratagene Eagle Eye system (La Jolla, CA). B, localization of the shedding cleavage around a cryptic furin-type convertase motif, 545 RRKERR. The wild type and E252A mutant of MT5-MMP are illustrated along with the C-terminally truncated mutant MT5 1-570 F (16,27). Amino acid sequence for the stem region and the attached FLAG tag is shown with the calculated molecular mass of 4.8 kDa. The shed ectodomain is ϳ3 kDa shorter than MT5 1-570 F and should have a C terminus near R 550 L, as indicated by an upward arrow.
other dibasic convertases could be the sheddases. Alternatively, both activation and shedding may be mediated by the same convertase. Furin, PACE4, or PC7/8 all prefer motifs with the Ϫ4 position as Arg and Ϫ6 position as Arg in addition to the dibasic Arg/Lys at Ϫ2 and Arg at Ϫ1 positions (39,40). A Lys at Ϫ4 position lowers the efficiency of cleavage by these three convertases (39,40), especially when the concentration of the convertase is limited. This difference in processing efficiencies may offer a mechanism for cells to balance the ratio between activated/membrane-bound and the activated/shed MT5-MMP by modulating the concentration of the convertase(s) in the TGN. At relatively high concentration of convertase(s), both the activation site and shedding site may be cleaved, thus, favoring the secretion or shedding of MT5-MMP ectoenzyme. On the other hand, the concentration of the convertase(s) may be relatively low, thus allowing the activation of MT5-MMP zymogen, but not enough to cleave the shedding site, thus favoring accumulation of active MT5-MMP on the cell surface. The fact that both shedding and activation take place in the TGN suggest that, once past this compartment without being shed, MT5-MMP should remain membrane-bound until reaching the plasma membrane. The presence of furin-type convertases in the plasma membrane (34) would also shed MT5-MMP on the cell surface into the extracellular milieu and thus down-regulate its function.