The Cytoplasmic Tail Dileucine Motif LL572 Determines the Glycosylation Pattern of Membrane-type 1 Matrix Metalloproteinase*

Membrane-type 1 matrix metalloproteinase (MT1-MMP; MMP-14) drives fundamental physiological and pathological processes, due to its ability to process a broad spectrum of substrates. Because subtle changes in its activity can produce profound physiological effects, MT1-MMP is tightly regulated. Currently, many aspects of this regulation remain to be elucidated. It has recently been discovered that O-linked glycosylation defines the substrate spectrum of MT1-MMP. We hypothesized that a mutual interdependency exists between MT1-MMP trafficking and glycosylation. Lectin precipitation, metabolic labeling, enzymatic deglycosylation, and site-directed mutagenesis studies demonstrate that the LL572 motif in the cytoplasmic tail of MT1-MMP influences the composition of the complex O-linked carbohydrates attached to the hinge region of the protein. This influence appears to be independent from major effects on cell surface trafficking. MT1-MMP undergoes extensive processing after its synthesis. The origins and the molecular characters of its multiple forms are incompletely understood. Here, we develop and present a model for the sequential, post-translational processing of MT1-MMP that defines stages in the post-synthetic pathway pursued by the protein.

Consequently, no cellular mechanisms that are involved in the generation of distinctive glycosylation species of MT1-MMP have been identified until now. We hypothesized in this context that a mutual interdependency exists between MT1-MMP trafficking and glycosylation. The function for the cytoplasmic tail of MT1-MMP in determining this protein's subcellular trafficking has been well established (for example see Refs. 20 -22). Here, we demonstrate that the cytoplasmic tail's dileucine motif LL 572 also influences the O-glycosylation pattern of MT1-MMP. This influence appears to be independent of major effects on MT1-MMP distribution, as demonstrated by metabolic pulse-chase experiments and assessments of steady-state surface expression levels. In addition, a comprehensive model for the sequential, post-translational processing of MT1-MMP is elaborated and presented.

EXPERIMENTAL PROCEDURES
General Reagents-Unless otherwise noted, all reagents were purchased from Sigma-Aldrich.
Sequence Analysis of MT1-MMP-The amino acid sequence of MT1-MMP was obtained from Swiss-Prot (P50281). Protein alignment and identification of domains were performed on the T-coffee server (www.ch.embnet.org/software/Tcoffee.html). The prediction of O-and N-glycosylation sites was performed by analyzing the full-length sequences on the NetOGlyc 3.1 (www.cbs.dtu.dk/services/NetOGlyc) and NetNGlyc 1.0 (www. cbs.dtu.dk/services/NetNGlyc) server (Center for Biological Sequence Analysis, Technical University of Denmark, Lyngby, Denmark) (23). The prediction of N-glycosylation sites was based on identification of Asn-Xaa-Ser/Thr consensus sequences.
Tissue Culture-Cell culture media and reagents were purchased from Invitrogen. COS cells were grown in flasks under standard conditions (37°C and 5% CO 2 ) in minimal essential medium supplemented with L-glutamine (GlutaMAX), 10% fetal calf serum, and penicillin/streptomycin.
Lectin Precipitation Assay-Cells were transfected with double tagged MT1-MMP constructs and lysed according to the protocol described above. For lectin precipitation, 900 l of cell lysate were preincubated with 25 g of biotinylated wheat germ agglutinin (WGA) or biotinylated, succinylated wheat germ agglutinin (sWGA) (Vector Laboratories, Burlingame, CA) for 4 h at 4°C before 60 l of a washed 50% slurry of streptavidin conjugated beads (Pierce Biotechnology, Rockford, IL) were added. After overnight incubation at 4°C on a rotating platform, beads were washed five times in 1% Triton X-100 lysis buffer. Proteins were released by adding 2ϫ sample buffer and heating (10 min and 56°C). Blots of WGA precipitates were incubated with anti-HA mAb. Western blots of lysate samples were incubated with anti-HA, anti-FL, and anti-␤-actin mAb.
Direct Metabolic Labeling of GalNAc-The basic principle and chemistry of the metabolic in vivo labeling of O-linked glycosides by bioorthogonal chemical tags have been described elsewhere in detail (26 -28). The method is based on the metabolic incorporation of an azide modified sugar, here 1,3,4,6tetra-O-acetyl-N-azidoacetyl-␣,␤-D-galactosamine (Ac 4 -GalNAz), into protein glycan structures. The glycoproteins which have incorporated the modified sugar are detected by the specific reaction of the azide modified sugar with an alkyne. In brief, COS cells were transfected with double tagged MT1-MMP constructs and incubated for 48 h under tissue culture conditions at a 40 M final concentration of Ac 4 -GalNAz (C33365, Invitrogen). Cells were lysed according to the protocol described above, and cell lysates were immunoprecipitated with anti-HA mAb (compare below). Proteins were released by boiling (5 min at 95°C) the protein A beads (Pierce) in 50 mM Tris-HCl, pH 8.5, containing 1% SDS. Covalent binding of glycoside incorporated Ac 4 -GalNAz to 5-carboxytetramethylrhodamine (TAMRA) alkyne (C33370, Invitrogen) followed by protein precipitation by methanol and chloroform was performed according to the manufacturer's instructions. Proteins were separated by SDS-PAGE gel electrophoresis. TAMRA fluorescence was detected in the polyacrylamide gels with a UVlight source and a charge-coupled device camera system with an ethidium bromide filter set (Gel iX Imager, Intas Science Imaging Instruments GmbH, Goettingen, Germany).
Cell-surface Biotinylation-Protocols and standard procedures for cell-surface protein biotinylation have been described previously (29). In brief, manipulations were performed on a rocking platform at 4°C with ice-cold reagents and buffers. Cell-surface biotinylation was carried out 24 h after transfection of cells in 6 well with double tagged MT1-MMP constructs. Cells were washed twice with PBS and once with biotinylation buffer (10 mM triethanolamine, 2 mM CaCl 2 , 125 mM NaCl, pH 8.9). 1 ml of biotinylation buffer supplemented with 2.5 mM sulfo-NHS-SS-biotin (EZ-Link, Pierce) was added twice per well for incubations of 25 min each. Cells were washed twice for 5 min with PBS supplemented with 100 mM glycine and once with PBS before they were lysed. Biotinylated proteins were captured with streptavidin beads (see protocol for the lectin precipitation assay).
Immunoprecipitation-Cell lysates and immunoprecipitates were prepared according to the procedures described for the lectin precipitation assay. For immunoprecipitations, cells were transfected with MT1-MMP constructs. 15 g of anti-HA mAb (MMS-101R, Covance, Berkeley, CA) were added to the cleared lysates. Immobilized protein A beads (Pierce) were washed, blocked with 3% bovine serum albumin in lysis buffer for 4 h and washed again prior to the addition of 50 l of a 50% slurry of these beads to each preincubated lysate-Ab mix.
Enzymatic Desialylation-Enzymatic desialylation with sialidase A (GK80040, Prozyme Inc., San Leandro, CA) was performed according to the manufacturer's instructions. In short, transfected cells were lysed, and 10 l of lysate was mixed and incubated with 3 l of 5ϫ reaction buffer and 2 l of sialidase A (activity Ͼ 5 units/ml) or solvent control for 12 h at 37°C. The reaction was stopped by addition of 6ϫ reducing sample buffer and heating (10 min for 56°C). Samples were immunoblotted with anti-HA mAb.
Metabolic Labeling and Pulse-chase Experiments-Experiments were performed according to the procedures described elsewhere (30). In short, COS cells were transfected with double tagged MT1-MMP constructs. 24 h after transfection, COS cells were washed twice with PBS and incubated at 37°C for 40 min in depletion media (serum, methionine, and cysteine-free minimal essential medium) before pulse labeling with 150 Ci of L-[ 35 S]methionine/cysteine in 1 ml of depletion medium per 35-mm dish for 25 min at 37°C. Cells were placed on ice after 30 min and washed twice with ice-cold chase media (minimal essential medium supplemented with 10% fetal bovine serum and 10 mM L-methionine and 10 mM L-cysteine, pH 7.4, at 5% CO 2 ). Cells were then switched back to 37°C and incubated with chase media. After the time intervals indicated in Fig. 5, individual dishes were placed on ice, washed with ice-cold PBS, and lysed. Cell lysates were immunoprecipitated with anti-HA mAb. Lysates of untransfected cells and lysates of transfected cells without addition of precipitating Ab were used as controls (Fig. 5).

Sequence and Glycosylation Analysis of MT1-MMP-
To generate a glycosylation-defective CHO mutant of MT1-MMP that could be used as an appropriate control for lectin precipitation assays, ezymatic dessialylation, and pulse-chase experiments, the glycosylated residues of MT1-MMP first had to be identified. Analysis of potential N-glycosylation sites in MT1-MMP with the NetNGlyc server software revealed the absence of the canonical Asn-Xaa-Ser/Thr N-glycosylation sequence, indicating that this protein cannot be subject to this post-translational modification. The prediction of O-glycosylated residues is less straightforward, because no consensus recognition sequence is known for O-glycosyltransferases (23). Five residues were predicted by the trained neural network of the NetOGlyc server as potential sites for O-glycosylation (Fig. 1b). All residues are within the small hinge domain, which connects the large catalytic and hemopexin domains. The prediction identifies 76% of glycosylated and 93% of non-glycosylated residues correctly (23). Consequently, the possibility of a false-positive prediction of a glycosylated residue is higher than missing one. If the G-score (the score from the best general predictor) of a residue is above the threshold of 0.5, it is predicted to be glycosylated. The confidence of the prediction increases with the score. Five residues of MT1-MMP received scores above 0.5 (Thr 291 , Thr 299 , Thr 300 , Ser 301 , and Ser 304 ), however, one of these residues (Ser 304 ) had a G-score of only 0.508 (Fig. 1b). To examine the potential glycosylation status of these residues two mutant constructs, CHO-4 and CHO-5, were prepared. CHO-5 includes all of the mutations present in CHO-4 (T291A, T299A, T300A, S301A) as well as the mutation S304A (Fig. 1b). Because the electrophoretic mobilities of the MT1-MMP putative glycosylation mutants CHO-4 and CHO-5 did not differ from one another ( Fig.  2a) we conclude that Ser 304 is not glycosylated. Furthermore, although enzymatic deglycosylation produces a dramatic shift in the electrophoretic mobility of wild-type MT1-MMP (Fig. 2d), no such shift was observed after the enzymatic deglycosylation of CHO-4 ( Fig. 2b), demonstrating that the mutations incorporated into CHO-4 are sufficient to abrogate O-glycosylation of the MT1-MMP protein. Accordingly, CHO-4 was used for all subsequent experiments.
Expression of MT1-MMP and Glycosylation Analysis of Its Metabolites-The extensive post-translational processing of MT1-MMP produces a diversity of discrete metabolites. It was critical for our study to distinguish the protein's glycosylated and unglycosylated species. The location of the HA tag close in the protein's sequence to its single transmembrane-spanning segment permits the detection of all membrane-tethered metabolites. Western blot analysis (with an anti-HA Ab) displays a distinct expression pattern of HA-tagged CHO-4 and WT MT1-MMP expressed in COS cells. Both proteins comprise several metabolites, of which they share only one with an apparent molecular mass of 63 kDa (Fig. 2c). Besides the common 63-kDa protein, expression of the CHO-mutant generates additional bands of 53 and two bands of ϳ34 kDa, whereas the WT protein's other metabolites have apparent molecular masses of ϳ69, 57, and 44 kDa (Fig. 2, c and d). Because enzymatic deglycosylation is expected to alter only the electrophoretic mobility of glycosylated proteins, we conclude that any of the multiple forms of MT1-MMP that exhibit a change in apparent molecular mass after sialidase A treatment possess sialic acid-containing glycosylation. Based upon this criterion, all of the MT1-MMP forms (Fig. 2d, arrows 1, 3, and 4) detected except for the 63-kDa protein (Fig. 2d, arrow 2) are glycosylated. Accordingly, we suggest that the 63-kDa protein is the biosynthetic precursor from which all of the glycosylated protein species must be derived. This hypothesis is explored further in the pulse-chase analysis presented below and in Fig. 5.
Autocleavage of MT1-MMP-Expression of the wild-type MT1-MMP protein in COS cells results in the appearance of a glycosylated 44-kDa metabolite that is likely to be the product of autocatalytic cleavage (Fig. 2, c and d). The catalytically inactive MT1-MMP construct E240A was used to rule out a possible role of endogenously expressed COS cell proteases in this proteolytic cleavage of MT1-MMP (Fig. 3a). Because expression of E240A did not result in the generation of a membranetethered 44-kDa product, we conclude that this product is generated by self-proteolysis.
It has been suggested that glycosylation exerts a major effect on the a, MT1-MMP has a distinct domain structure. In the constructs used for the present studies, the FLAG epitope tag (FL) was inserted between the furin cleavage site and the N terminus of the catalytic domain (ct). After the autocleavage of MT1-MMP and shedding of the catalytic domain the FLAG tag is thus released into the extracellular media. To ensure the capacity to detect the remaining membrane-associated portion of the cleaved protein, a HA tag (HA) was introduced in the stalk domain (st) that links the hemopexin domain to the transmembrane domain. This tag localization enables identification of most membrane-tethered MT1-MMP metabolites. b, glycosylation analysis. The protein sequence of MT1-MMP was analyzed with the NetNGlyc 1.0 Server and NetOGlyc 3.1 server for potential O-and N-linked glycosylation sites. No N-glycosylated residues were found. The prediction of O-glycosylated residues is based upon a trained neural network approach. If the score of the best general predictor, the G-score, is above the threshold of 0.5 then the residue is predicted to be glycosylated. All putative glycosylation sites are in the hinge domain (hn) of the protease (sp ϭ signal peptide, pro ϭ pro-domain, tm ϭ transmembrane domain, ct ϭ cytoplasmic tail).  1, 3, and 4). This analysis confirms that the 63-kDa band (no shift, arrow 2) represents the only unglycosylated form of the WT protein. DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 35413
self-proteolysis of MT1-MMP (31). This conclusion was mainly based on the observation that the prevalence of the 44-kDa autocleavage product increased after inhibition of O-glycosylation in cells by a synthetic inhibitor (31). In our hands, the MT1-MMP WT and the glycosylation-defective mutant CHO-4 proteins yielded similar amounts of autocleavage product, when the anti-HA-Ab was used for detection (Figs. 2c, 4e, and 6a). In contrast, results obtained utilizing the antihinge domain Ab (AB815, Chemicon), which has been used in previous studies for immunoblotting, suggested enhanced generation of self-cleavage product in CHO-4 compared with WT (arrow, Fig. 3b). Interestingly, enzymatic desialylation of MT1-MMP WT in vitro also considerably increased the detection of the autocleavage product with the anti-hinge domain Ab (arrow, Fig. 3c). Because MT1-MMP activity was blocked during the in vitro desialylation, this increase cannot be attributed to enhanced autocleavage induced by desialylation. Instead, we conclude that the affinity of the anti-hinge region antibody varies dramatically with the glycosylation status of this domain.
Glycosylation Analysis of MT1-MMP Mutants-To investigate a possible connection of the glycosylation of MT1-MMP and its trafficking, the effects of alterations introduced into MT1-MMP's cytoplasmic tail motifs were investigated by lectin precipitation and metabolic labeling with Ac 4 -Gal-NAz. Western blot analysis of total cell lysates using the anti-HA and anti-␤-actin antibodies was employed to ensure that any differences in lectin binding were not due to unequal expression levels of the MT1-MMP constructs, or to variations in cell lysis efficiency or in the total protein loaded per lane (Fig. 4,  h-j). Fig. 4i depicts the results of Western blotting using an antibody directed against the FLAG tag epitope, which is inserted at the N terminus of the full-length MT1-  MMP protein (Fig. 1), and in Fig. 4j the blot is probed with an antibody directed against actin, which serves as a loading control. The data presented in this figure demonstrate that all of the constructs whose behaviors will be analyzed in subsequent figures are expressed at similar levels in transfected COS cells.
The black arrowhead in Fig. 4 indicates the position of the 44 kDa MT1-MMP self cleavage product. As demonstrated by desialylation, this metabolite is O-glycosylated and no unglycosylated species of this protein species can be observed (Fig. 2 d). Differences in metabolic labeling of O-linked glycosides and lectin precipitability were particularly apparent through direct comparison of the 44 kDa autocleavage fragment. For reasons of clarity and figure composition only this band is displayed in Fig. 4 c to g. Major differences were detected in the susceptibility of the various constructs to lectin precipitation Fig. 4c and metabolic labeling by GalNAz (Fig. f). The CHO-4 mutant showed only minimal interaction with WGA (Fig. 4c) and sWGA (Fig. 4d) and no metabolic labeling by GalNAz (Fig. 4f), consistent with the conclusion that all relevant glycosylation sites are abrogated in this construct.
The differences in lectin precipitability and metabolic incorporation of GalNAz indicate that the presence of the dileucine LL 572 motif has a major impact on the glycosylation of MT1-MMP (Fig. 4, a-f). The truncation constructs ⌬CT, ⌬567, and ⌬571 incorporate less GalNAz and are less precipitated by WGA and sWGA as compared with the full-length WT protein.
In contrast, ⌬575 and ⌬579 display GalNAz labeling and susceptibility to lectin precipitation that is comparable to that of the WT MT1-MMP protein (Fig. 4, c, d, and f). The ⌬571 and ⌬575 mutants differ from one another in length and in the presence of the 571 LLYC 574 sequence. In interpreting the patterns detected with these constructs, the different precipitation behaviors and metabolic GalNAz labeling of the ⌬579 and ⌬571-4 constructs are instructive. These constructs have the same total length but differ from one another in the presence of a defined subset of amino acids, namely the 571 LLYC 574 sequence. Furthermore, the cytoplasmic tail's second dileucine motif LL 579 is disrupted in ⌬579 and intact in ⌬571-4. Conversely, LL 572 is intact in ⌬579 and absent in ⌬571-4. Therefore, these data suggest that the LL 572 motif modulates the glycosylation status of the MT1-MMP polypeptide.
To further specify the function of 571 LLYC 574 , the dileucine and tyrosine motifs embodied within this sequence were each individually disrupted by alanine mutagenesis. The LL 571/572 AA mutation results in a pronounced reduction in metabolic GalNAz integration and susceptibility to WGA and sWGA pull down as compared with that observed with the WT protein and the Y573A and C574A mutants (Fig. 4, c, d, and f).
In principle, the lectin precipitation of MT1-MMP could be indirect and thus the differing susceptibilities of the constructs to lectin precipitation could be explained by differences in their affinity for another glycosylated protein that is pulled down by WGA and sWGA. As an additional control, and to test this possibility the preceding experiment was performed in reverse. HA-tagged MT1-MMP constructs were precipitated first with anti-HA Ab. GlcNAc-or sialic acid-modified proteins on the blot were then visualized by blotting with WGA (Fig. 4e). This approach is hampered by the fact that WGA binds to any pro-tein on the blot containing this modification, including the IgG used in the immunoprecipitations. Hence, Western blots were blocked with an optimized concentration of 2.5% bovine serum albumin, which served as a compromise allowing effective blocking of nonspecific protein binding to the blot without resulting in excessive binding of WGA to the bovine serum albumin blocking agent. Although this compromise results in a relatively high background signal, the specific bands corresponding to the 44-kDa autocleavage product can be readily identified (arrow, Fig. 4e). The data from this experiment are essentially consistent with the results from WGA precipitations. The clones that possess the LL 572 motif in their C-terminal tails are O-glycosylated in the hinge domain.
Surface Biotinylation Studies-Previous studies of the cytoplasmic tail of MT1-MMP have focused primarily on its influence on the trafficking and subcellular localization of the protein. We wondered, therefore, whether there is a connection between steady-state surface expression and the glycosylation of MT1-MMP. No such association was observed (Fig. 4g). The distinct behavior manifest by the 571 LLYC 574 construct with respect to its susceptibility to WGA and sWGA precipitation did not extend to alterations in the extent of its presence at the plasma membrane. The cell-surface expression levels of the 571 LLYC 574 -omitting construct ⌬571-4 and the LL 571/572 AA mutant are equivalent to that of WT. The surface expression level of ⌬579 was increased as compared with WT and in particular as compared with that of the ⌬571-4 protein. As noted above, the cytoplasmic tail's LL 579 motif is disrupted in ⌬579 and intact in ⌬571-4, whereas LL 572 is absent in ⌬571-4 and present in ⌬579. The interruption or removal of LL 579 (⌬CT, ⌬567, ⌬571, ⌬575, and ⌬579) but not LL 572 (LL 571/572 AA, ⌬571-4) resulted in general in the enhanced steady-state surface expression of MT1-MMP.
Metabolic Labeling and Processing of MT1-MMP-A metabolic pulse-chase experiment was performed to investigate the sequential order in which the various forms of the MT1-MMP protein appear following synthesis and to explore possible differences in the processing of the LL 572 and CHO-4 mutant in comparison to the full-length wild-type protein. Western blot analysis demonstrated equal and constant total expression of the various MT1-MMP constructs during the entire pulsechase protocol (Fig. 5a). The first band that became apparent after 10 min of 35 S pulse labeling incubation has a molecular mass of 63 kDa (Fig. 5b, arrows 1). This band represents the only common metabolite of the WT protein and CHO-4 (Fig. 2c). It is unglycosylated, because it has the same electrophoretic mobility in all three constructs, including CHO-4, and it did not shift after sialidase A treatment of MT1-MMP WT (Fig. 2d,  arrow 2). The band's molecular mass of 63 kDa matches the calculated molecular mass of pro-MT1-MMP after the loss of its 23-amino acid signal peptide (32). Pro-MT1-MMP rapidly underwent an increase in molecular mass in LL 572 and WT but not in CHO-4 (Fig. 5, arrows 2), which we interpret to reflect glycosylation of the protein (Fig. 6). This higher molecular mass band (apparent molecular mass of ϳ69 kDa) must be sialylated, because its electrophoretic mobility was altered after incubation with sialidase A (Fig. 2d, arrow 1). This form is also absent in the CHO mutant (Fig. 2c). In the same post-synthetic time DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51 interval, CHO-4 was processed to a 53-kDa fragment (Fig. 5b,  arrow 3) that corresponds to the calculated molecular mass of active MT1-MMP without the pro-domain. Activation of MT1-MMP has previously been demonstrated to occur intracellularly by limited proteolysis between Arg 111 and Tyr 112 , respectively, its pro-and catalytic domain, by furin or furin-like enzymes (Fig. 6) (10, 33). Concurrently or with a short delay, WT and LL 571/572 AA underwent the same apparent cleavage, resulting in a 57-kDa protein (Fig. 5b,  arrows 4). The 57-kDa fragment is glycosylated, as demonstrated by its behavior in SDS-PAGE following desialylation (Fig. 2d, arrow 3), and must therefore be a product of the corresponding higher molecular mass band, but not the unglycosylated 63-kDa pro-enzyme (Fig. 5b). It corresponds essentially to the 53-kDa fragment found in CHO-4 (Fig. 2c). WT and LL 571/572 AA underwent autocleavage to a 44-kDa fragment (Figs. 5b, arrows 5, and 6), whereas CHO-4 produced at least two fragments of ϳ34 kDa (Figs. 5, arrows 6, and 6). Again, the 44-kDa autocleavage product was glycosylated and sialylated (Fig. 2d,  arrow 4). It was, therefore, a direct cleavage product of the activated, "mature" 57-kDa protein (Fig. 6). No fragments matching the weights of the unglycosylated autocleavage products of CHO-4 were observed with WT, LL 571/572 AA, or with any cytoplasmic tail mutant. We conclude, therefore, that glycosylation precedes autocleavage but is not a prerequisite for further MT1-MMP processing. CHO-4 cleavage products match exactly the predicted molecular mass of previously described MT1-MMP fragments, as calculated from their amino acid sequences (summarized in Fig. 6) (32, 34 -36). As noted above, the higher resolution achieved with the large SDS-PAGE gels used for separating proteins in the pulse-chase experiments revealed that the selfproteolysis product of the CHO-4 construct consists actually of at least two distinct subspecies of different molecular weights (Fig. 5, arrows 6).

DISCUSSION
The pathways by which MMPs regulate development, homeostasis, and disease are not fully understood. Insight into their function in diverse biological processes will require the elucidation of their own intricate regulation. A more thorough knowledge of the mechanisms involved in this complex regulation may lead to the development of specific therapeutic interventions that modify proteolytic activities involved in distinct processes by interfering with their regulation rather than by blocking them  Fig. 2, b and c) allow the assignment of the protein bands to distinct processing steps. In brief, these steps include: 1, synthesis of the protein; 2, glycosylation (absent in CHO); 3 and 4, activation with loss of the pro-domain; and 5 and 6, generation of a membrane-tethered, autocatalytic cleavage product (see "Results" and "Discussion" for further details) (control ϭ untransfected COS; * ϭ transfected controls without precipitating Ab). globally. Here, we present a new model for the metabolic processing of MT1-MMP and demonstrate that the presence of the dileucine motif LL 572 in its cytoplasmic tail modulates the processing of the protein as revealed by metabolic labeling of O-linked glycosides by azide modified D-galactosamine and WGA and sWGA precipitation.
WGA is a lectin that is positively charged at a pH in the physiological range (37). It binds glycoconjugates containing N-acetylglucosamine (GlcNAc) and sialic acids, whereas its succinylated form is negatively charged at a physiological pH and fails to interact with sialic acids (17,37). Sialic acids are acidic monosaccharides that are typically located at the terminal positions of glycoconjugates (38,39). The sialic acid family comprises currently over 50 naturally occurring members that display a considerable structural and biochemical diversity (39,40). Because sWGA does not bind sialic acids and the precipitation of MT1-MMP by WGA and sWGA displays an essentially similar pattern, we conclude that the dileucine LL 572 motif's effect cannot be limited to the terminal residues of the MT1-MMP sugar complexes. Nonetheless, changes in the affinity of an anti-hinge domain antibody after sialidase A treatment of MT1-MMP in vitro emphasize the functional relevance of sialic acids for the antigenic presentation of the hinge region (Fig. 3, b and c). This effect becomes particularly apparent in the 44-kDa autocleavage product.
It had previously been suggested that O-glycosylation serves to stringently regulate the self-cleavage of MT1-MMP (31). The data presented in this report indicate, however, that apparent glycosylation-induced effects on autocleavage are actually attributable in large measure to the effects of glycosylation on the ability of the antibody used in those studies to detect the cleavage product. Our data demonstrate that glycosylation does indeed mask autocleavage sites within the hinge domain of MT1-MMP, and the absence of glycosylation enables selfcleavage of these alternate cleavage sites. Glycosylation has little effect, however, on the total extent of self-proteolysis. The preferential autocleavage site for MT1-MMP has previously been identified to reside between Gly 284 and Gly 285 (32,34,35). Recent evidence indicated that mutation of the preferential autocleavage site Gly 284 /Gly 285 enforced downstream cleavage between Gln 296 and Ser 304 (41). Strikingly, this stretch includes three out of four potentially glycosylated residues (Fig. 1b). In the past, the low sequence homology of hinge domains among MMPs has been exploited for the generation of specific antibodies. Because glycosylation appears to be a common feature of hinge domains (data not shown), our data suggest that the results obtained with such probes need to be interpreted with caution.
The cytoplasmic tail of MT1-MMP has generally been credited as the primary determinant of the trafficking pattern pursued by this protein (20,24,42). The results from metabolic pulse-chase and cell surface biotinylation experiments presented here indicate that the effect of LL 572 on the O-glycosylation pattern of MT1-MMP is likely to be independent of changes in trafficking and steady-state surface expression. Previously, the LLY 573 sequence has been demonstrated to bind the 2 subunit of adaptor protein-2 in a yeast two-hybrid assay (24). Membrane protein interaction with adaptor protein-2 is known to induce internalization via clathrin-coated pits. However, the precise binding motif had not been fully defined, because both tyrosine and dileucine motifs are potential adaptor protein-2 binding sites (43)(44)(45)(46). Our results indicate that the mutation of the dileucine motif LL 572 had no effect on the expression level of MT1-MMP at the cell surface, whereas mutation of the other dileucine motif LL 579 increased the population of MT1-MMP resident at the cell surface (24).
Most notably, a number of studies have linked dileucine motifs to a function in protein folding that is independent from cell surface trafficking. For instance, the dileucine motif at the C terminus of the human multidrug resistance P-glycoprotein (Pgb or ABCB1) has been demonstrated to be essential for the correct folding of this protein (47). Mutation of the dileucine motif caused the expression of an immature protein. Although the underlying mechanism responsible for this effect is still unclear, it appears that the dileucine motif did not function as a plasma membrane targeting signal. Furthermore, dileucine motifs in the C terminus of the vasopressin V 2 receptor and ATP-sensitive potassium channels (K ATP ) have been shown to be required for the exit from the endoplasmic reticulum (48,49). In polarized epithelial cells, such as Madin-Darby canine kidney cells, a dileucine motif has been proven to be necessary and sufficient to target E-cadherin to the basolateral cell surface (50). Mutation of the motif caused aberrant apical expression of E-cadherin. Still, the basic mechanisms that define these effects of dileucine motifs have not yet been delineated. In our case, MT1-MMP with a mutated or interrupted dileucine LL 572 is delivered to the cell surface, but displays a glycosylation pattern that is distinct from that of the wild-type protein. This effect could be attributed, for example, to perturbation of an interaction with a chaperone that depends on the dileucine motif, resulting as a consequence in incorrect protein folding. This might hamper downstream trafficking to or within the Golgi apparatus or protein-protein interactions that ultimately modulate the O-glycosylation pattern.
Differences in the nature and extent of pro-MMP-2 activation and substrate degradation have been extensively described for cytoplasmic tail truncation mutants of MT1-MMP, and have been interpreted in relation to the putative effects of tail mutations on MT1-MMP cell surface trafficking and expression (20,24). Our data suggest that these functional differences may also be attributable in part to the novel influence that the cytoplasmic tail exerts upon the extent and composition of the O-glycosylation of MT1-MMP, as glycosylation has been shown to be a significant determinant of the substrate profile of MT1-MMP (17). Conversely, effects of glycosylation on microdomain trafficking have been discussed, because glycans contain signals that are responsible for the partitioning of glycoproteins into subcellular domains (51,52). Interestingly, truncation of the cytoplasmic tail of MT1-MMP has been shown to induce aberrant and persistent inclusion of this protein into lipid rafts (53). On a functional level, the diversity of MT1-MMP glycosylation subtypes found in different cell lines may reflect the involvement of these proteins in regulating an equally diverse array of biological processes (17).
To the best of our knowledge, a cytoplasmic tail motif has not previously been implicated in the regulation of the O-glycosy- DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51 lation pattern of a protein. Because protein-coupled oligosaccharides are not primary gene products but are instead the product of post-translational modifications, our understanding of the factors that regulate their composition and assembly remains somewhat rudimentary (54,55). Hence, the unique regulation of the glycosylation pattern of MT1-MMP by a dileucine motif in this protein's cytoplasmic tail sheds light on a novel mechanism involved in the O-linked glycosylation and post-translational modification of MT1-MMP in particular and in the maturation of transmembrane proteins in general.