Tracking the Cartoon mouse phenotype: Hemopexin domain–dependent regulation of MT1-MMP pericellular collagenolytic activity

Following ENU mutagenesis, a phenodeviant line was generated, termed the “Cartoon mouse,” that exhibits profound defects in growth and development. Cartoon mice harbor a single S466P point mutation in the MT1-MMP hemopexin domain, a 200-amino acid segment that is thought to play a critical role in regulating MT1-MMP collagenolytic activity. Herein, we demonstrate that the MT1-MMPS466P mutation replicates the phenotypic status of Mt1-mmp–null animals as well as the functional characteristics of MT1-MMP−/− cells. However, rather than a loss-of-function mutation acquired as a consequence of defects in MT1-MMP proteolytic activity, the S466P substitution generates a misfolded, temperature-sensitive mutant that is abnormally retained in the endoplasmic reticulum (ER). By contrast, the WT hemopexin domain does not play a required role in regulating MT1-MMP trafficking, as a hemopexin domain-deletion mutant is successfully mobilized to the cell surface and displays nearly normal collagenolytic activity. Alternatively, when MT1-MMPS466P–expressing cells are cultured at a permissive temperature of 25 °C that depresses misfolding, the mutant successfully traffics from the ER to the trans-Golgi network (ER → trans-Golgi network), where it undergoes processing to its mature form, mobilizes to the cell surface, and expresses type I collagenolytic activity. Together, these analyses define the Cartoon mouse as an unexpected gain-of-abnormal function mutation, wherein the temperature-sensitive mutant phenocopies MT1-MMP−/− mice as a consequence of eliciting a specific ER → trans-Golgi network trafficking defect.

Type I collagen, the dominant extracellular protein found in mammals, undergoes extensive proteolytic remodeling in the course of growth and development as well as multiple disease states, ranging from inflammation to cancer (1)(2)(3). Although the mammalian proteome includes more than 500 distinct enzymes, only a small subset of proteinases display type I collagenolytic activity (1)(2)(3). In mice, the ability to cleave native type I collagen within its triple-helical domain is restricted largely to the secreted matrix metalloproteinases (MMPs), 3 MMP-8, MMP-13, and possibly MMP-2; the membraneanchored matrix metalloproteinases, MT1-MMP and MT2-MMP; and the cysteine proteinase, cathepsin K (1-3). Nevertheless, whereas the expression of each of these proteinases has been targeted in mouse models (4 -8), only MT1-MMPnull animals exhibit profound defects in type I collagen remodeling in vivo that are associated with early morbidity and mortality (9,10).
Not unexpectedly, the unique proteolytic functions assigned to MT1-MMP have catalyzed comprehensive efforts to delineate the critical structural determinants that define its ability to operate as the dominant pericellular type I collagenase operative in mammalian systems (3,11). Currently, the membraneanchored proteinase is divided structurally into at least six discrete regions: an N-terminal prodomain, a catalytic domain, a short linker sequence followed by the hemopexin domain, a single-pass transmembrane region, and a short cytosolic tail (3,11). Independent of the obvious functional importance of its catalytic domain, increasing interest has focused on the ability of the MT1-MMP hemopexin domain to modulate proteolytic activity (3,11). Using a variety of structure/function-designed approaches, the MT1-MMP hemopexin domain has been reported to control (i) the trafficking of the enzyme from the trans-Golgi network to the cell surface, (ii) MT1-MMP association with cell-surface transmembrane protein binding partners, (iii) MT1-MMP homodimerization with consequent effects on proteolytic activity, and (iv) MT1-MMP-type I collagen binding interactions (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30). Alternatively, more subtle roles for the hemopexin domain in regulating MT1-MMP activity have also been identified (31)(32)(33), thereby complicating efforts to assign a definitive role to its functional activity.
In an ENU mutagenesis screen designed to identify immunological phenodeviants, Beutler and colleagues (http:// mutagenetix.utsouthwestern.edu/, allele: Cartoon) 4 recently identified a mutation giving rise to mice with craniofacial defects, stunted growth, infertility, and a markedly shortened lifespan, a phenotype similar to that described for MT1-MMPnull mice (http://mutagenetix.utsouthwestern.edu/, allele: Cartoon). Genome sequencing subsequently identified a single Ser 446 3 Pro substitution in the MT1-MMP hemopexin domain, a finding seemingly consistent with the myriad functions previously assigned to this domain. However, the impact of this point mutation on MT1-MMP function remains undefined, and the mechanism by which this mutant hemopexin domain regulates MT1-MMP activity in intact cell systems has not been established. To more specifically characterize the impact of the Cartoon mouse mutation on MT1-MMP function, we have compared and contrasted the phenotype of WT, MT1-MMP Ϫ/Ϫ , and Cartoon mice and defined the effect of the Ser 446 3 Pro mutation on MT1-MMP activity in intact cell systems. Herein, we report that whereas the Cartoon mouse mutation recapitulates key in vivo features of the Mt1-mmpnull phenotype, changes in MT1-MMP function do not arise as a consequence of a direct loss of proteolytic activity. Rather, the S466P substitution effects a loss-of-function phenotype by conferring the hemopexin domain with new properties wherein the Cartoon mutant is retained in the endoplasmic reticulum (ER) and fails to traffic to the cell surface. By contrast, by either deleting the entire hemopexin domain or culturing Cartoon mutant-expressing cells at lower temperatures that are permissive for ER 3 trans-Golgi network trafficking of the misfolded MT1-MMP S466P mutant, both the hemopexin domain-deleted MT1-MMP and Cartoon proteinase are displayed on the cell surface, where they retain collagenolytic activity and function.

Cartoon mice recapitulate an Mt1-mmp ؊/؊ phenotype
Cartoon mice display an overall phenotype similar to Mt1mmp Ϫ/Ϫ mice with stunted growth and craniofacial anomalies characterized by a shortened head and snout as well as dwarfism ( Fig. 1A) (http://mutagenetix.utsouthwestern.edu/, allele: Cartoon) 4 (9,10). Indeed, Cartoon mice, like their MT1-MMP-null counterparts, also exhibit an osteopenic phenotype, increased cartilage formation, and the complete absence of secondary ossification centers (Fig. 1B). Similarly, a profound loss of dermal adipose tissue, a characteristic finding in Mt1mmp Ϫ/Ϫ mice (34), is likewise observed in the Cartoon mouse mutants (Fig. 1C). Hence, the Cartoon mouse displays phenotypic changes similar, if not identical, to those observed in MT1-MMP-null mice.
In normal fibroblasts, MT1-MMP is mobilized to the cell surface where it proteolytically remodels pericellular collagen and promotes tissue-invasive activity (35)(36)(37). By contrast, MT1-MMP-null fibroblasts display a complete inability to either degrade or invade type I collagen-rich extracellular matrix barriers (35)(36)(37). As such, primary dermal fibroblasts were isolated from WT, Cartoon, and Mt1-mmp Ϫ/Ϫ mice in order to characterize their respective functional activities. Dermal fibroblasts recovered from Cartoon mice display normal morphology and cytoskeletal organization that are indistinguishable from MT1-MMP ϩ/ϩ or MT1-MMP Ϫ/Ϫ fibroblasts ( Fig. 2A). Further, relative to the complete deficiency of the proteinase in MT1-MMP Ϫ/Ϫ cells, MT1-MMP protein levels in cell lysates are comparable between WT and Cartoon mouse fibroblasts (Fig. 2B). However, when cultured atop a bed of fluorescently labeled type I collagen fibrils, only WT, and not Cartoon or MT1-MMP Ϫ/Ϫ fibroblasts, display a collagenolytic phenotype (Fig. 2, B and C). Likewise, as collagenolytic activity is a prerequisite for supporting invasive activity through native type I collagen hydrogels (35)(36)(37), Cartoon fibroblasts, like MT1-MMP Ϫ/Ϫ fibroblasts, are unable to mount an invasive response (Fig. 2, D and E). Hence, Cartoon mouse fibroblasts display a loss of pericellular collagenolytic activity that phenocopies the functional properties of MT1-MMP-null cells.

Characterization of MT1-MMP S466P activity
To define the role of the hemopexin point mutation in regulating MT1-MMP activity, mouse WT or S466P mutant expression vectors were constructed and transfected into COS-1 cells that do not express detectable levels of the endogenous protein (38). Following transfection, and as observed in Cartoon mouse fibroblasts, both the WT and mutant protein (i.e. MT1-MMP S466P ) are expressed at comparable levels in cell lysates (Fig. 3A). However, in apparent agreement with earlier studies reporting a required role for the MT1-MMP hemopexin domain in activating its downstream proteolytic target, pro-MMP-2 (12,21,23), only WT MT1-MMP-transfected cells effectively processed the MMP-2 zymogen to its active form in the extracellular compartment as assessed by gelatin zymography (Fig. 3A). As the highly specific anti-MT1-MMP mAb used here is directed toward an epitope localized near the catalytic domain, a 46-kDa autocatalytic degradation product that remains anchored to the cell surface, but no longer retains the catalytic domain, cannot be detected (32,39). As such, an HAepitope tag was inserted into the juxtamembrane region to allow tracking of MT1-MMP turnover (32,39). Interestingly, whereas the WT proteinase generated the autocatalytic degradation product, the ϳ46-kDa fragment is not detected in MT1-MMP S466P -transfected COS cells (Fig. 3B). These results are not confined to mouse MT1-MMP, as similar findings are found when COS-1 cells are transfected with a human MT1-MMP S466P mutant carrying an HA-epitope tag (Fig. 3C). As expected, the generation of the 46-kDa degradation product derived from WT MT1-MMP is blocked in the presence of the pan-specific MMP inhibitor, BB-94 (35), whereas mutant MT1-MMP does not undergo further processing in the absence or presence of BB-94 (Fig. 3D). A faint ϳ40 kDa band is often detected in MT1-MMP S466P -transfected COS-1 cells in either the absence or presence of BB-94 (but not in Cartoon fibro-blasts) and appears to reflect a degradation product of the overexpressed mutant protein. In any case, the inabilities of MT1-MMP S466P to activate the MMP-2 zymogen or undergo autocatalytic degradation are both consistent with a presumed loss of hemopexin-dependent MT1-MMP activity. Indeed, whereas COS-1 cells transfected with WT MT1-MMP readily degrade subjacent collagen, MT1-MMP S466P mutant-transfected cells are predictably devoid of detectable collagenolytic activity (Fig. 3, E and F).

Cell-surface trafficking of Cartoon mouse MT1-MMP
WhereasMT1-MMP-dependentdefectsinpro-MMP-2processing and collagenolytic activity are consistent with activities assigned previously to the hemopexin domain, MT1-MMP must undergo trafficking to the cell surface to function as a pericellular proteinase (3,11). To directly assess MT1-MMP routing to the cell surface, COS cells were transfected with either WT or the MT1-MMP S466P mutant, and surface proteins were biotinylated and then captured by streptavidin-affinity chromatography before Western blotting. As expected, in cells expressing WT MT1-MMP, both active and degraded forms of the proteinase are detected on the cell surface with the generation of the 46-kDa product blocked completely in the presence of BB-94 ( Fig. 4A) (32). In marked contrast, the MT1-MMP S466P mutant cannot be detected on the cell surface in either the absence or presence of BB-94 by surface biotinylation or immunostaining (Fig. 4, A and B). Confirming this result, Cartoon mouse fibroblasts are likewise unable to traffic mutant MT1-MMP to the cell surface (Fig. 4C). Given that the absence of MT1-MMP S466P on the cell surface might be alternatively explained by an accelerated rate of internalization, lysosomal routing, and degradation, we stabilized membrane expression

Hemopexin domain control of MT1-MMP
levels by deleting the MT1-MMP cytosolic tail that contains key internalization signals (13,32,40). As expected, when WT MT1-MMP is expressed as a tail-deleted mutant (i.e. MT1-MMP⌬CT), cell-surface expression is marginally increased in tandem with increased MMP-2 activation (Fig. 4, D and E). By contrast, deleting the tail of the MT1-MMP S466P mutant yields only barely detectable levels of the enzyme at the cell surface (Fig. 4, D and E). Consistent with these findings, tail-deleted MT1-MMP displays a trend toward increased collagenolytic activity relative to the WT proteinase, whereas MT1⌬CT S466Ptransfected COS cells remain unable to degrade subjacent collagen fibrils to a detectable degree (Fig. 4, F and G).

Hemopexin domain control of MT1-MMP Dysregulated trafficking of Cartoon mouse MT1-MMP
Recent studies suggest that the MT1-MMP hemopexin domain plays a regulatory role in sorting the proenzyme to the trans-Golgi network, where the enzyme undergoes proprotein convertase-dependent processing to its proteolytically active form before its final routing to the cell surface (22). However,

Hemopexin domain control of MT1-MMP
when COS cells are transfected with an MT1-MMP hemopexin deletion-mutant domain (i.e. MT1⌬PEX), the proteinase maintains its ability to traffic to the cell surface, where it remains catalytically active, as reflected in its ability to (i) undergo autocatalytic degradation, (ii) activate pro-MMP-2, and (iii) degrade subjacent collagen fibrils (Fig. 5, A-C). Although the level of collagenolytic activity displayed by MT1⌬PEX is modestly depressed relative to that of the WT enzyme (32), the hemopexin-deleted mutant maintains significant activity relative to the Cartoon mutant (Fig. 5C).
Independent of its ability to support tissue-invasive activity, MT1-MMP has also been shown to regulate bone marrow-derived mesenchymal stem cell (MSC) differentiation programs (41). When suspended in 3D type I collagen hydrogels, MT1-  (Fig. 5, D and E). Similarly, following transduction of MT1-MMP Ϫ/Ϫ MSCs with an MT1-MMP S466P expression vector, 3D collagen-embedded MSCs remain locked in an adipogenic lineage commitment program (Fig. 5, D and E). By contrast, when MT1-MMP Ϫ/Ϫ MSCs are transduced with the WT proteinase or MT1⌬PEX, adipogenesis is inhibited, whereas lineage commitment is redirected toward osteoblastogenesis (Fig.  5, D-F). Although MT1⌬PEX is less active than WT MT1-MMP in terms of either collagenolytic or MSC differentiationinducing activity (Fig. 5, C-F), these results demonstrate that the MT1-MMP hemopexin domain does not play a required role in regulating the enzyme's proteolytic or functional activity.
In the absence of direct evidence supporting a required role for the hemopexin domain in regulating MT1-MMP processing, trafficking, or activity, we next considered the potential impact of the S466P substitution on MT1-MMP structure by first interrogating the recently solved crystal structure of the MT1-MMP hemopexin domain (23). In WT MT1-MMP, Ser 466 is characterized as a buried moiety within the mid-region of a ␤-strand of a 4-fold propeller structure (Fig. 6). This serine residue is positioned in a closely packed environment with other buried residues, and the insertion of a bulky proline residue would be predicted to disrupt protein secondary structure by inhibiting the ability of its backbone to adopt a ␤-strand conformation while creating a potential steric clash with Ala 417 in a neighboring strand (Fig. 6) (42). As ␤-strand structural changes can impact trafficking of secreted proteins (43)(44)(45)(46), we compared the intracellular localization of MT1-MMP and MT1-MMP S466P . As expected, WT MT1-MMP co-localized with markers for the ER and cis-Golgi compartments (i.e. calnexin and GM130, respectively) (Fig. 7, A and B). By contrast, MT1-MMP S466P is confined almost entirely to the ER (Fig. 7, A  and B). Further, whereas WT MT1-MMP internalized from the cell surface co-localizes with the early endosomal marker, EEA1, MT1-MMP S466P cannot be detected in this compartment (Fig. 7, A and B).
As the processing of the MT1-MMP zymogen to its active form normally occurs in the trans-Golgi network (39,47), the localization of MT1-MMP S466P to the ER raises the possibility that the proteinase remains locked in its zymogen form as a proenzyme. To monitor MT1-MMP processing in situ, a FLAG sequence was inserted into both WT MT1-MMP and MT1-MMP S466P downstream of the RXKR 111 proprotein convertaserecognition sequence (39). Using this approach, as WT MT1-MMP undergoes processing, the FLAG sequence is positioned at the newly exposed N terminus, where it can be recognized specifically by the FLAG M1 mAb. Indeed, whereas the M1-positive product of processed MT1-MMP is readily detected in WT-transfected cells, the MT1-MMP S466P -expressing cells fail to expose the FLAG N terminus, confirming a failure to undergo proprotein convertase-dependent processing (Fig.  7C). Although confinement of pro-MT1-MMP S466P to the ER might be predicted to trigger an unfolded protein stress response (48,49), no significant changes in p-ERK, p-eIF2␣, or

MT1-MMP S466P is a temperature-sensitive mutant that retains pericellular collagenolytic activity
The inability of MT1-MMP S466P to traffic to the cell surface precludes efforts to assess its proteolytic activity as a membrane-anchored proteinase. Regardless of whether the intracellular confinement of the mutant proteinase occurs as a consequence of the generation or exposure of a cryptic ER retention signal, the trafficking of ER-retained, misfolded proteins can sometimes be rescued at permissive temperatures, thereby allowing mutant proteins to traffic to the cell surface (43,45,50). As such, MT1-MMP S466P -transfected COS cells were either incubated under standard conditions at 37°C or, alternatively, cultured at 25°C for 12 h before returning the cells to 37°C. As expected, at 37°C, MT1-MMP S466P -transfected COS cells fail to traffic the mutant to the GM130 ϩ cis-Golgi network (Fig. 8A). In marked contrast, the 25 3 37°C switch allows the mutant proteinase to bypass the ER block, traffic to the cis-Golgi network, and then move to the cell surface (Fig. 8, A-C). Similarly, despite lower levels of expression, Cartoon fibroblasts likewise traffic the endogenous MT1-MMP mutant to the cell surface (Fig. 8B). As such, we next sought to assess the ability of the membrane-anchored MT1-MMP S466P mutant to express pericellular proteolytic activity. To this end, transfected COS cells were again cultured under nonpermissive conditions at 37°C or, alternatively, allowed to undergo the 25 3 37°C switch, before plating the cells atop fluorescently labeled gelatin-or type I collagen-coated surfaces at 37°C. Whereas MT1-MMP S466P -transfectedCOScellspreculturedunderstandard 37°C conditions predictably failed to display gelatinolytic or collagenolytic activity, the 25 3 37°C switch allowed mutant-expressing cells to display both proteolytic activities , and proteinase distribution in the ER, cis-Golgi, and early endosome compartments was assessed by their co-localization with calnexin, GM130, and EEA1, respectively, in fixed permeabilized cells. COS-1 cells were then imaged by confocal laser microscopy. Whereas WT MT1-MMP is found in each of the three compartments, MT1-MMP S466P is confined to the ER. C, COS-1 cells were transiently transfected with WT MT1-MMP or MT1MMP S466P expression vectors that contain a FLAG sequence (DYKDDDK) inserted downstream of the MT1-MMP RRKR 111 proprotein convertase recognition sequence. Cells were then lysed, and MT1-MMP expression was assessed by Western blotting using either an anti-HA mAb or a FLAG-M1 antibody that only recognizes the FLAG sequence when displayed at the N terminus. Under these conditions, the WT MT1-MMP construct containing both FLAG and HA inserts can be resolved into a closely migrating doublet representative of pro-MT1-MMP and active MT1-MMP as well as its autocatalytic degradation product (39). By contrast, MT1-MMP S466P is primarily expressed as the proenzyme alone. Following immunoblotting with the FLAG-M1 antibody, the processing of WT MT1-MMP to its prodomaindeleted mature/active form is detected. D, COS-1 cells were transfected with MT1-MMP or MT1-MMP S466P and lysed 24 h later, and levels of p-ERK/total ERK, p-elf2␣/total elf2␣, and p-JNK/total JNK as well as ␤-actin were assessed by Western blotting. Bars, 20 m. (Fig. 8D). Hence, MT1-MMP S466P elicits a loss-of-function mutation by transforming MT1-MMP into an ER-retained, temperature-sensitive mutant that retains collagenolytic activ-ity when conformation-specific defects in cell-surface trafficking are circumvented.
As we now describe, Cartoon mice phenocopy many of characteristics assigned to MT1-MMP-null mice, including major defects in bone formation, an inability to form secondary ossification zones, and the disrupted development of peripheral white fat depots (9,10,34,41). Furthermore, in apparent agreement with earlier studies stressing a required functional role for the MT1-MMP hemopexin domain (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23), the S466P mutant proved incapable of activating pro-MMP-2 or expressing type I collagenolytic activity following expression in COS cells. However, rather than defining a defect in MT1-MMP proteolytic activity per se, further studies demonstrate that the S466P mutation interferes with MT1-MMP trafficking to the cell surface. Whereas this outcome is consistent with a report that the hemopexin domain can control MT1-MMP exocytosis (22), the former experiments were performed by introducing domain swaps wherein the MT1-MMP hemopexin domain was replaced with the hemopexin domain of MT4-MMP, a glycosylphosphatidylinositol-anchored MMP whose structure is distinct from that of type I transmembrane MT-MMPs (11). Here, we show that deleting the entire hemopexin domain of MT1-MMP does not interfere with MT1-MMP trafficking or function at the cell surface, a finding consistent with earlier work from our laboratory as well as others where the MT1-MMP hemopexin domain was replaced with that of either the MT3-MMP or MMP-2 hemopexin domain without affecting cell-surface trafficking (19,39). As such, the earlier results reported with the

Hemopexin domain control of MT1-MMP
MT4-MMP hemopexin domain swap most likely arose as a consequence of unanticipated domain clashes. Indeed, as opposed to the Cartoon mouse mutation, the MT1-MMP⌬PEX deletion mutant retains not only its ability to activate pro-MMP-2, but also to support collagenolytic and invasive activity as well as more complex functions, including MSC differentiation. Nevertheless, these results should not be misconstrued to suggest that the hemopexin domain is without function, at least in terms of tuning proteolytic activity. Using transmembranedeleted mutants, Zhao et al. (31) reported that the ability of secreted, WT MT1-MMP to hydrolyze triple-helical substrates (as defined by k cat /K m values) is decreased when the hemopexin domain is deleted, but only by 3-fold. Nevertheless, it should be stressed that although this study is consistent with our findings, these authors did not examine the ability of the mutant to degrade native collagen as a membrane-anchored proteinase in an intact cell system (31). Taken together, these studies highlight the fact that the hemopexin domain more likely serves a modulatory, as opposed to necessary, role in defining MT1-MMP functional activity.
Although the presence of the MT1-MMP hemopexin domain is not required for its export to the cell surface, we found that the single S466P point mutation precluded the export of Cartoon MT1-MMP from the ER to the trans-Golgi apparatus, where the proenzyme normally undergoes proprotein convertase-dependent processing to its active form (39,47). The C-terminal hemopexin domain of MT1-MMP is composed of a sheet of four anti-parallel ␤-stands that form a four-bladed propeller-like structure (23). As the insertion of proline residues into ␤-sheet strands precludes normal folding (42,46), the associated conformational changes are not permissive for ER 3 Golgi trafficking. Interestingly, a number of human genetic disorders that are distinguished by defects in intracellular sorting and trafficking are also characterized by proline substitutions in ␤-sheet structures (43,45,69). Whereas the mutant MT1-MMP protein does not appear to trigger an unfolded protein response, we note that ER retention is not necessarily associated with increased rates of degradation (70). In this regard, a recent report has concluded that bacterially expressed recombinant MT1-MMP S466P does not display major changes in structural conformation (33), but it is apparent from our studies that significant alterations in protein folding do occur under physiologic conditions. The further characterization of MT1-MMP S466P as a temperature-sensitive mutant also allows us to conclude that the hemopexin domain point mutation does not, in and of itself, interfere with proprotein convertase-dependent processing or trafficking to the cell surface or preclude the expression of type I collagenolytic activity. Direct kinetic analyses of the collagenolytic activity of WT versus mutant MT1-MMP cannot be readily determined, but we note that a recombinant MT1-MMP S466P transmembrane deletion mutant has been reported to retain full enzymatic activity against synthetic triple-helical substrates (33).
Finally, by establishing Cartoon fibroblast cultures, we confirmed that these cells share each of the functional defects observed in our model COS cell system. Indeed, the inability of Cartoon fibroblasts to degrade or invade type I collagen hydrogels is identical to that observed in MT1-MMP-null fibro-blasts. Nevertheless, it is interesting to note that Cartoon mice live longer than MT1-MMP Ϫ/Ϫ mice in identical C57BL/6J backgrounds (i.e. whereas MT1-MMP Ϫ/Ϫ mice rarely live beyond 3 weeks, Cartoon mice display a modest increase in longevity with partial morbidity observed by 3.5 weeks with no mice surviving beyond 6 weeks). Whereas it remains possible that small amounts of MT1-MMP S466P are folded correctly and can traffic to the cell surface under select conditions in vivo, it is also noteworthy that MT1-MMP can exert protease-independent functions that potentially affect signal transduction cascades as well as transcriptional programs (59,67,68). Further studies will be needed to resolve these issues, but the findings described herein characterize the Cartoon mouse as an unexpected "gain-of-abnormal-function" mutation that elicits a specific, but reversible, defect in MT1-MMP trafficking.

Experimental procedures
Cell culture and mouse lines COS-1 cells (ATCC) were routinely maintained in DMEM supplemented with 10% heat-inactivated fetal bovine serum (FBS) (both from Life Technologies) and a 1% penicillin-streptomycin solution (Invitrogen). All cells were maintained in a 5% CO 2 , 95% air atmosphere at 37°C unless indicated otherwise. Primary mouse fibroblasts were isolated from dorsal dermal explants of 2-4-week-old male WT, Mt1-mmp Ϫ/Ϫ , or Cartoon mice (C57BL/6J background) as described (35,36) and cultured in DMEM supplemented with 10% heat-inactivated FBS and 1% penicillin-streptomycin-fungizone solution (Gibco). Bone marrow-derived mesenchymal stem cells were isolated from WT and Mt1-mmp Ϫ/Ϫ mice and cultured in 3D type I collagen hydrogels as described (41). In brief, bone marrow cells were isolated from mouse hind limbs and cultured in DMEM supplemented with 10% heat-inactivated FBS. Adherent colonies were sorted by flow cytometry with antibodies directed against Sca-1, CD29, CD45, and CD116, and the harvested cells were cultured and used for up to five passages as described (72). For 3D culture, 5 ϫ 10 5 cells/ml were embedded in 2.2 mg/ml rat tail-derived type I collagen (36) and cultured in Transwell dishes with 0.4-m pore size. Where indicated, cells were cultured in the presence or absence of 10 M BB-94 (Abcam). All cell lines and cultured cells were mycoplasma-negative. All mouse work was performed with institutional animal care and use committee approval and in accordance with protocols approved by the University of Michigan Institutional Animal Care and Use Committee.

Construction of expression plasmids and transfection
Subcloning of HA-tagged mouse or human MT1-MMP cDNA as well as cDNAs encoding mutant human MT1-MMP with a cytosolic tail deletion (Met 1 -Arg 563 ; MT1-⌬CYT) or a hemopexin domain deletion (MT1⌬PEX; Cys 318 -Gly 535 deleted) was performed as described previously (32). A FLAGtagged variant of human MT1-MMP cDNA was generated wherein the epitope tag was inserted directly downstream of the C terminus of the proprotein recognition motif at Arg 111 by using overlapping primer sets containing the FLAG sequence: forward, 5Ј-TACCCATACGATGTTCCAGATTACGCTGA-GGGGACTGAGGAG-3Ј; reverse, 5Ј-AGCGTAATCTGGAA-

Hemopexin domain control of MT1-MMP
CATCGTATGGGTAATCGGGCCGGCCCCC-3Ј (39). HAand FLAG-tagged versions of human and mouse MT1-MMP Cartoon mutation (Ser 466 3 Pro 466 ) were generated using sitedirected mutagenesis produced with PCR primers: forward, CCCAGAGGGCCATTCATGG-3Ј; reverse, TGCCCATGA-ATGGCCCTCTG-3Ј. Each mutant was sequenced to verify the generation of the desired mutation. COS-1 cells were transiently transfected with either a control vector (pCR3.1; Invitrogen) or with the indicated expression vectors using FUGENE6 (Roche Applied Science) according to the manufacturer's instructions. In selected experiments, COS-1 cells were co-transfected with a Lifeact expression vector (Addgene) to visualize F-actin.

Gelatin zymography
Pro-MMP-2 was transiently expressed in COS-1 cells, and the conditioned medium was harvested after 18 h. Aliquots of the conditioned medium containing recombinant pro-MMP-2 were then incubated with COS-1 cells overexpressing each of the indicated expression vectors for 24 h. Aliquots of conditioned medium were then subjected to gelatin zymography after a 12-h incubation period (32). Gelatinolytic activity is linear with incubation time over this period (data not shown).
For immunofluorescence, cells were fixed in 4% paraformaldehyde, washed in PBS, and permeabilized with Triton X-100 (Sigma-Aldrich). Following blocking with 3% goat serum and 1% BSA in PBS, samples were incubated with primary antibodies overnight at 4°C. Alexa Fluor 488 -and 594 -conjugated secondary antibodies (Molecular Probes) were used for protein detection.

Cell-surface biotin labeling
Cell-surface biotinylation was performed as described (32). Briefly, cells were rinsed twice with PBS and incubated with 0.5 mg/ml Sulfo-NHS-SS-Biotin (Pierce) in PBS at 4°C for 1 h. The reaction was terminated by washing with 150 mM glycine/TBS, and cell lysates were recovered for pulldown with streptavidin-agarose (Pierce). The captured material was resolved by SDS-PAGE, and surface-labeled MT1-MMP was assessed by Western blotting.

Type I collagen degradation
Acid-extracted type I collagen was prepared from rat tail tendons and dissolved in 0.2% acetic acid to a final concentration of 2.7 mg/ml. To generate matrix-coated surfaces, collagen was mixed with 10ϫ Eagle's minimum essential media and 0.34 N NaOH in an 8:1:1 ratio with 25 mM Hepes at 4°C, and 100 l of the mixture was uniformly spread over the surface of 2-cm 2 Lab-Tek II chamber slides (Nalge Nunc International, Naperville, IL). Fibrillogenesis was induced by incubating the collagen-coated slides for 45 min at 37°C, and collagen films were labeled with Alexa Fluor 488 (Molecular Probes). Post-fibrillogenesis labeling does not alter the sensitivity of type I collagen to collagenolytic attack relative to unlabeled type I collagen (data not shown). Fibroblasts or COS-1 cells (0.5 ϫ 10 4 ) were seeded at low density atop collagen or gelatin films and incubated for 3 days in DMEM, 10% FBS at 37°C. Degradation rates increase in linear fashion between 1 and 3 days under these culture conditions. Fluorescence images were captured by laser confocal microscopy. Collagen degradation was quantified by the area of zones without fluorescent signal with results expressed as the mean Ϯ S.E. of three experiments.

Microscopy
Confocal imaging of collagen degradation and cellular immunofluorescence was performed with a spinning disc Nikon Eclipse Ti confocal microscope using a ϫ20 objective lens, numerical aperture 0.75, or a ϫ100 objective lens, numerical aperture 1.45. Images of Alexa Fluor 488 and Alexa Fluor 594 signals were captured at 25°C with a Yokogawa CSU-W1 camera using Micromanager MM Studio (version 1.4.23) software and ImageJ for image processing. Cellular immunofluorescence signals were also imaged using a Leica DM IRB spinning-disc confocal microscope with a ϫ63 objective lens, numerical aperture 1.4, and images were captured with a PerkinElmer Ultra-View Vox system camera using Volocity version 4.0 software. Equal photomultiplier tube intensity and gain settings were used in acquiring images. All other fluorescence and brightfield images were captured using a Spot digital camera (Diagnostic Instruments, Inc.) through a Leica upright microscope. Image-processing software (Photoshop version 7, Adobe) was used to overlay images and to enhance equally image color and clarity.

Structural analysis
The Ser 466 3 Pro mutation was modeled into the structure of the MT1-MMP hemopexin domain using the COOT program and then subjected to structural idealization using the REFMAC5 program as described (71). Contacts were analyzed using COOT.

Statistical analysis
Statistical analyses were performed using unpaired Student's t test. All experiments were performed three or more times.