Extracellular Collagenases and the Endocytic Receptor, Urokinase Plasminogen Activator Receptor-associated Protein/Endo180, Cooperate in Fibroblast-mediated Collagen Degradation*

The collagens of the extracellular matrix are the most abundant structural proteins in the mammalian body. In tissue remodeling and in the invasive growth of malignant tumors, collagens constitute an important barrier, and consequently, the turnover of collagen is a rate-limiting process in these events. A recently discovered turnover route with importance for tumor growth involves intracellular collagen degradation and is governed by the collagen receptor, urokinase plasminogen activator receptor-associated protein (uPARAP or Endo180). The interplay between this mechanism and extracellular collagenolysis is not known. In this report, we demonstrate the existence of a new, composite collagen breakdown pathway. Thus, fibroblast-mediated collagen degradation proceeds preferentially as a sequential mechanism in which extracellular collagenolysis is followed by uPARAP/Endo180-mediated endocytosis of large collagen fragments. First, we show that collagen that has been pre-cleaved by a mammalian collagenase is taken up much more efficiently than intact, native collagen by uPARAP/Endo180-positive cells. Second, we demonstrate that this preference is governed by the acquisition of a gelatin-like structure by the collagen, occurring upon collagenase-mediated cleavage under native conditions. Third, we demonstrate that the growth of uPARAP/Endo180-deficient fibroblasts on a native collagen matrix leads to substantial extracellular accumulation of well defined collagen fragments, whereas, wild-type fibroblasts possess the ability to direct an organized and complete degradation sequence comprising both the initial cleavage, the endocytic uptake, and the intracellular breakdown of collagen.

membrane and the fibrillar matrix collagens all incorporate into dense, insoluble protein networks that form a critical barrier against processes of cell migration such as those connected to tissue remodeling, including the invasive growth of cancer. Consequently, the degradation of these matrices is one of the rate-limiting steps in cancer invasion (1).
Importantly, however, collagen may also be degraded through an intracellular turnover pathway (11,14). Recent studies have shown that an endocytic route of collagen breakdown, mediated by the collagen internalization receptor, uPARAP/Endo180 (15)(16)(17) (hereafter designated uPARAP), is a rate-limiting factor in collagenolysis and invasive growth of breast tumors in mice (18). This new observation underscores the importance of understanding the function of uPARAP and the interplay between cellular and extracellular events in collagen and matrix turnover.
uPARAP is a type-1 transmembrane protein and a member of the mannose receptor protein family of constitutive endocytosis receptors (reviewed in Refs. 19 and 20). The extracellular portion includes a cysteine-rich domain, a fibronectin type-II domain, and eight consecutive C-type lectin-like domains (20). The receptor binds directly to various types of collagen through interactions governed, wholly or in part, by the fibronectin type-II domain (15,(21)(22)(23). uPARAP also has carbohydrate binding activity, mediated through the second C-type lectinlike domain (24), and on certain cell types it interacts with the urokinase plasminogen activator receptor (15,25) in a process implicated in signal transduction in connection with cell migration (26,27).
The molecular interaction between uPARAP and collagen is reflected in part in a decreased initial adhesion of uPARAPdeficient cells on a collagen matrix, but a much more striking effect is noted on the ability of fibroblasts to internalize solubilized collagen, which is lost completely upon gene targeting of uPARAP (17,23,28). uPARAP has also been shown to be responsible for the uptake of collagen in hepatic stellate cells (29). Following endocytosis from clathrin-coated pits (30,31), uPARAP is routed to early endosomes and later recycled to the cell surface, whereas, the initially bound collagen is routed to lysosomes for degradation (23,28). The latter process is primarily undertaken by cysteine proteases (11,32).
Consistent with a role in matrix turnover, the expression of uPARAP is clearly associated with sites of tissue remodeling. Thus, the receptor is expressed in osteoblasts of developing bone (33) and in myofibroblasts and certain other cell types in wound healing tissue (34). In connection with cancer, expression is observed in the myoepithelium of mammary carcinoma in situ, in the myofibroblasts of invasive mammary carcinoma, and in fibroblast-like cells in oral cancer (35,36).
Little is known, however, about the interplay between the intracellular and extracellular processes that govern the breakdown of extracellular matrix components. In this work, we show that fibroblasts, a dominant cell type in collagen turnover, direct an organized degradation pathway that combines extracellular cleavage and uPARAP-mediated endocytosis. These observations assign a new function to uPARAP as an efficient endocytic receptor for early collagen cleavage products.

EXPERIMENTAL PROCEDURES
Proteins and Cultured Cells-Skin fibroblasts from newborn homozygous uPARAP-deficient mice and littermate control wild-type mice were isolated and cultured as published previously (17). Skin fibroblasts from mice homozygously deficient for MMP-2 (37), MMP-13 (38), or MMP-14 (7) were isolated and cultured in the same manner, in each case accompanied by wild-type control cultures obtained from littermate control mice. Recombinant, soluble uPARAP fusion protein (suPARAP), comprising the extracellular portion (Met 1 -Ser 1402 ) of the human uPARAP sequence fused to a purification tag (the third domain of the urokinase receptor), was expressed in Drosophila Schneider S2 cells and purified as described (39). The monoclonal mouse anti-uPARAP antibody 2.h.9:F12 was raised against purified suPARAP and produced as described previously (36). The following proteins were purchased from the commercial sources indicated: native trypsin-resistant collagen type-I from rat tail (BD Biosciences), FITC-collagen type I from MD Biosciences (Zürich, Switzerland), cysteine protease inhibitor E-64d and recombinant, human pro-MMP-13 (Calbiochem, San Diego, CA), 4-aminophenylmercuric acetate (APMA) (Sigma-Aldrich), interleukin 1␤, and tumor necrosis factor-␣ (Peprotech, Rocky Hill, NJ), and polyclonal rabbit antibody against collagen type-I (Rockland, Gilbertsville, PA).
Cleavage of Collagen under Non-denaturing Conditions-For activation of pro-MMP-13, the pro-enzyme was diluted to 50 g/ml in reaction buffer (100 mM Tris-HCl, 10 mM CaCl 2 , 1 M ZnCl 2 , 100 mM NaCl, pH 7.0). APMA was added from a freshly prepared 20 mM stock solution in 0.1 M NaOH to a final concentration of 1 mM, after which the sample was incubated for 30 min at 37°C. For the removal of APMA after the activation step, the sample was subjected to four rounds of buffer exchange, each cycle including centrifugation through an Amicon Ultrafree-MC 10,000 NMWL Filter Unit, to 10% of the starting volume and replenishment of reaction buffer to 100% volume. For cleavage of collagen, collagen was diluted to 250 g/ml in reaction buffer, after which the freshly activated MMP-13 was added to a final concentration of 8.75 g/ml (ϳ175 nM). The samples were incubated at 25°C for 24 h.
Internalization Studies with Collagen and Collagen Fragments-Labeling of proteins with 125 I and radioligand internalization assays with cultured newborn mouse fibroblasts were performed as described before (17) and included incubation of cells with radioligand in 24-well tissue culture plates for 3 h at 37°C, followed by isolation of the intracellular fraction and quantification of the internalized radioactivity. Before the assay, the protein concentration and the specific radioactivity of each of the radiolabeled protein preparations was estimated by SDS-PAGE and silver staining, by careful comparison with a narrow dilution series of a known standard of the same protein preparation in the non-labeled form, run on the same gel. 30 ng of labeled intact or cleaved collagen was added to each well with near-confluent cells in a volume of 300 l. In some experiments, the labeled collagen ligands were preincubated at elevated temperature before the assay, as indicated. Incubation of cultured fibroblasts with FITC-labeled intact and cleaved collagen-I (15 g/ml) followed by confocal fluorescence microscopy with a Zeiss LSM 510 microscope was performed as described previously for fluorescence-labeled, intact collagen-IV (28), using an incubation period of 22 h in the presence of 20 M E-64d.
Surface Plasmon Resonance Analysis-Interaction analyses with intact and cleaved collagen and immobilized suPARAP were done using a BIAcore2000 instrument (BIAcore, Uppsala, Sweden). Before interaction analysis, the monoclonal antibody 2.h.9:F12, against uPARAP (36), was coupled to the surface of a CM5-type BIAcore chip by amine-directed coupling, following the instructions provided by the manufacturer. This chip was then used as a catching support for suPARAP, injected subsequently, to obtain a level of immobilized suPARAP of ϳ5500 resonance units. Due to the binding characteristics of the 2.h.9: F12 antibody, this non-covalent immobilization step was irreversible under the conditions used, because the elution conditions for collagen (below) led to no release of bound suPARAP from the chip. Cleaved or intact collagen type-I was preincubated in assay buffer (0.01 M Hepes, 0.15 M NaCl, 1 mM CaCl 2 , 0.005% surfactant P20, pH 7.4) at 37°C, or at elevated temperature as indicated, after which various concentrations of the protein were injected. After each round of binding and dissociation, the chip was regenerated by injection of 10 mM glycine/ HCl, pH 2.0, to obtain complete dissociation of all bound collagen material. A parallel flow channel, containing coupled antibody but no suPARAP, served as a control for unspecific binding and buffer bulk subtraction. No binding of intact or cleaved collagen was obtained with antibody alone. The BIAcore interaction analyses were otherwise performed as described (40), monitoring binding and dissociation at 20°C in assay buffer at a flow rate of 10 l/min. Curve fitting to kinetic models and calculation of kinetic parameters were done using the BIAeval software, version 3.2 RC1, supplied with the instrument.
Analysis of Collagen Turnover in Growing Fibroblasts-A stock solution of collagen type-I in 0.02 M acetic acid was brought to neutral pH by addition of the appropriate volume of phosphate buffer to obtain a final concentration of 300 g/ml collagen, after which a trace amount of 125 I-labeled collagen type-I was added to obtain a radioactivity of ϳ1,500,000 cpm/ ml. Immediately thereafter, each well of a 24-well tissue culture plate was coated with 350 l of the tracer-containing collagen solution. The collagen was gelled by incubation at 37°C for 3 h and allowed to dry in a flow hood. For rehydration, the dried gel was washed with three changes of water, after which the wells were incubated with phosphate-buffered saline at room temperature for 24 h. Phosphate-buffered saline was removed, and 1 ϫ 10 5 cells were added to each well in a volume of 300 l of Dulbecco's modified Eagle's medium with 10% fetal calf serum, covering the bottom of the well. The cells were allowed to attach for 4 h, after which the medium was removed and replaced with 400 l of Dulbecco's modified Eagle's medium, including 1 nM interleukin 1␤ and 10 nM tumor necrosis factor-␣. After 5 days of cell culture, the conditioned medium from each well was collected. The media were analyzed by SDS-PAGE followed by Western blotting or automated phosphorimaging analysis as described previously (17). After harvest of the media, the cells and residual matrix were washed with phosphate-buffered saline. Cells were then detached by trypsinization, after which the matrix was stained with Coomassie Blue for visualization of zones of collagen clearance, as described before (7).

uPARAP-dependent Internalization and Degradation of Col-
lagen Fragments-Because the extracellular routes of collagen turnover are initiated by a well defined cleavage in the triple helical structure (41), we wanted to learn whether the same initial event might serve as a starting point for the endocytic collagen degradation route in fibroblasts. Consequently, we tested whether defined collagen fragments, resulting from treatment of native collagen with a mammalian collagenase, can serve as ligands for uPARAP-mediated internalization.
For the subsequent internalization studies, the intact collagen and the collagen fragments were labeled with 125 I (Fig. 1B). Analyses of uPARAP-mediated internalization were then undertaken, studying the endocytic activity of fibroblasts from newborn mice using radiolabeled collagen and collagen fragments as the ligands. This setup allowed studies on endocytosis in a system which, in the case of intact collagens, has been shown previously to be completely uPARAP-dependent (17).
Furthermore, the use of parallel cultures of fibroblasts from wild-type and littermate control uPARAP-deficient mice (uPARAP Ϫ/Ϫ ) enabled us to also test this uPARAP dependence for unknown samples, such as the collagen fragments.
This study revealed that both the intact and the cleaved collagen are taken up by wild-type (uPARAP ϩ/ϩ ) cells ( Fig. 2A, black columns). Strikingly, however, the uptake of the cleaved protein (column 5) was much more efficient than the internalization of the untreated collagen (column 1). This increased efficiency was indeed the result of proteolytic cleavage, because preincubation of intact collagen at 25°C without collagenase-3 did not affect the internalization (column 3). The uptake of both cleaved and uncleaved collagen was completely uPARAP-dependent, because fibroblasts from littermate uPARAP Ϫ/Ϫ mice displayed no internalization (gray columns). To ensure that the uptake of the collagen fragments was indeed part of a degradation pathway, we also studied the effect of blocking lysosomal cysteine proteases in the fibroblasts. To this end, a similar experiment was set up with the inclusion of E-64d (Fig. 2B), an inhibitor that has been shown previously to block lysosomal degradation of internalized collagen-IV (28). Comparison of parallel samples obtained in the absence and presence of E-64d revealed that the blocking of intracellular degradation led to a marked increase in intracellular radioactivity in the uPARAP ϩ/ϩ cells (Fig. 2B, black columns). Whereas this was the case both for intact collagen (columns 1 and 3) and cleaved collagen (columns 5 and 7), the amount of collagen fragments accumulated in the cells was much larger than the amount of intact collagen, also in the presence of the inhibitor. The presence of E-64d did not affect collagen uptake in uPARAP Ϫ/Ϫ cells (gray columns).
To enable the visualization of the collagen fragments after endocytosis, FITC-labeled collagen type-I was treated with collagenase-3, leading to complete cleavage into fluorescent products with an electrophoretic migration pattern indistinguishable from that obtained with unlabeled collagen (Fig. 1A, lanes  3 and 4). The fluorescent fragments, as well as FITC-labeled intact collagen, were added to live fibroblasts from wild-type mice or littermate uPARAP-deficient mice, after which endocytosis was allowed to proceed in the presence of E-64d (Fig. 3). Examining the wild-type cells by fluorescence microscopy after 22 h of endocytosis, a strong fluorescence was observed in intracellular vesicles in the cells incubated with the cleaved FITC-collagen (Fig. 3B). The same cellular distribution was observed with FITC-labeled intact collagen-I (Fig. 3D), but strikingly, the amount of accumulated, fluorescence-labeled intact collagen was much lower than that of the collagen fragments, although the amounts of added fluorescent protein were identical. Again, this endocytosis was completely uPARAP-dependent, because no uptake or intracellular accumulation occurred with uPARAP-deficient cells (Fig. 3, A and C). The perinuclear, vesicular distribution of fluorescent protein, obtained with both cleaved and intact collagen-I, was indistinguishable from that obtained with fluorescence-labeled collagen-IV (result not shown), confirming results reported previously where the latter ligand was shown to accumulate in lysosomes in response to the arrest of lysosomal degradation (28).
Altogether, these results showed that the defined proteolytic fragments of collagen-I are taken up much more efficiently than intact collagen by a uPARAP-dependent mechanism, through which they are directed to further lysosomal degradation.
uPARAP-mediated Binding of Native and Cleaved Collagen in a Purified System-Molecular binding studies with collagen fragments and recombinant uPARAP were performed by surface plasmon resonance-based interaction analysis. A pilot study with intact collagen showed that a monoclonal antibody against uPARAP, designated mAb 2.h.9:F12, was well suited for immobilizing recombinant soluble uPARAP on the sensor surface in an orientation allowing collagen binding, with no collagen binding to the antibody alone, as demonstrated in the absence of uPARAP (data not shown). Upon injection of native collagen type-I (Fig. 4A, upper panel), the binding curves obtained in this system were in good accordance with a simple 1:1 Langmuir binding model. The optimal fit to this model was FIGURE 2. uPARAP-dependent internalization of labeled collagen and collagen fragments. A, increased uptake after cleavage. 125 I-Labeled untreated, mock incubated, or cleaved collagen type-I (30 ng/well) was added to fibroblast cultures derived from wild-type (uPARAP-expressing) mice (ϩ/ϩ, black columns) or from littermate mice with homozygous uPARAP deficiency (Ϫ/Ϫ, gray columns). After 3 h of incubation at 37°C, the intracellular fraction of each cell sample was isolated and the amount of internalized labeled protein was determined by gamma counting. All samples were analyzed in triplicate. The data are represented as the internalized fraction of the added radioactivity; note that the same amount of protein material was added to all samples. Error bars indicate standard deviations. B, internalized collagen fragments are routed to intracellular degradation. Fibroblast cultures were incubated with 125 I-labeled intact or cleaved collagen type-I as in A, except that incubation was performed in the absence or presence of 20 M E-64d, as indicated. obtained with the lowest concentration (2 g/ml) of collagen tested, with only marginal (Ͻ2 resonance units) deviations from the theoretical curve, as shown in the residual plot (Fig.  4A, lower panel). For this binding curve, an apparent K d in the sub-nanomolar range was estimated (see figure legend for kinetic parameters and curve fitting), in accordance with competition studies performed previously with uPARAP on intact cells (15).
Binding analyses were then performed on the cleaved collagen in the same system (Fig. 4B). Again, a strong interaction with the immobilized uPARAP was observed, but, as expected for a fragment mixture, the binding curves could in this case not be fitted satisfactorily to a 1:1 Langmuir binding model. Nevertheless, the sensorgrams still revealed some important new features. Apparently, the complexes formed were somewhat less stable than those obtained with intact collagen (compare the dissociation phases, ϳ440 -800 s of Fig. 4, A and B), but an even much more pronounced difference was noted in the association phase. Comparing the early injection phase (ϳ70 -120 s) of Fig.  4 (A and B), it was evident that one or several components in the cleaved collagen had a much higher association rate with uPARAP than the intact collagen.
The conformational properties of collagens are highly sensitive to limited proteolysis. This is connected to the fact that collagen undergoes quite well defined denaturation transitions as a function of temperature (44). Thus, whereas intact triple helical collagen is conformationally stable at body temperature, the 1 ⁄ 4 and 3 ⁄ 4 fragments are both known to be conformationally unstable, gradually entering a gelatin-like state at 37°C (45,46). Intact collagen undergoes a very similar conformational transition but only after heating to higher temperatures (45). In the standard conditions employed above (see "Experimental Procedures"), all samples were kept at 37°C before injection to present the collagen in the same state as adopted in the internalization assays. Consequently, a possible explanation for the increased uPARAP association rate in the samples with cleaved collagen could be the acquisition of a gelatin-like state in the fragments. This hypothesis was supported by studies on heat-denatured collagen and fragments. Thus, preincubation of intact collagen at 65°C before injection led to a binding curve with uPARAP characterized by a fast initial binding phase (Fig. 4C, upper curve), clearly distinct from the curve obtained with native, intact collagen (Fig. 4A) but very similar to the situation obtained with collagen fragments, irrespective of whether the latter had been incubated at 37°C (Fig. 4B) or at 65°C (Fig. 4C,  lower curve).
Altogether, these experiments showed that collagen fragments, preincubated at body temperature, displayed an increased uPARAP association rate relative to intact collagen. This increase, however, could be mimicked by heat-denatured, intact collagen. Thus, the acquisition of a gelatin-like state in the fragments was responsible for the effect observed.
Increased Uptake of Cleaved Collagen Is Mimicked by Heatdenatured, Uncleaved Collagen-An increased association rate with uPARAP might explain the initial finding that cleaved collagen was internalized more efficiently than intact collagen by uPARAP-mediated endocytosis. If so, a high internalization FIGURE 4. Surface plasmon resonance analysis of the binding of collagen and collagen fragments to uPARAP. The monoclonal antibody, 2.h.9:F12, was immobilized on a BIAcore chip surface after which the antibody was used as a capturing support for recombinant suPARAP, injected subsequently. The indicated collagen derivatives were then injected, and each round of binding and dissociation was recorded through a 360-s injection phase and a 360-s dissociation phase. Each sensorgram is shown after bulk subtraction, recorded in a parallel flow cell with immobilized antibody and no pre-injection of suPARAP; no binding of collagen or collagen fragments was detected with antibody alone. The concentration of collagen (micrograms/ml) is indicated for each curve. RU, resonance units. A, binding of uncleaved, native collagen type-I to recombinant suPARAP. The sensorgrams obtained with three concentrations of collagen, preincubated at 37°C before BIAcore analysis, are shown in the upper panel. The lower panel shows the residual plot obtained from the curve with 2 g/ml collagen after fitting to a theoretical Langmuir 1:1 binding model in the BIAeval program. The kinetic parameters derived from this curve were k (assoc) ϭ 9.1 ϫ 10 5 s Ϫ1 M Ϫ1 ; k (diss) ϭ 1.5 ϫ 10 Ϫ5 s Ϫ1 and an apparent (kinetically derived) dissociation constant of K d ϭ 0.16 ϫ 10 Ϫ10 M. B, binding of cleaved collagen type-I. BIAcore analysis was performed as in A, except that the injected protein was cleaved collagen-I, which had been preincubated at 37°C in the same manner. The sensorgrams shown represent the same concentrations of collagen material (micrograms/ml) as in A. Note the steep association phase in this set of sensorgrams. C, sensorgrams showing the binding of intact (upper curve) and cleaved (lower curve) collagen type-I, both subjected to heat-induced denaturation by incubation at 65°C for 20 min before injection. Both samples were injected at a concentration of 2 g/ml. efficiency might also be expected for heat-denatured collagen, because this ligand displayed the same steep association phase as the collagen fragments. Consequently, we compared the fibroblast-mediated internalization of radiolabeled collagen and collagen fragments after preincubation at different temperatures (Fig. 5).
This experiment revealed that the gelatin-like product, obtained by heating uncleaved collagen-I to 65°C, was internalized much more efficiently than the unheated protein (compare columns 1 and 3). Actually, the uptake observed was very close to that obtained with the same amount of unheated cleaved collagen (column 5). Furthermore, in the case of the cleaved collagen, heating to 65°C before the assay led to only a marginal increase in the internalization efficiency (column 7). Thus, in the latter case, the major part of the molecular transition responsible for the high efficiency had occurred already at 37°C, closely reflecting the molecular behavior observed above during the BIAcore binding studies. Importantly, none of these binding reactions were the result of a loss of specificity in the denatured state, because the cellular internalization was completely uPARAP-dependent for all the protein samples. This was shown using uPARAP-deficient cells as the negative control (Fig. 5, columns 2, 4, 6, and 8).
Fibroblasts Direct the Sequential Cleavage of Collagen and the uPARAP-mediated Endocytosis and Degradation of Collagen Cleavage Products-Finally, we wanted to learn whether a sequential collagen degradation mechanism, involving both initial extracellular cleavage and endocytosis-dependent complete breakdown, is actually operative in living cells. Therefore, we seeded uPARAP-expressing and uPARAP-deficient fibroblasts on a reconstituted, trypsin-resistant collagen-I matrix and analyzed the content of the culture supernatants after 5 days of cell culture using Western blotting with an anti-collagen antibody for the detection of released collagen fragments (Fig. 6A). A striking difference was observed. In the culture supernatant of the uPARAP-deficient cells, well defined collagen fragments had accumulated (lane 4). A strong signal was observed at the electrophoretic position of the 1 ⁄ 4 collagen fragment (compare with the sample of cleaved purified collagen; lane 2) and the 3 ⁄ 4 fragment could also be detected, although with a lower band intensity. Strikingly, both of these fragments were absent in the culture supernatant of the wild-type cells (lane 3). In the latter sample, the only high molecular weight FIGURE 5. Endocytosis of cleaved and heat-denatured collagen. Samples of 125 I-labeled cleaved and intact collagen, respectively, were either preincubated at 65°C for 20 min before the assay or used for internalization assay directly, as indicated. Internalization assays with uPARAP ϩ/ϩ and uPARAP Ϫ/Ϫ cells were then carried out as described in the legend to Fig. 2. Error bars indicate standard deviations of triplicate samples. FIGURE 6. Fibroblasts direct the sequential cleavage of collagen and uptake of collagen fragments. A and B, fibroblasts from newborn uPARAP ϩ/ϩ and littermate uPARAP Ϫ/Ϫ mice, respectively, were grown on a reconstituted collagen type-I matrix with a tracer of 125 I-labeled collagen type-I included. After 5 days of cell culture, the conditioned culture supernatants were harvested and analyzed by non-reducing SDS-PAGE, followed by Western blotting with a polyclonal antibody against collagen-I (A) or by phosphorimaging analysis of the gel for localization of radiolabeled collagen and its degradation products (B). In each panel, the culture supernatants from the uPARAP ϩ/ϩ and uPARAP Ϫ/Ϫ cells are shown in lanes 3 and 4, respectively. The following purified reference proteins are shown: A: lane 1, unlabeled, intact collagen type-I; lane 2, unlabeled collagen type-I, cleaved with collagenase-3 as described in the legend to Fig. 1A. B: lane 1, 125 I-labeled, intact collagen type-I; lane 2, 125 I-labeled, cleaved collagen type-I. The 1 ⁄4 and 3 ⁄4 collagen fragments are indicated. C, after harvest of the cell culture media analyzed in A and B, cells were detached, and the remaining collagen matrix was stained with Coomassie Blue. Upper sample: uPARAP Ϫ/Ϫ cells. Lower sample: uPARAP ϩ/ϩ cells. D, upper row, fibroblasts homozygously deficient for MMP-13 (left), MMP-14 (center), and MMP-2 (right) were cultured as above, after which cells were detached followed by Coomassie staining of the matrix, performed as in C. Lower row, for each genotype, a parallel wild-type control sample was included, representing a fibroblast culture obtained from a littermate control wild-type mouse. Note the complete lack of collagen clearance in the MMP-14-deficient cell sample and the partial effect in the MMP-2-deficient sample. component recognized by the antibody was intact collagen-I, which was present in very low amounts. This material was likely to reflect diffusion into the medium of minor contents of noncross-linked collagen from the matrix, because the same component was found in wells incubated with growth medium in the absence of cells (data not shown). In the low molecular mass region (Ͻ15 kDa), unidentified collagen fragments resulting from extensive degradation were evident. These degradation products, however, were present in the culture supernatants of both of the cell types.
In addition to being the primary cell type involved in collagen clearance, fibroblasts also synthesize collagen (47). Therefore, it was important to verify that the collagen fragments observed above were indeed derived through cleavage of the extracellular collagen compartment and were not just degradation products of aberrant neosynthetic collagen. This was ascertained through the incorporation of 125 I-labeled collagen in the reconstituted collagen matrix. An autoradiogram of an SDS gel with the culture supernatant samples (Fig. 6B) revealed the same well defined collagen fragments in the uPARAP Ϫ/Ϫ supernatants as observed in Western blotting. Again, these fragments were absent in the uPARAP ϩ/ϩ supernatant.
These experiments demonstrated that fibroblasts grown on a collagen matrix direct an initial collagen cleavage step, which leads to well defined fragments that accumulate in the extracellular space if uPARAP is not present on the cells. On the other hand, when the uPARAP pathway is operative, these fragments are efficiently endocytosed and degraded.
According to this notion, whereas uPARAP has a primary role in the clearance of the cleaved collagen, the endocytic event would not be expected to be the major determinant in the initial attack on the collagen matrix under these conditions. This was confirmed by comparing the zones of collagen clearance in the solid matrix after cell culture. Coomassie staining of the collagen matrix after detachment of the cells (7) revealed extensive zones of clearance, which were indistinguishable when comparing uPARAP-expressing and uPARAP-deficient cells (Fig.  6C). We therefore studied the effect of deletion of various candidate collagenase genes on this initial event, using primary fibroblast cultures from the respective gene-deficient mice and matched wild-type cell cultures from littermate control mice in each case (Fig. 6D). In accordance with results published previously (7), deficiency for the membrane-bound collagenase, MMP-14 (MT1-MMP), completely abolished collagen clearance (center wells). In contrast, deficiency for the soluble collagenase, MMP-13 (collagenase-3), led to zones of clearance indistinguishable from those in the wild-type control (wells to the left). Because MMP-14 acts both to degrade collagen directly (9, 48 -50) and to activate MMP-2 (gelatinase-A (51)), which is a third protease with collagenolytic activity (3), we also studied the effect of MMP-2 deficiency (wells to the right). In this case, the gene deficiency led to a clearly discernible reduction in the degree of clearance but not a complete abolition. Thus, part of the function of MMP-14 in this system might be exerted through MMP-2 activation, but because the effect of MMP-14 deficiency was stronger than that of MMP-2 deficiency, a direct MMP-14-mediated collagen cleavage seemed to be central to the process, as also suggested previously (7).
Altogether, these results document the existence of a sequential collagen degradation pathway, initiated by pericellular collagenolytic activity and completed by uPARAP-mediated fragment uptake and intracellular degradation.

DISCUSSION
In this work, we show that uPARAP-mediated endocytosis of defined collagen fragments serves as a distinct step in a consecutive mechanism of clearance that involves extracellular collagen cleavage, uPARAP-mediated internalization, and lysosomal degradation. This observation adds an important new role of uPARAP to the previously established reactivity toward intact collagen (15,17,18,22).
In the cell culture system studied, the uPARAP-dependent step may be indispensable, because the fragments from the initial cleavage accumulated in its absence. On the other hand, fragments of a much lower molecular weight did also appear with uPARAP-deficient cells, thus suggesting that parallel degradation systems are also operative (Fig. 6, A and B). Most likely, this alternative route is exclusively extracellular and involves the attack of a gelatinase such as MMP-2 on the fragments initially formed by collagenase cleavage (12). However, irrespective of any alternative mechanism, the lack of uPARAP makes the degradation of the initial fragments a rate-limiting step in the total process. Furthermore, comparing the defined, major fragments found in the cell culture system with uPARAP-deficient cells, the accumulation of the 1 ⁄ 4 fragment was much more pronounced than that of the 3 ⁄ 4 fragment. Therefore, any alternative route of degradation may be directed primarily against the larger fragment in this system. An important question is which enzymes in the extracellular compartment serve to perform the initial cleavage of the triple helical collagen in the consecutive mechanism of collagen degradation described here. Because different proteases with collagenolytic activity exist, and because these enzymes have different expression patterns, it is likely that this cleavage can be performed by more than one enzyme in vivo and that the details of this event have at least some tissue specificity. Fibroblasts are generally recognized as a primary cell type in collagen turnover (47,52). In our assay system with fibroblasts in vitro, MMP-14 was indispensable, whereas MMP-2 deficiency had only a partial effect. MMP-13, a third collagenase with expression in fibroblasts (5), could be deleted without any detectable effect in this assay system.
In addition to the sequential process described here, it is likely that alternative, parallel routes of collagen breakdown are also operative in vivo, depending on the physiological process in question. Thus, whereas MT1-MMP is centrally engaged in the initial collagen cleavage in the combined mechanism in our cell culture system, both uPARAP and MT1-MMP also take part in collagen turnover processes independently. This follows from the fact that mice with double deficiency for both of these proteins have an even more compromised collagenolytic function than mice deficient for either component alone. 4 Furthermore, it is well established that uPARAP is also active against intact, native collagens (17,23,28). As shown in Fig. 2, uPARAP-mediated endocytosis occurs with both cleaved and intact collagen in vitro, and in malignant mammary tumors of wild-type mice, intact fibrillar collagen was observed inside the tumor-embedded fibroblasts, whereas no intracellular collagen could be found in the tumors of mice deficient for uPARAP (18). The latter observation points to a uPARAP-dependent internalization process comprising regular particulate collagen, which requires a phagocytic mechanism as also reported previously (14). This phagocytic pathway and the involvement of uPARAP therein may be completely distinct from the consecutive mechanism described in this work.
The uPARAP-mediated endocytosis of the collagen fragments, just like the uptake of intact collagen, probably results from a direct binding of the ligands in question to uPARAP on the cell surface. Using BIAcore-based interaction analysis, we observed a strong binding between immobilized uPARAP and cleaved as well as uncleaved collagen. Although the cleaved collagen did not fulfill the model requirements for a complete kinetic analysis, it was evident that at least one component in the cleaved sample bound very rapidly to uPARAP, and indeed, this process was substantially faster than the association phase obtained with intact collagen (Fig. 4, A and B). Because the endocytosis of cleaved collagen in fibroblasts was also faster than the uptake of intact collagen, it is tempting to speculate that this is the result of a faster binding step. In this connection, it is noteworthy that uPARAP recycles rapidly and constitutively between the plasma membrane and the endosomal compartment, resulting in the fast reappearance of the unoccupied receptor on the cell surface (30). Therefore, a rapid ligand binding step may be important for maximal endocytic efficiency.
Although a detailed investigation of the binding motifs within the isolated collagen fragments is beyond the scope of the present work, our studies do add to previous studies on the uPARAP-collagen interaction. Thomas and coworkers (22) showed that uPARAP binds to a region in the C-terminal part of the collagen I ␣1 chain, although it was not excluded that additional binding determinants may exist. The region in question is part of the 1 ⁄ 4 fragment formed by collagenase treatment, and, consistent with the reported observation, we could clearly detect the accumulation of this fragment in the supernatant of uPARAP-deficient fibroblasts. As noted above, however, some accumulation of the 3 ⁄ 4 fragment was also observed with the uPARAP-deficient cells, and because the latter fragment was indeed cleared from the supernatant by uPARAP-expressing cells, it is likely that this product also displays some binding to uPARAP. However, irrespective of any difference in the affinity of uPARAP for each of the two fragments, it is clear that the large conformational rearrangements, known to occur upon collagen cleavage, are important for the effects observed in this work. This is clearly illustrated by the parallel behavior of collagen fragments and heat-denatured collagen, respectively, in direct binding as well as endocytosis. It should be noted that the ability to bind both native collagen and gelatin is quite a widespread property among proteins with fibronectin type-II domains (53) and that, in several cases, a preference for gelatin has been noted in this respect (54 -56).
The complete lack of internalization of cleaved as well as intact collagen in fibroblasts from uPARAP-deficient mice suggests that uPARAP is the exclusive collagen clearance receptor in fibroblasts. It is noteworthy, however, that collagen interactions have also been reported for other members of the same protein family, i.e. the mannose receptor (57,58) and the phospholipase A 2 receptor (59), which, thus, may serve to fulfill a similar function in other cell types. This has indeed been suggested in the case of the mannose receptor (58), although an actual endocytic function of this receptor with collagen or its fragments remains to be demonstrated.
Altogether, our work suggests that the interplay between extracellular proteolysis and specific endocytic mechanisms is a crucial factor in the directed breakdown of extracellular matrix material. This organized process is an important subject in the attempts to elucidate and manipulate matrix turnover in conjunction with several pathological processes, including those of cancer invasion.