Collagenolysis-dependent angiogenesis mediated by matrix metalloproteinase-13 (collagenase-3).

We have demonstrated previously that new blood vessel formation induced by angiogenic growth factors in onplants placed on the chorioallantoic membrane (CAM) of the chick embryos is critically dependent on the cleavage of fibrillar collagen by a previously unidentified interstitial collagenase. In the present study we have used a quantitative CAM angiogenesis system to search for and functionally characterize host avian collagenases responsible for the collagen remodeling associated with angiogenesis. Among the matrix metalloproteinases (MMPs) identified in the CAM onplant tissue, the chicken MMP-13 (chMMP-13) was the only enzyme whose induction and expression coincided with the onset of angiogenesis and blood vessel formation. The chMMP-13 cDNA has been cloned and recombinantly expressed. The chMMP-13 protein has been purified, characterized in vitro, and examined in situ in the CAM. MMP-13-positive cells appear in the CAM shortly after angiogenic stimulation and then accumulate in the collagen onplant tissue. Morphologically, the chMMP-13-containing cells appear as hematopoietic cells of monocyte/macrophage lineage. In vitro, the chMMP-13 proenzyme is rapidly and efficiently activated through the urokinase plasminogen activator/plasminogen/plasmin cascade into a collagenase capable of cleaving native but not the (r/r) mutant collagenase-resistant collagen. Surprisingly, nanogram levels of purified chMMP-13 elicit an angiogenic response in the CAM onplants comparable with that induced by the angiogenic growth factors. The chMMP-13-mediated response was efficiently blocked by select protease inhibitors indicating that plasmin-activated chMMP-13 can function as an angiogenic factor in vivo. Altogether, the results of this study extend the physiological role of MMP-13, previously associated with cartilage/bone resorption, to the collagen remodeling involved in the angiogenic cascade.

ing through the vascular basement membrane, proliferation and invasion into the surrounding tissue matrix, and coalescing of endothelial cells with other vascular tissue components to form tubular arrays all represent events that evoke the induction of proteolytic enzymes as contributors to such remodeling. The targeted substrates for the induced proteases would be the proteinaceous components of the vascular basement membrane and the extracellular matrix within the angiogenic tissue. The most dominant proteins in the vascular tissue milieu are the interstitial collagens. A number of non-collagen extracellular proteins also contribute to the architecture of vascular tissue, but these molecules (e.g. the adhesive glycoproteins), although functionally very significant, compositionally are often minor when compared with the collagens.
Proteolytic remodeling of the structural collagen in vascular tissue would not be a straightforward catalytic event. Triple helical, fibrillar collagen is resistant to many proteolytic enzymes, including the potent digestive proteases trypsin and chymotrypsin and to plasmin, the major extracellular serine protease present in body fluids and tissues. Furthermore, among the members of an extensive family of matrix metalloproteinases (MMPs), 1 only a limited number can cleave the highly structured fibrillar collagens under physiological conditions (2,3). This cleavage can be carried out by several soluble MMPs, including MMP-1 (collagenase-1), MMP-8 (collagenase-2), MMP-13 (collagenase-3), and possibly MMP-2 (gelatinase A) and by at least one member of the subfamily of membraneanchored MMPs, MMP-14 (membrane type 1 matrix metalloproteinase). These enzymes are the main candidates that within vascular tissue could catalyze the initial cleavage in type I collagen, at the Gly 775 -Ile 776 bond, causing a partial unfolding of the fibrillar collagen and thereby making it susceptible to the gelatinases and to other proteases in the vascular tissue. Thus it is likely that this initial and highly specific cleavage reaction would constitute a rate-limiting step in the remodeling of collagen-enriched tissues.
In the case of invasive tumor cell remodeling of collagenenriched tissue, recent studies (4,5) indicate that MMP-14 is the major collagen-cleaving enzyme. In contrast, when dermal keratinocytes migrate through collagen, MMP-13 would appear to be the principal collagen-cleaving enzyme (6). However, it has not yet been clarified which specific collagen-cleaving MMP(s) might be the rate-limiting enzyme(s) in angiogenesislinked tissue remodeling and which vascular-associated cell types produce the enzyme(s). By using a variety of in vitro angiogenic models, it has been shown that specific MMPs, in particular MMP-14 and MMP-2, are up-regulated in both endothelial and stromal cell cultures (1,(7)(8)(9). In contrast, a number of in vivo studies using gene-depleted null mice have indicated that MMP-9 expression and activity correlate positively with the progression of angiogenesis (10,11). However, MMP-9, although a potent gelatinase, is clearly not an interstitial collagenase (2,12) and therefore could not initiate the rate-limiting cleavage of fibrillar collagen in vascular tissue. Thus it would appear that additional in vivo studies are needed to evaluate the angiogenic involvement of specific collagenases.
For in vivo modeling of angiogenesis, the chorioallantoic membrane (CAM) of the developing chick embryo historically has been a frequently used model system (13,14). It provides an accessible, facile in vivo system to examine molecular events that occur during new blood vessel formation induced by angiogenic growth factors or tumors. The CAM system has implicated MMP-2 as one possible contributory proteolytic enzyme (9,15). This enzyme has the catalytic potential to cleave interstitial collagen (12) but mainly has been examined in the CAM as a potential modulator of specific integrins (15).
Our laboratory has developed and utilized a quantitative CAM angiogenesis assay that involves placing select collagens onto the CAM in the absence or presence of angiogenic factors (16). Within 24 h various stromal cells, macrophage-like cells, and endothelial cells influx the collagen onplant from the underlying CAM, thereby expanding the vascular tissue. At the end point of the assay, only the new blood vessels that have formed in the collagen/CAM onplants are scored. It has been demonstrated that the new blood vessel formation in the CAM is MMP-dependent (16). Furthermore, the use in this assay of a specific, collagenase-resistant, mutant collagen (r/r) has implicated an undefined interstitial collagenase as the rate-limiting angiogenic MMP (16).
In the present study we have used the CAM angiogenesis system to screen for host avian collagenases responsible for the collagen cleavage and remodeling associated with new blood vessel formation. We have identified chicken MMP-13 (chMMP-13), induced in the CAM angiogenic onplants, as a likely candidate. The chMMP-13 cDNA has been cloned from the total RNA isolated from the onplant tissue and recombinantly expressed. The chMMP-13 protein has been purified, characterized in vitro, and then examined in situ in the CAM and in other chick tissues. We demonstrate that the chMMP-13 zymogen is rapidly and efficiently activated into an interstitial collagen-cleaving enzyme by the uPA/plasminogen cascade. Moreover, purified chMMP-13, when added at nanogram levels onto the CAM, can function directly as a stimulator of new blood vessel formation.

MATERIALS AND METHODS
CAM Angiogenesis Assay-Angiogenesis assays were performed as described previously (16). Fertilized White Leghorn chicken eggs were received from SPAFAS (North Franklin, CT) and incubated in a humidified incubator at 38°C. At day 4, eggshells were carefully removed, and embryos were incubated throughout the length of the experiment under shell-less conditions, in a covered dish placed in a humidified air incubator at 38°C and 60% humidity. Onplants were generated by overlaying two gridded plastic meshes and embedding them into 30 l of 1.6 mg/ml collagen of different types and origin. Where indicated, collagen was supplemented with basic fibroblast growth factor (FGF) at 16.7 g/ml and VEGF at 5 g/ml (PeproTech Inc., Rocky Hill, NJ). Collagen onplants were placed on the CAM of 10-day-old shell-less embryos. Test proteins or reagents were incorporated into the onplants at the concentrations indicated in the text. Onplant-bearing embryos were incubated for an additional 66 h at which time the extent of onplant vascularization was quantified. Newly formed vessels were identified by analyzing the upper plane of the onplant with a dissecting microscope (i.e. scoring only those vessels that sprouted up from the CAM and reached the plane of the upper mesh). The angiogenic index of the onplant tissue was determined as the percentage of grids that contained newly formed blood vessels from the total number of grids in the upper mesh. Data were processed using Graphpad Prism software (Graphpad Software Inc. San Diego). Statistical significance was estimated with the Mann-Whitney test. To facilitate angiogenic scoring in some experiments, 100 l of india ink were injected intravenously to visually enhance the appearance of the vasculature. Images were taken at ϫ6.3 by using a digital video camera mounted on an Olympus SZ60 dissecting microscope (Olympus, Melville, NY).
Immunohistochemistry and Histology-Onplants and embryo tissues were collected immediately after scoring, fixed in 10% zinc formalin, and paraffin-embedded. 4 -6-m-thick sections were deparaffinized and processed either for histological or immunohistochemical analyses. For routine histology, the sections were stained with Gills III hematoxylin and eosin. Giemsa-based dyes were used to discriminate the cells of hematopoietic origin. For immunohistochemistry, the chMMP-13 antigen was retrieved by heating the sections in 0.01 M citrate buffer, pH 6.0, for 5 min in a microwave oven. Nonspecific binding was blocked with 3% bovine serum albumin and 5% normal goat serum in phosphate-buffered saline. The sections were subsequently probed with a polyclonal huMMP-13-specific antibody (Triple Point Biologics Inc., Forest Grove, OR), which is directed to the highly conserved hinge domain of the enzyme and cross-reacts with chMMP-13. Bound MMP-13 antibody was visualized with the chromogen diaminobenzidine, resulting in dark brown staining. Counterstaining was performed with Gills III hematoxylin. Images were taken using a DVC digital camera mounted on an Olympus BX60 microscope (Olympus, Melville, NY) and processed with Adobe Photoshop (Adobe Systems, Seattle, WA) and Macromedia Freehand software (Macromedia Inc., San Francisco).
RNA Isolation-RNA was isolated from cells and tissues using either Trizol reagent (Invitrogen) or RNeasy kit (Qiagen Inc., Valencia, CA) following the manufacturers' instructions. RNA was suspended in RNase-free water and stored at Ϫ80°C.
Screening for Expression of chMMPs in the CAM Angiogenic Tissue-The full-length sequence of the chMMP-13 cDNA was determined by using a combination of gene family PCR (17), gene-specific primers, and 5Ј-RACE. Collagen onplants were dissected from the CAM and snapfrozen in liquid nitrogen. Total RNA was purified from the onplants, and 1 mg of RNA was reverse-transcribed to first strand cDNA with random hexamer primers. After cDNA synthesis, the PCR was carried out with degenerate primers from the two most conserved domains found in MMPs: the Cys-switch and the zinc-binding regions. The forward primer was 5Ј-AGCCIMGITGYGGIRWICCIGA-3Ј; the reverse primer was 5Ј-GATGICCIADYTCRTGIRCIGCIAC-3Ј (I ϭ deoxyinosine, Y ϭ C ϩ T, R ϭ A ϩ G). The reaction was cycled at 95°C for 5 min, followed by 35 1-min cycles at 94°C for denaturation, 1 min at 40°C for annealing, and 2 min at 72°C for elongation. The PCR products were cloned into pCR4-Topo vector. Those clones containing an insert of ϳ450 bp were sequenced.
Cloning of chMMP-13 Full-length cDNA-The 5Ј 279-bp fragment coding for the first 93 amino acids of the chMMP-13 cDNA was cloned using the SMART RACE cDNA amplification system (Clontech, Palo Alto, CA). Briefly, mRNA was purified from FGF-treated chick embryonic fibroblasts and then reverse-transcribed using Superscript II (Invitrogen) at 42°C for 1.5 h primed with SMART II oligonucleotides and 5Ј coding sequences. The first strand cDNA was subjected to touchdown PCR by using universal primers provided in the kit and genespecific primer 5Ј-AAGTTCTGCTTCAACCATCTGCGGG-3Ј. The PCR was carried out under the following cycling conditions: 5 cycles at 94°C for 5 s and 72°C for 3 min, 5 cycles at 94°C for 5 s, 70°C for 10 s and 72°C for 3 min, followed by 27 cycles 94°C for 5 s, 68°C for 10 s, and 72°C for 2 min. This reaction yielded a 1-kb fragment that was cloned into pCR4-Topo vector and sequenced.
Expression of Recombinant chMMP-13 Protein-The cDNA containing the entire coding sequence for chMMP-13 was cloned into the expression vector pEE12 and used to transfect the murine myeloma cell line NS0 as described (18). Wells containing NSO cells expressing the chMMP-13 protein were identified by using gelatin and casein zymography of conditioned medium, and then cell lines were isolated and cloned by limiting dilution.
Purification of Recombinant Proteins-For isolating recombinant chMMP-13, serum-free conditioned medium from NS0 cell clones expressing chMMP-13 was dialyzed against 20 mM Tris-HCl, pH 7.5, 10 mM CaCl 2 buffer and then applied to a UnoS column, equilibrated with the same buffer, and attached to a Biologic DuoFlow FPLC system (Bio-Rad). Bound protein was eluted with a linear gradient of 0 -1 M NaCl in the same buffer and analyzed by SDS-PAGE, Western blotting, and zymography. Chicken MMP-2, MMP-9, and TIMP-2 were purified as described previously (12,19,20).
Collagenase Assays-Assays for the cleavage of native, acid-soluble collagens by specific MMPs were performed essentially as described (12). Briefly, collagens were solubilized in 10 mM acetic acid and diluted to the indicated concentrations in 40 mM Tris-HCl, pH 7.5, 200 mM NaCl, 10 mM CaCl 2 , 0.05% Brij-35 buffer (calcium assay buffer). Individual MMPs or organomecurial p-aminophenylmercuric acetate (APMA) was added to collagen mixtures at the concentrations indicated in the text. Reactions were performed at 23°C to maintain the native triple helical collagen status and then terminated by the addition of 10ϫ reducing SDS sample buffer. The samples were analyzed for specific collagen cleavage by SDS-PAGE.
SDS-PAGE, Western Blotting, and Zymography-Proteins separated by SDS-PAGE were detected in the gels by either Coomassie Blue R-250 or silver staining (21). For Western blotting, separated proteins were transferred to nitrocellulose or PVDF membranes. After blocking with 5% non-fat dry milk in phosphate-buffered saline, 0.05% Tween 20 (PBST), the membranes were incubated overnight at 4°C with 1 g/ml primary MMP-13 polyclonal antibody (Triple Point Biologics Inc., Forest Grove, OR), washed four times with PBST, incubated for 1 or 2 h at room temperature with horseradish peroxidase-conjugated goat antirabbit IgG (Pierce) diluted at 1:2,500 in PBST, 5% milk, and developed with ECL reagent (Amersham Biosciences). Casein (0.67 mg/ml) and gelatin (0.28 mg/ml) zymography was performed on 10 or 12% SDS-PAGE gels as described (12). Following electrophoresis, gels were washed twice in 2.5% Triton X-100, incubated overnight at 37°C in calcium assay buffer, and stained with Coomassie Blue R-250.
Activation of chMMP-13 Zymogen by Plasmin-Chicken plasmin was generated by incubation of purified plasminogen (2.5 M) with recombinant uPA (5% conditioned medium from HeLa cells transfected with the chicken uPA cDNA) in 25 mM Tris, pH 8.0, 0.1% Triton X-100 (TTX buffer) for 1 h at 37°C. Where indicated, aprotinin was added at a final concentration of 10 M to block the activity of generated plasmin. To analyze whether plasminogen or uPA alone could induce chMMP-13 activation, these proteins were preincubated individually and thereafter mixed with the chMMP-13 proenzyme. The chMMP-13 zymogen was added at 33 nM to each of reaction mixtures, which were further incubated at 37°C. At the indicated time intervals, the reaction was stopped with ϫ10 SDS sample buffer. Proteins from 20-l aliquots were separated by SDS-PAGE on 10% gels under nonreducing conditions, transferred to a PVDF membrane, and probed with 1 g/ml anti-MMP-13 rabbit antibody. Bound antibodies were detected with the secondary horseradish peroxidase-conjugated goat anti-rabbit IgG, and MMP-13 bands were visualized by ECL.
To determine the actual amount of plasmin generated from 2.5 M plasminogen upon incubation with recombinant chuPA, a standardized assay for plasmin-mediated cleavage of a specific tripeptide substrate was employed. The plasmin substrate S2251 (American Diagnostics, Greenwich, CT) was incubated with serial dilutions of purified human plasmin (1.5-100 nM), and S2251 cleavage was monitored during a 1-h incubation at 37°C according to the manufacturer's instructions. In parallel, aliquots of the chicken plasminogen/uPA incubation mixture were serially diluted and similarly assayed with the plasmin substrate S2251. The concentration of plasmin in the incubation mixture was then determined based on a kinetic analysis of the ensuing cleavage reaction compared with the standardized cleavage reaction with puri-fied plasmin. The analysis confirmed that chuPA was the rate-limiting factor in conversion of plasminogen to plasmin, yielding ϳ10 -25 nM of plasmin from 2.5 M chicken plasminogen. In addition, it was also determined that at a concentration as low as 2 nM, purified human uPA was capable of generating 1 nM plasmin from 250 M of human plasminogen (data not shown), indicating that the chicken uPA/plasminogen cascade operates at a similar efficiency in generating plasmin as the human uPA/plasminogen cascade.
Collagenolytic Activity of the Plasmin-activated chMMP-13 Enzyme-The chMMP-13 zymogen was activated by plasmin as described above. After incubation for 1 h at 37°C, aprotinin was added at 10 M to all reaction mixtures to block any residual collagenolytic activity of plasmin or plasmin-generating reagents. Where indicated, recombinant chTIMP-2 was added at a stoichiometric concentration of 33 nM to inhibit proteolytic activity of the generated chMMP-13 enzyme. The collagenolytic ability of the plasmin-activated chMMP-13 was assessed by the 3/4 -1/4 cleavage of type I collagen. A 20-l aliquot of the above described reagent mixtures was mixed with 20 g of type I rat collagen in 30 l of calcium assay buffer and incubated for 18 h at 23°C. Proteins were separated by SDS-PAGE on a 10% gel, and protein bands were visualized by Coomassie staining.

MMP-13 Expression in the Chicken CAM Correlates with
Growth Factor-induced Angiogenesis-A quantitative method for scoring new blood vessel formation in the chick embryo was used to examine the involvement of specific MMPs in angiogenesis. The method involves placement of collagen-enmeshed grids on the CAM and then scoring for the newly formed blood vessels that distinctly appear in the upper grids of the collagen onplant (see Ref. 16 and "Materials and Methods"). The development of functional, blood-bearing vessels in the collagen onplants was stimulated by the addition of the angiogenic growth factors FGF and VEGF. The number of grids containing newly formed blood vessels increased 5-10-fold in the presence of FGF/VEGF as compared with control conditions lacking exogenous angiogenic factors (Fig. 1A). This growth factorinduced blood vessel formation was inhibited almost completely by purified chTIMP-2 ( Fig. 1A). In conjunction with the reported sensitivity of CAM angiogenesis to synthetic broad spectrum MMP inhibitors (16), the latter result with a natural MMP inhibitor strongly indicated an involvement of avian MMP(s) in the formation of new vessels in the CAM. Moreover, CAM angiogenesis was substantially reduced when the normal wild type (wt) collagen in the onplants was substituted with a collagenase-resistant mutant (r/r) collagen (Fig. 1A). This finding suggests that during angiogenesis, one of the rate-limiting MMPs is an interstitial collagenase that specifically cleaves wt collagen at the site mutated in the r/r collagen.
To identify potential MMP candidates, the temporal expression of various chicken MMPs was examined in both control and growth factor-containing onplants. Total RNA was extracted from the CAM onplants at 15 h (before any new blood vessels have formed), at 45 h (when new vessels have begun to sprout), and at 66 h (the peak of new vessel formation) (16). The RNA was reverse-transcribed and subjected to PCR analysis by using degenerate primers targeting the two most conservative domains found in MMPs, the cysteine switch and zinc-binding regions.
The chicken MMP-1, -8, and 14 mRNAs were not detected in the CAM extracts. In contrast, the mRNA for chMMP-2 (gelatinase A), chMMP-13 (collagenase-3), and chMMP-16 (MT3-MMP) were readily detected. However, mRNA expression levels of chMMP-2 and chMMP-16 were not affected by the angiogenic growth factors and did not correlate with the kinetics of CAM angiogenesis (Fig. 1B). Furthermore, as shown previously, the tissue distribution of chMMP-2, which can function as an interstitial collagenase (12), did not correlate with the sites of new blood vessel development in the CAM (16). In contrast to chMMP-2 and chMMP-16, chMMP-13 mRNA ex-pression was up-regulated by the angiogenic growth factors and exhibited a temporal pattern that closely correlated with the kinetics of new blood vessel formation (Fig. 1B). Thus, among identified MMPs, an interstitial collagenase, i.e. chMMP-13, was the only MMP candidate whose expression correlated positively with CAM angiogenesis.
Cloning of Full-length Chicken MMP-13 cDNA from CAM Onplants-To examine in more detail the possible role of chMMP-13 in angiogenesis, we cloned, expressed, isolated, and biochemically characterized the recombinant enzyme. A previously reported partial cDNA sequence of chMMP-13 (22) covered the region from the beginning of the catalytic domain to the 3Ј-untranslated region. In order to generate a full-length chMMP-13 cDNA, total RNA was isolated at 66 h from growth factor-treated CAM onplants. The reverse-transcribed RNA was then employed as a template for 5Ј-RACE using primers specific for known regions of the chMMP-13 cDNA. The resulting full-length coding sequence and deduced amino acid sequence are shown in Fig. 2A. The newly identified 360 nucleotides in the 5Ј end of the transcript contained an initiation ATG at the beginning of a complete open reading frame of 1413 nucleotides. Our sequence is nearly identical to the partial sequence in the overlapping regions as reported previously (22). One difference involves an A to C change at nucleotide 310, which results in a predicted activation site for chMMP-13 at VGE2YNF instead of VGE2YKF as reported previously (22). The newly deduced activation site in chMMP-13 is identical to that in huMMP-13 (23). Fig. 2B shows an alignment of the chicken and human MMP-13 deduced amino acid sequences with the different domains and regions of the enzyme delineated. Overall chMMP-13 is 71% identical to huMMP-13 at the protein level ( Table I). The highest homology exists between the catalytic domains (82%), whereas the propeptide and hemopexin domain homology is only 58 and 67%, respectively. The overall homology and interdomain homology between chMMP-13 and the six known mammalian MMP-13 molecules is presented in Table I. The interspecies comparison indicates that chMMP-13 is as similar (or dissimilar) to other mammalian MMP-13 as it is to huMMP-13.
Expression of chMMP-13 in Embryonic Tissues-The expression of chMMP-13 mRNA in various embryonic tissues was examined by RT-PCR. A specific signal for chMMP-13 could be detected after 35 cycles in most embryonic tissues tested, with the highest levels observed in lungs, intestine, and stomach (Fig. 3A). In contrast, chMMP-2 mRNA levels appeared to be abundant and are readily detected in all of the tested tissues, even at 30 cycles.
To confirm the tissue-dependent expression of chicken MMP-13 at the protein level, we examined the distribution of chMMP-13 antigen by immunohistochemistry. Fig. 3B shows strong MMP-13-positive staining in lung and stomach and weaker staining in intestine, i.e. tissues that yielded the highest RT-PCR signals (Fig. 3A). The specific immunohistochemical signal appeared to localize in a select population of round, mononuclear cells in all three tissues. Distinct staining for the chMMP-13 antigen also appeared in regions of bone development (data not shown), consistent with previous reports (24,25) that MMP-13 is associated with cartilage resorption and bone remodeling.
Expression, Isolation, and Partial Characterization of Recombinant chMMP-13-To analyze the putative angiogenic properties of chMMP-13, the full-length cDNA for chMMP-13 was recombinantly expressed in the mouse NSO myeloma cell line, using vectors and methods that previously had been used successfully to produce 0.1-1 mg quantities of avian MMPs (17,19). In contrast to mock-transfected cell cultures, the serumfree conditioned medium isolated from chMMP-13-transfected NSO cells demonstrated a major 60-kDa protein (Fig. 4A, lanes  1 and 2), which strongly reacted with the MMP-13-specific antibody in Western blotting (Fig. 4B, lane 2). The 60-kDa protein also exhibited caseinase and gelatinase activities in the respective substrate gel zymographs (lanes 2 in Fig. 4, C and D). A weakly stained 50-kDa protein, also immunologically reactive with anti-MMP-13, was present in the conditioned medium. This protein was resolved better in the substrate gels and manifested distinctly enhanced gelatinase and caseinase activities, indicating that it represented the processed, active form of chMMP-13, whereas the 60-kDa protein appeared to be the zymogen form of chMMP-13.
FIG. 1. MMP-13 expression in the CAM correlates with the angiogenic response to growth factors. A, growth factor-induced angiogenesis in the CAM onplants depends on cleavage of collagen by a TIMP-2-sensitive interstitial collagenase. The onplants were prepared with wild type collagen (wt Col), supplemented with or without angiogenic growth factors (FGF/VEGF), or mutant collagenase-resistant collagen (r/r Col). Angiogenic response was determined 66 h after the placement of collagen onplants onto the CAM and presented as the percentage (%) of grids with blood vessels from a total number of grids in the individual onplant. Where indicated, chTIMP-2 was incorporated into wt collagen at 1 g/ml. B, expression of chMMP-13 is increased during angiogenic response in the CAM collagen onplants. The chMMP-2, chMMP-16, and chMMP-13 mRNA expression during growth factor-induced angiogenesis was determined using RT-PCR of total RNA that was extracted from onplants harvested at the indicated time points. The level of expression is presented as the MMP/GAPDH PCR product ratio for each MMP. The overall progression of angiogenesis is depicted graphically as a function of time.
The conditioned medium from the chMMP-13 transfected cultures was passed through a fast protein liquid chromatography cation exchange column, and bound proteins were eluted with a 0.1-1.0 M NaCl gradient. Thirty five fractions were collected, and greater than 80% of the total MMP-13 enzymatic activity was isolated in four fractions (Fig. 4, lanes 3-6). The two central fractions contained a mixture of ϳ75% chMMP-13 zymogen and 25% active enzyme (Fig. 4B, lanes 4 and 5), whereas the leading fraction included mainly zymogen (Fig.  4B, lane 3), and the trailing fraction contained mainly active enzyme (Fig. 4B, lane 6). A 25-30-kDa immunologically reactive fragment of chMMP-13, identified in the conditioned medium, was distinctly resolved in one of the fractions (Fig. 4B, lane 4) but was not enzymatically active in the zymographs.
In order to activate the proenzyme form of chMMP-13, the isolated fraction that contained mainly the 60-kDa chMMP-13 zymogen was treated with 1.0 mM APMA. Aliquots were removed at various times and analyzed for conversion to the 50-kDa active form (Fig. 5A). Conversion to the lower molecular weight, processed form of chMMP-13 was detectable within 5-10 min, and nearly complete conversion occurred by 90 -120 min.
To assess the interstitial collagenase activity of chMMP-13, the APMA-generated, active form of the enzyme was then incubated at 23°C with purified chicken collagens types I-III and V (Fig. 5B). True collagenase activity of chMMP-13, as measured by the generation of distinct 3/4 and 1/4 collagen fragments, was manifested with types I-III chicken collagens but not with type V collagen. The activity was especially pronounced with type III collagen. That chMMP-13 exhibited true interstitial collagenase activity was further demonstrated by incubating the activated enzyme at 23°C with either wild type (wt) or mutant (r/r) collagenase-resistant mouse type I collagen, the latter being the collagen that was unable to support CAM angiogenesis (Fig. 1A). The chMMP-13 clearly cleaved wt collagen, generating 3/4 and 1/4 fragments but failed to cleave r/r collagen (Fig. 5C). This result, in combination with the

TABLE I Full-length and domain homology between chicken
and mammalian MMP-13 proteins The deduced amino acid sequences of MMP-13 from the indicated species were aligned with the deduced amino acid sequence of chMMP-13 by using BLOSUM 62 matrix software. Domains were delineated as described previously (39). Homology is indicated in percent. The GenBank TM accession numbers for MMP-13 are as follows: human, NM_002427; mouse, NM_008607; rabbit, AF059201; rat, XM_217083; horse, AF034087; and bovine, AF072685.  Human  71  58  82  82  67  Mouse  71  56  83  88  66  Rabbit  73  57  85  82  69  Rat  71  53  82  83  68  Horse  73  58  84  82  69  Bovine  72  57  84  70  68 enhanced expression of chMMP-13 during FGF/VEGF-mediated CAM angiogenesis (Fig. 1), provided further evidence for chMMP-13 as a critical collagen-remodeling enzyme during new blood vessel formation. Efficient and Rapid Activation of the chMMP-13 Proenzyme through the uPA/Plasminogen Cascade-Several reports (26,27) have indicated that huMMP-13 can be activated by members of the MMP family, including MMP-2, MMP-3, and MMP-14. Of the activating MMPs, chMMP-2 and chMMP-9 were the only chicken homologues available in purified form (17,19) for testing as possible activators of chMMP-13. When active chMMP-2 and chMMP-9 were incubated for up to 16 h with the chMMP-13 proenzyme, no processing of the 60-kDa MMP-13 zymogen was detected (Fig. 6A). Even if an active form of huMMP-3 was incubated for 16 h with the chMMP-13 proen- The highest levels of chMMP-13 message are identified in the lung, intestine, and stomach, but lower levels of chMMP-13 are detectable in all tissues examined. The chMMP-2 and chGAPDH messages are readily detected in all tissue samples. No cDNA template was incorporated in the negative control. B, immunohistochemical analysis of chMMP-13 protein expression. Lung, stomach, and intestine from 13day-old embryos were fixed, paraffin-embedded, and processed for immunostaining by using MMP-13-specific polyclonal antibody (brown). The nuclei are counterstained with hematoxylin (blue). Whereas some nonspecific staining is seen in glandular structures both after staining with MMP-13 antibody and staining without primary antibody (control), specific staining for the enzyme is most apparent in round scattered cells (arrows). These cells resemble hematopoietic cells of monocyte/macrophage lineage and are readily distinguishable in the lung (upper panel), stomach (middle panel), and intestine (lower panel).

FIG. 4. Expression, purification, and zymographic analysis of recombinant chMMP-13.
Full-length chMMP-13 was expressed in murine myeloma NSO cells as described under "Materials and Methods." Serum-free conditioned medium (SFCM) from control and MMP-13-transfected cells (lanes 1 and 2, respectively) and fractions 28 -31 eluted from a UnoS cation exchange column (lanes 3-6) were assessed for total protein content by silver staining (A) and for chMMP-13 protein by Western blotting (B). Activation status of the expressed and purified enzyme was assessed by casein (C) and gelatin (D) zymography. An arrow indicates the chMMP-13 60-kDa zymogen, and an arrowhead indicates the 50-kDa activated enzyme. Fragments of chMMP-13 protein identifiable by silver staining, and apparently containing the hinge domain as revealed by Western blotting, are indicated with asterisks. Undefined active fragments visible in the gelatin zymogram are indicated with (E). zyme, no apparent processing occurred. However, if chMMP-13 zymogen was incubated with purified chicken plasminogen plus chicken uPA that can efficiently generate plasmin from the plasminogen, rapid processing of the 60-kDa chMMP-13 proenzyme was observed in less than 1 min and was nearly completed by 30 min (Fig. 6B). The 50-kDa processed form of chMMP-13 was quite stable under these plasmin-generating conditions; however, some proteolytic fragmentation of FIG. 5. chMMP-13 is an interstitial collagenase. A, APMA-mediated activation of the chMMP-13 zymogen. Purified 60-kDa proenzyme of chMMP-13 was incubated with 1 mM APMA at 23°C for the indicated time periods and then analyzed for conversion to the 50-kDa processed enzyme by SDS-PAGE followed by silver staining. B, collagenolytic activity of the chMMP-13 enzyme. Purified chMMP-13 zymogen was incubated with the indicated native chicken collagens at 23°C in the presence (ϩ) or absence (Ϫ) of APMA for 16 and 32 h. Following incubation, the reactions were terminated and analyzed by SDS-PAGE under reducing conditions. Positions of molecular weight markers are indicated on the left. Native collagen ␣ chains (Col) and the 1/4 and 3/4 cleavage products are indicated on the right. C, chMMP-13 is a true interstitial collagenase. The ability of APMA-activated chMMP-13 to cleave wild type (wt) and collagenase-resistant mutant (r/r) murine type I collagen was assessed as described in B for chicken collagens. chMMP-13 occurred after 60 -120 min.
The concentration of plasminogen (2.5 M) utilized for the processing of the chMMP-13 zymogen (33 nM) in Fig. 6B, although in large molar excess over the zymogen, is the natural physiological concentration of plasminogen found in plasma and other tissue fluids (28). In addition, the actual enzyme/ substrate molar ratio is not unusually high because the conversion efficiency of plasminogen to plasmin by chuPA under the conditions used was only 0.5-1.0%, as determined by the measuring of plasmin activity after 1 h of preincubation of 2.5 M plasminogen with chuPA (data not shown, see "Materials and Methods"). Thus, not more than 25 nM of active plasmin is present during the processing of 33 nM of chMMP-13 zymogen.
In order to determine whether even lower concentrations of plasmin could efficiently process the chMMP-13 zymogen, thereby approaching expected physiological enzyme/substrate ratios, the chicken plasminogen was added at decreasing concentrations after activation with chuPA (Fig. 6C). Levels of plasminogen as low as 20 nM (yielding plasmin at 0.2 nM) in 1 h converted almost all of the 60-kDa chMMP-13 zymogen to a major 50-kDa processed form with trace levels of a 53-55-kDa processed form (Fig. 6C, left panel). The 4 nM plasminogen conditions caused ϳ50% conversion of chMMP-13 zymogen after 1 h. Further incubation under these conditions (4 nM plasminogen), representing an enzyme (0.04 nM plasmin) to substrate (33 nM proMMP-13) molar ratio of 1:825, resulted in almost complete conversion of chMMP-13 zymogen to the 50-kDa processed form in 2 h (Fig. 6C, right panel). That plasmin is indeed the converting enzyme is indicated by the complete abrogation of the processing events by aprotinin, a potent inhibitor of plasmin (Fig. 6C, last lane). Thus conversion of chMMP-13 proenzyme to its 50-kDa processed form in 2 h, at enzyme to substrate ratios approaching 1:1000, demonstrates that plasmin is a highly efficient, physiological processing enzyme for chMMP-13.
The requirements for the chMMP-13 activation cascade and a demonstration that generated chMMP-13 enzyme was catalytically active are presented in Fig. 7. The activation status of the chMMP-13 protein was analyzed by Western blotting after a 1-h incubation of the zymogen with the various components of the activating cascade (Fig. 7, upper panel). Both uPA and plasminogen were required for efficient chMMP-13 processing (lanes 2 and 6), because either one alone failed to reduce significantly the levels of the 60-kDa chMMP-13 zymogen (Fig. 7,  lanes 3 and 4). Plasminogen alone generated a small amount of the 50-kDa species (lane 3), but this was because of trace levels of plasmin present in the plasminogen preparation (data not shown). That the generated plasmin is the main activator of the chMMP-13 proenzyme in this system again is indicated by results showing that the plasmin inhibitor, aprotinin, completely prevented the generation of the 50-kDa chMMP-13 form (Fig. 7, lane 5).
To verify the interstitial collagenase activity of the generated 50-kDa form of chMMP-13, all the above-indicated samples were added to triple helical collagen and further incubated for 24 h at 23°C. The results are illustrated in corresponding lanes of the lower panel of Fig. 7. Only when pro-MMP-13 was incubated with uPA and plasminogen without the plasmin inhibitor did the generated 50-kDa enzyme mediate a 3/4 -1/4 cleavage of triple helical collagen (Fig. 7, lane 2Ј). If TIMP-2 was added to the active 50-kDa protein, no collagen cleavage was observed (Fig. 7, lane 6Ј), indicating that the cleavage of triple helical collagen was because of active MMP-13 and not plasmin or uPA.
Monocyte-like Cells Containing chMMP-13 Are Present at Elevated Levels in the Angiogenic CAM Tissue-The presence of chMMP-13 protein during angiogenesis was examined using immunohistochemical analysis of normal CAM and CAM with collagen onplants (untreated and growth factor-containing) (Fig. 8). A few chMMP-13 positive cells were present in normal CAM tissue from day 13 embryos (on average less than 1.0 cell per field). Control onplants undergoing low to moderate levels of angiogenesis contained about 10 -15 chMMP-13 positive cells per field. In contrast, 35-40 chMMP-13 positive cells per field could be found in the FGF/VEGF onplants, undergoing extensive formation of new blood vessels (Fig. 8A). Thus, growth factor-induced angiogenesis in the CAM is accompanied by the appearance of cells carrying MMP-13 protein.
A low power view of a representative FGF/VEGF CAM onplant is shown in Fig. 8B, panel a. Numerous chMMP-13 positive cells are scattered among the mesenchymal cells in the vascularized CAM tissue. The cells with chMMP-13 protein are often clustered near small blood vessels. At higher magnification, chMMP-13 positive cells adjacent to blood vessels (Fig. 8B,  panels c and d) can be easily distinguished from the wispy, pale-stained stromal cells that make up the majority of the CAM tissue and from the endothelial cells, which line the blood vessels and do not exhibit any chMMP-13-specific staining. In contrast, MMP-13-positive cells are round, characterized by a relatively high volume of condensed cytoplasm, and distinct oval or bean-shaped nuclei with large nucleoli. Differential staining of adjacent sections with Giemsa dyes indicated that the chMMP-13 positive cells are of hematopoietic origin and appear to belong to the monocyte-macrophage lineage (data not shown). The increased numbers of MMP-13-positive cells in the FGF/VEGF-containing tissue (Fig. 8A) suggest that these cells respond to growth factors by influx and/or differentiation in the CAM and precede or accompany the formation of new blood vessels.
Involvement of Active chMMP-13 in CAM Angiogenesis-The findings that elevated levels of chMMP-13 are present in tis- Thereafter, 20 l from each reaction were added to 20 g of type I collagen in 30 l of calcium assay buffer. Following incubation at 23°C for 24 h, the samples were processed by SDS-PAGE under reducing conditions (lanes 1Ј-6Ј). The gels were then analyzed for the cleavage status of collagen. In the upper panel the status of chMMP-13 activation was analyzed by Western blotting (WB) as described in Fig. 6. In the lower panel the collagenolytic activity of the plasmin-activated chMMP-13 was assessed by monitoring the 3/4 -1/4 cleavage of type I collagen by Coomassie Blue staining of the gel. The chMMP-13 proenzyme (arrow), active enzyme (arrowhead), intact collagen (Col), and the cleaved collagen fragments (3/4 and 1/4) are indicated on the left. sues undergoing growth factor-mediated angiogenesis (Fig. 8) and that specific inhibitors of active MMPs substantially reduce the new blood vessel formation in the CAM (Fig. 1) suggested that active chMMP-13 may be directly involved in this angiogenesis model system. This hypothesis implies that in vivo the chMMP-13 zymogen would have to be activated in order to exert its angiogenic action.
The zymogen form of chMMP-13 has been efficiently activated in vitro by physiological concentrations of plasmin (Fig.  6). Because this activation could be completely inhibited by aprotinin (Fig. 7), we further analyzed the effects of this inhibitor on the in vivo angiogenic response in the CAM. Therefore, aprotinin was incorporated into the FGF/VEGF-containing onplants or added directly onto the onplants placed on the CAM. The aprotinin-treated embryos exhibited a complete reduction in growth factor-stimulated angiogenesis down to control levels of untreated animals (Fig. 9A).
To test directly the angiogenic capability of chMMP-13, the purified zymogen was mixed with the collagen in control onplants. The onplants containing only 15 ng of the chMMP-13 exhibited a 2-3-fold enhancement in the levels of new blood vessel formation (Fig. 9B). The enhanced angiogenesis mediated by purified chMMP-13 was prevented by aprotinin, indicating that in order to exert its angiogenic activity, the exogenous chMMP-13 zymogen has to be activated in vivo by endogenous plasmin or a plasmin-like enzyme. Most important, when wt(ϩ/ϩ) collagen was substituted with the collagenase-resistant (r/r) collagen in the onplants, chMMP-13 did not stimulate angiogenesis (Fig. 9B), a result consistent with the notion that the interstitial collagenase activity and collagen remodeling ability of chMMP-13 indeed provide the angiogenic stimuli.
The angiogenesis stimulated by purified chMMP-13 was examined microscopically. The upper panels of Fig. 9C, which are planar views of the upper grids in the collagen onplants, illustrate enhanced levels of new blood vessel formation in vivo within those collagen onplants containing chMMP-13 (compare Fig. 9C, panels a and b). The injected india ink contrast enhances the new blood vessels (Fig. 9C, arrows), and they appear as dark channels or capillary beds within the opaque collagen and are clearly more numerous and more dense in Fig. 9C, panel b. A dramatic reduction in the number of these chMMP-13-induced blood vessels resulting from the addition of aprotinin also is clearly observed (Fig. 9C, panel c). Histological sections cut through the onplants and stained with hemotoxylin and eosin are shown in the lower panel of Fig. 9C as cross-sections of onplants and the engulfing CAM tissue. In contrast to control onplants (Fig. 9C, panel aЈ), small angiogenic blood vessels (indicated by arrowheads) are observed within the upper regions of the onplant supplemented with purified chMMP-13 (Fig. 9C, panel bЈ). Remarkably, very few or no vessels are observed in the onplant treated with aprotinin (Fig. 9C, panel cЈ). However large numbers of viable stromal cells have accumulated in both of the chMMP-13-containing onplants (Fig. 9C, panels bЈ and cЈ); thus indicating that aprotinin did not exert cytotoxic or cytostatic effects. The lack of onplant vascularity in the presence of aprotinin (Fig. 9C, panel cЈ) indicates that this inhibitor of plasmin could specifically prevent the formation of those vascular structures that are induced in the CAM by chMMP-13 zymogen incorporated into collagen onplants. This finding is consistent with a requirement for specific activation by plasmin of chMMP-13 in order to exert the angiogenic effects of the enzyme. DISCUSSION In this study, we have shown that the enzymatic activity of an endogenous interstitial collagenase is required for the ini-tiation of the extracellular matrix remodeling events that accompany new blood vessel formation in the CAM. The evidence for the specific requirement of a collagenase in angiogenesis is based on the observations demonstrating that the migration of endothelial cells and the completion of vessel formation are prevented or substantially diminished when wild type collagen was substituted with collagenase-resistant (r/r) collagen in the CAM onplant matrix (Fig. 1A) (16). A limited subset of the MMP family represents the enzymes that specifically can cleave triple helical collagens, i.e. the true collagenases, including MMP-1, -8, -13, -14, -16, and possibly MMP-2 (2).
Among the avian collagenase homologues, chicken MMP-1, MMP-8, and MMP-14 were not found in the CAM angiogenic tissue despite an extensive search using a combination of PCR with degenerate primers, data base analysis, and potentially cross-reacting antibodies. On the contrary, chMMP-2, chMMP-13, and chMMP-16 were readily detected in the angiogenic CAM by RT-PCR or immunohistochemistry (Fig. 1B) (16). However, the mRNA levels for chMMP-16 did not correlate with the progression of CAM angiogenesis (Fig. 1B). Although chMMP-2 has been shown previously to efficiently hydrolyze fibrillar collagen (12), the expression levels, tissue distribution, and time and growth factor dependence of chMMP-2 did not correlate with angiogenesis induced in the CAM. Therefore, chMMP-13 would be the likely enzyme initiating specific collagen cleavage in growth factor-induced angiogenic remodeling, although it is possible that the endogenous chMMP-16 and chMMP-2 could contribute to CAM angiogenesis in other catalytic manifestations.
Several lines of evidence implicate chMMP-13 as the endogenous collagenase responsible for initiating the critical matrix remodeling events in the CAM. First, the chMMP-13 expression is characterized by time-, tissue-, and growth factor-dependent induction. Second, activation of the chMMP-13 proenzyme in vivo can be achieved through a physiologically relevant mechanism involving the uPA/plasminogen/plasmin cascade. Third, an active chMMP-13 efficiently and specifically cleaved interstitial collagens, while being incapable of cleaving mutant r/r collagen (Figs. 1 and 5-8). Finally, in addition to the close coordination between CAM angiogenesis and endogenous chMMP-13 expression, purified chMMP-13 added directly to the CAM onplants at nanogram quantities induced substantial levels of new blood vessel formation (Fig. 9). To our knowledge, the latter finding represents a novel demonstration that a single exogenous MMP can induce angiogenesis in vivo. Remarkably, exogenous chMMP-13 stimulated blood vessel formation in the CAM onplants in the absence of supplemental FGF/VEGF.
The precise mechanisms underlying angiogenic stimulation by active chMMP-13 have yet to be identified. It is unlikely that the chMMP-13 activity simply substitutes for the effects of FGF and/or VEGF. The more likely event(s) is that exogenous chMMP-13, upon activation in the CAM tissue, cleaves the fibrillar collagen and further initiates a cascade of events that result in the accessibility of endogenous FGF and VEGF or other angiogenic stimulatory factors. However, the cleavage of fibrillar collagen has to be the initiating event because the addition of chMMP-13 to CAM onplants containing collagenase-resistant r/r collagen failed to induce angiogenesis (Fig.  9B). The cleavage of wild type interstitial collagens by chMMP-13 could either expose angiogenic cryptic sites or generate angiogenic collagen fragments. An even more complex scenario may involve chMMP-13-dependent cleavage of collagen thereby creating a matrix pathway for endothelial cells as well as chMMP-13-induced cleavage of indeterminate compo-nents that allows for the subsequent release of angiogenic factors associated with the matrix components.
The above proposed hypotheses involving the cleavage of specific proteins by chMMP-13 and the subsequent release or generation of angiogenic products are supported by the exist-ence of MMP-mediated fragmentation of specific bioactive targets (reviewed in Refs. 3, 29, and 30). A few findings that relate to the specific growth factor-induced angiogenesis described herein include the increase of VEGF expression and production in MMP-14-transfected tumor cells (31), the release of bioactive FIG. 9. chMMP-13 mediates collagenolysis-dependent angiogenesis in CAM collagen onplants. A, growth factor-induced angiogenesis in the CAM depends on the activity of an aprotininsensitive serine proteinase. Collagen onplants with or without angiogenic growth factors (FGF/VEGF) were placed on the CAM. Where indicated, 10 M aprotinin was incorporated into the onplants. Levels of angiogenesis were determined as the percentage of grids with newly formed blood vessels. B, purified chMMP-13 mediates angiogenic response in the CAM collagen onplants. Buffer or purified chMMP-13 zymogen (15 ng) was incorporated into control (no growth factors) wild type collagen onplants (wt Col) with or without 10 M aprotinin. Where indicated, collagenase-resistant collagen (r/r Col) was used with or without purified chMMP-13 (15 ng). Levels of angiogenesis were quantitatively determined as the percentage of grids with newly formed blood vessels. Statistical significance of data is as follows: **, p Ͻ 0.01; ***, p Ͻ 0.001; Ø, p Ͼ 0.05. C, morphological analysis of chMMP-13-containing CAM onplants. To visualize the physiological changes in the angiogenic tissues, the onplants were prepared for histological examination. Immediately after the quantitative assessment of angiogenesis, a group of embryos was injected with india ink to contrast and visualize the vasculature in the onplant under a binocular dissecting microscope (a-c, ϫ6.3) or were processed for histological examination (aЈ-cЈ, ϫ20). The india ink outlines the new blood vessels (arrows) within the grids of the collagen onplants in the planar views of panels a-c. Hematoxylin and eosin staining of 4-m sections of CAM onplants illustrates the vascularization and the stromal cell infiltration of collagen onplants in the cross-sections of aЈ, bЈ, and cЈ. The CAM-onplant interface is depicted by the dashed line. Arrowheads indicate the newly formed blood vessels in the upper plane of the onplants, and shortened arrows indicate established vasculature in the CAM. The plastic mesh of the onplant is indicated with #.
VEGF by MMP-9 (11), the exposure of an angiogenic cryptic site in type IV collagen by MMP-9 (32), or the generation of active growth factors by MMP-9 (33), MMP-2 (34), or MMP-13 (35). The latter study demonstrated that chMMP-13 in vitro is capable of generating directly the active form of a potent growth factor.
It is interesting that endogenous chMMP-13 is detected in the CAM vascular tissue at the time of angiogenic induction. Neither chMMP-13 protein nor cells expressing the chMMP-13 antigen were found in the normal, untreated CAM, whereas abundant chMMP-13-containing cells were identified in the vascularized CAM 2.5 days after the addition of FGF/VEGF (Fig. 8A). Despite expectations, our immunohistochemical analyses have not identified mesenchymal stromal cells and endothelial cells in the onplants as the purveyors of the chMMP-13 collagenase. Instead, a distinct population of cells whose morphology suggested a hematopoietic origin was surprisingly the major if not the only source of chMMP-13 (Fig.  8B). These chMMP-13-positive cells apparently enter the vascular CAM in response to FGF or VEGF. We suggest that other forms of angiogenic induction such as tumor formation, wound repair, female reproductive cycling, and acute and chronic arthritic occurrences may also involve the influx of monocyte-like responder cells containing varying levels of specific proteolytic enzymes that could directly contribute to vascular tissue remodeling. In this regard, inflammation-responsive hematopoietic cells such as monocytes, macrophages, and neutrophils have been shown to import specific MMPs catalytically involved in matrix remodeling of tumor tissue (30,36). Therefore, our discovery of the interstitial collagenase, chMMP-13, in hematopoietic cells of the monocyte-macrophage lineage, which enter the CAM tissue in response to angiogenic factors, is consistent with recent findings in tumor biology.
The chMMP-13 molecule, like all members of the MMP family is expressed as an inactive zymogen, which must be converted to its active form in order to remodel the collagen-enriched vascular tissue. Extensive studies have demonstrated that huMMP-13 (collagenase-3) is activated by a select number of other MMP family members. By using in vitro cell culture models and purified reactants in solution, it has been shown that MMP-3, MMP-2, and MMP-14 efficiently convert the 60-kDa huMMP-13 zymogen to the 48-kDa active enzyme, proceeding through a 50 -55-kDa autolytic intermediate (26,27,37,38). Surprisingly, the plasminogen/plasmin cascade was excluded from the physiologic pathway of huMMP-13 activation because plasmin generated by concanavalin A-stimulated human fibroblasts did not activate the huMMP-13 proenzyme. In agreement, the plasmin inhibitor aprotinin failed to prevent huMMP-13 activation by cells that expressed active MMP-2 and -14 (27). Our results with the chMMP-13 zymogen are in sharp contrast. The 60-kDa chMMP-13 proenzyme resisted activation in vitro by the tested chicken and human MMPs, including soluble chMMP-2 and huMMP-3 (Fig. 6A), and in culture by cell surface-expressed huMMP-14, 2 but was very efficiently activated by the plasminogen/plasmin cascade (Fig. 6, B and C).
The processing of the chMMP-13 zymogen by 10 -25 nM chicken plasmin proceeds immediately, within 1 min, is complete in 30 min, and generates a 48 -50-kDa form of chMMP-13 capable of cleaving triple helical interstitial type I collagen into its signature 3/4 -1/4 fragmentation polypeptides (Fig. 7). Even much lower concentrations of plasmin (Ͻ1.0 nM) can generate the 48 -50-kDa active form of chMMP-13 in 1-2 h (Fig. 6C). Furthermore, plasmin appears to be involved in the activation of chMMP-13 in vivo because aprotinin, added to the CAM onplants, completely abolished the angiogenic response induced either by FGF/VEGF or by the chMMP-13 zymogen (Fig.  9). These findings clearly indicate that the uPA/plasminogen/ plasmin cascade indeed is involved in the activation of chMMP-13 under physiological conditions.
The activation pathways of huMMP-13 and chMMP-13 may contrast because of specific differences in the sequence and structure of the two homologues. Both the APMA activation site, 55 FFG2LEV 60 , where cleavage results in an activation intermediate, and the autoactivation site, 81 VGE2YN 85 , where cleavage generates a fully processed enzyme, are identical in human and chicken zymogens (Fig. 2B). However, the regions harboring the initiating MMP-14 and plasmin cleavage sites are not homologous in the two proenzymes: 33 LAG2ILK2EN 40 for huMMP-13 (26,27) and 33 PIGIMKKS 40 for chMMP-13 (where G 35 and K 38 are the putative P 1 amino acids involved in the MMP-14 and plasmin cleavages, respectively). Thus, in sharp contrast to the 100% homology of the APMA and autoactivation sites, these regions in the human and chicken MMP-13 zymogens are only 38% homologous. Therefore, effectiveness of MMP-13 activation by MMP-14 or by plasmin may depend on how these nonhomologous, nonconserved regions of the propeptide domains are presented contextually to tissue environments enriched in either MMP-14 or plasmin. Evidently, those environments could be quite different for the human and chicken MMP-13 zymogens. In this regard, activation of the huMMP-13 proenzyme, expressed mainly in chondrocytes and tumor cells (24), was suggested to be mediated predominantly by MMP-14, whereas a physiological role for MMP-13 activation by plasmin was excluded (27).
chMMP-13 also has been identified in chondrocytes from the developing long bones of 15-day-old embryos (25). However, the findings of the present study show that the majority of chMMP-13-positive cells in other chick embryo organs, including lungs, intestine, and stomach, appear to belong to the monocyte/ macrophage lineage (Fig. 3B). Similarly appearing cells were virtually absent in the normal CAM at day 13 of development, yet a significant number of cells entering the CAM tissue in response to FGF/VEGF indeed express MMP-13 and morphologically belong to the monocyte/macrophage lineage (Fig. 8B). With a view that activated chicken monocytes are also able to produce and secrete uPA, 2 activation of the chMMP-13 zymogen could then be accomplished very efficiently in a tissue environment that is enriched in plasminogen. The highly vascularized embryonic CAM with a circulating concentration of plasminogen at 100 -150 g/ml (28) indeed can provide such a setting. Thus, chMMP-13-positive monocyte/macrophage cells in the CAM collagen onplants could serve as a source for both chMMP-13 zymogen and uPA, the physiological initiator of the avian plasmin activation cascade. Most important, mammalian MMP-13 generated through the plasmin pathway has been implicated in the dissolution of fibrillar collagen by MMP-14deficient murine keratinocytes (6), thus representing a clear precedent for remodeling of interstitial collagen not by MMP-14 but by a plasmin-activated MMP-13. The characterization herein of chMMP-13 as a plasmin-activated interstitial collagenase, characterized by its unique appearance in vascular tissue upon angiogenic stimulation and its demonstrated ability to directly contribute to new blood vessel formation, now clearly extends the collagen-remodeling role of this MMP to neovascularization.