ABSTRACT
Osteoclasts resorb the extracellular matrix of bone by secreting enzymes and acid into a sealed-off compart-ment that they form upon attachment to the bone sur-face. Although the lysosomal cysteine proteinases can degrade collagen after the demineralization of bone at low pH, several lines of evidence suggest that collage-nase (matrix metalloproteinase-1, EC 3.4.24.7) may also be involved in this process. The question of whether col-lagenase is present in the osteoclast and/or in the bone-resorbing compartment has however not been resolved. We have prepared an antimouse collagenase antiserum and affinity-purified an IgG fraction that specifically immunoblots and immunoprecipitates (pro)collagenase. Using these antibodies, we demonstrate by immunolo-calization the presence of (pro)collagenase both in the osteoclasts and in the extracellular subosteoclastic bone-resorbing compartment. These specific localizations were observed not only in mice but also in rat and rabbit osteoclasts and using not only the antibody we have pre-pared but also antibodies raised in other laboratories against rat (Jeffrey et al., J. Cell. Physiol. 143, 396-403, 1990) and rabbit (Brinckerhoff et al., J. Biol. Chem. 265, 22262-22269, 1990) collagenase. Intracellular collage-nase was observed in the osteoclasts whether the cells were plated on bone or cultured on glass coverslips. It is proposed that osteoclastic collagenase is secreted in the resorbing compartment where it may cooperate with the lysosomal cysteine proteinases in the degradation of the collagen component of the matrix during the resorp-tion of bone.
INTRODUCTION
The resorption of bone requires the removal of both the mineral and the organic (collagenous and non-collagenous) matrix components. This degradative process occurs in an extracellular compartment acidified by proton transport at the osteoclast ruffled border membrane (Baron et al., 1985; Blair et al., 1989a; Chatterjee et al., 1992). The low pH allows dissolution of the mineral phase, exposes the organic matrix and favors the degradative action of lysosomal enzymes (Vaes, 1968, 1988; Baron, 1989). In this acid envi-ronment, the lysosomal cysteine proteinases (i.e. non-specific collagenolytic acid proteases) secreted by the osteo-clast in the bone-resorbing compartment are not only involved in the degradation of collagen (Delaissé et al., 1980, 1984, 1987; Everts et al., 1988; Rifkin et al., 1991) but could also be sufficient for its complete degradation without requiring the participation of collagenase (i.e. matrix metalloproteinase-1 or MMP-1, EC 3.4.24.7, a member of a distinct class of enzymes characterized by its specific ability to cleave collagen within the triple-helical body of its native molecule).
Collagenase may nevertheless be implicated in the bone-resorption process. Despite early reports that collagenase could not be detected in osteoclasts (Sakamoto and Sakamoto, 1984; Blair et al., 1986a, 1989b), agents that stimulate bone resorption in organ culture have been reported to enhance the accumulation of collagenase in bone (Delaissé et al., 1988); inhibitors of matrix metallopro-teinases prevent the resorption of cultured bone explants (Delaissé et al., 1985, 1988) and interfere with the processes leading to collagen degradation in the subosteoclastic resorp-tion compartment (Everts et al., 1992). On the other hand, the secretion of collagenase by osteoblasts is also regulated in vitro (Heath et al., 1984; Otsuka et al., 1984; Sakamoto and Sakamoto, 1984; Partridge et al., 1987) and complete extraction of the collagenase present in bone requires dem-ineralization (Eeckhout et al., 1986; Delaissé et al., 1988), which led to the hypothesis that it could be localized within the mineralized matrix, where it may have been deposited during osteogenesis (Gillet et al., 1977; Mechanic et al., 1982). Hence, although collagenase may be involved in the resorption of bone, its localization relative to osteoclasts and resorbing areas remains undetermined.
The present study was therefore designed to obtain col-lagenase-specific antibodies and further determine by immunolocalization the relationship between collagenase and bone-resorbing cells. Our results demonstrate the pres-ence of this enzyme in the osteoclasts and in the subosteo-clastic bone-resorbing compartment, thereby suggesting that osteoclastic collagenase plays a direct role in the degra-dation of the bone matrix.
MATERIALS AND METHODS
Materials
Chemicals for electrophoresis were purchased from Bio-Rad Lab-oratories (Richmond, CA). CNBr-activated Sepharose, Protein A-Sepharose, Protein G-Sepharose and chromatography media were from Pharmacia LKB Biotechnology (Uppsala, Sweden). Biotiny-lated anti-goat IgG and streptavidin-biotinylated peroxidase were from Amersham (UK). All other reagents used were of analytical grade.
Source, purification and assay of procollagenase
Conditioned medium of newborn mouse calvaria cultured in the presence of 0.3 mg/ml heparin (Vaes, 1972) was the source of procollagenase. The following successive purification steps were done in the presence of 50 mM cacodylate buffer, pH 7, 5 mM CaCl2, 0.01% Triton X-100 and 0.02% NaN3: (1) affinity chro-matography on heparin-Sepharose; elution with a linear (0.1 to 1.3 M) NaCl gradient; (2) ion-exchange chromatography on DEAE-Sepharose; elution with a linear (0.1 to 0.6 M) NaCl gra-dient; (3) gel filtration on Ultrogel AcA 54 in the presence of 1 M NaCl; (4) dialysis of purified procollagenase (3,000 to 4,000 units/mg protein) against 0.15 M NaCl. Procollagenase was acti-vated by trypsin and assayed as previously described (Eeckhout et al., 1986).
Preparation of antibodies to (pro)collagenase and immunoaffinity columns
A 1 ml (80 μg) sample of purified (partially autoactivated) pro-collagenase was emulsified with an equal volume of complete Freund’s adjuvant and injected into three intradermal sites of an adult goat. Two further injections were given on days 14 and 28. A bleed of 400 ml, taken on day 56, served as a source of anti-(pro)collagenase serum and IgG. The IgG from preimmune (IgG/PR) and immune (IgG/AC) sera were prepared by ammonium sulfate precipitation (40%, pH 6.8) of the whole sera, puri-fied using DEAE-Sephacel® and dialyzed against PBS containing 0.02% NaN3. IgG concentration was determined spectrophoto-metrically, using A (1%; 1 cm, 280 nm)=14.0.
IgG/PR and IgG/AC were coupled to CNBr-activated Sepharose 4B® (5 mg IgG/ml drained gel) as described by the manufacturer. The IgG (PR or AC)-Sepharose columns were equilibrated with CCTN buffer (0.05 M cacodylate, pH 7.0, 5 mM CaCl2, 0.05% (w/v) Triton X-100 and 0.03 M NaN3) supplemented with 0.15 M NaCl. The antigens (conditioned medium of cultured mouse calvaria containing 18 units (pro)collagenase) were preincubated in the column, for 1 hour at room temperature, with 4 ml drained gel. The columns were sequentially eluted with 8 ml 0.15 M NaCl in CCTN buffer (fraction 1), 8 ml 1 M NaCl in CCTN buffer (fraction 2) and 8 ml 3 M KSCN in CCTN buffer (fraction 3). Samples of total antigen and of each fraction were analyzed by immunoblotting (see below) and by gelatin and casein zymogra-phy (Lefebvre et al., 1991).
Affinity purification of anti-(pro)collagenase antibodies from immunoblots
Electrophoretic immunoblotting was performed essentially as described by Towbin et al. (1979) using Tween-20 as a blocking agent (Batteiger et al., 1982). (Pro)collagenase-enriched culture media were dissolved in 1% (w/v) SDS non-reducing sample buffer and resolved by electrophoresis (Laemmli, 1970) in a 11.3% polyacrylamide slab gel (13 cm × 18 cm × 0.15 cm) with two molecular mass markers slots and a single sample slot extend-ing most of the width of the 4.6% polyacrylamide stacking gel. Electrophoresis was at 55 V until the tracking dye was within 1 cm of the bottom of the separating gel.
Electrophoretic transfer of the resolved proteins was done at 180 mA for 4 to 5 hours in a Trans-Blot® Cell (Bio-Rad). The transfer efficiency was checked by staining the polyacrylamide gel with Coomassie brilliant blue. The nitrocellulose membrane (NCM) was blocked with PBS containing 0.05% (v/v) Tween-20 (PBS-Tween). The two lateral strips bearing the molecular mass markers were excised and stained with Ponceau S solution or Indian ink. The remaining central NCM part was divided into strips (0.5 to 1.0 cm), either vertical ones for staining or hori-zontal ones for affinity purification of antibodies.
Antigenic proteins bound to nitrocellulose were detected by incubating the strips with the IgG diluted in PBS-Tween under slight agitation, either for 3 hours at 35°C or for 16 hours at 4°C and 1 hour at 35°C. The strips were washed in PBS-Tween (3 changes during 15 to 30 minutes, total), incubated for 2 hours at room temperature with the second antibody (biotinylated anti-goat IgG) diluted 400-fold in PBS-Tween and washed again as before. The biotinylated antibodies were revealed by an incubation of 30 minutes at room temperature with peroxidase-streptavidin diluted 400-fold in PBS-Tween, a 10 minute wash with PBS-Tween fol-lowed by a similar wash with 20 mM Tris-HCl, 0.5 M NaCl, pH 7.5, and a final incubation with 4-chloronaphthol and H2O2.
A horizontal strip corresponding to the antigenic bands of 59-65 kDa was excised from unstained blots and used for the affin-ity purification of antibodies (Olmsted, 1981). The strip was incu-bated for 3 hours at 35°C with antibody (20 to 100 μg/ml) diluted in PBS-Tween and washed during 5 to 10 minutes (3 changes) in PBS-Tween. The specific antibodies were eluted by incubating the strip for 2 minutes at room temperature in 3 ml 0.2 M glycine-HCl, pH 2.8, containing 0.01 mg/ml gelatin. The eluted material was immediately neutralized (final pH was 7.0 to 7.7) with a titrated volume (200 μl) of 1 M Tris-base containing 1.6 mg/ml sodium azide and stored at 4°C. The NCM strip was rinsed with PBS-Tween containing 0.1 mg/ml sodium azide and reused for several absorption cycles. Concentration of affinity-purified anti-bodies was accomplished by dialysis under vacuum. The same technique was used to elute antibodies binding to other antigenic bands present on our immunoblots (see below).
Immunocytochemistry on bone tissue
Four-day-old Wistar rat pups, 5-day-old C57 mouse pups or 10-day-old New Zealand rabbits were perfused via the femoral artery with PBS for 1 minute followed by paraformaldehyde (2%)/lysine (0.75 M)/sodium periodate (0.01 M) (PLP) for 5 minutes. The proximal tibiae were dissected out and slices cut out of the pri-mary spongiosa area under the growth plate. The slices were fixed in PLP for an additional 4 hours at 4°C and then washed in PBS containing 10% DMSO as a cryoprotectant. The tissue was sub-sequently quick frozen and 40 μm sections were prepared on a Bright cryostat (Huntingdon, England) using tungsten carbide knives. Some sections were returned to PLP fix containing 4% EDTA and 5% PVP for 2 hours at 4°C with agitation. The remain-ing sections were rapidly decalcified by incubation either (rou-tinely) in 4% EDTA, 0.44% NaOH, 5% PVP or (when indicated) in 40% formic acid, 8% sodium formate (Vermeulen et al., 1989) for 2 to 16 hours at 4°C with agitation. In the experiments designed to detect intracellular signals, the specimens were treated (90 minutes at 37°C) with chondroitinase ABC at a concentration of 0.24 unit/ml PBS in order to enhance the immunoreactivity of the antigen (Pelletier et al., 1990). All sections were washed (2 hours) in PBS and incubated overnight in the respective primary antibodies diluted at about 40 μg/ml PBS + 0.1% BSA.
After washing (2 hours) in PBS + 0.1% BSA, the sections were incubated with Fab fragments of peroxidase-labeled rabbit anti-goat IgG (Biosys, France) for the mouse anti-collagenase primary, goat anti-rabbit IgG for the rat anti-collagenase primary and goat anti-rabbit for the anti-collagen primary. Sections were incubated at a dilution of 1:100 in PBS + 1% BSA for 2 hours at 20°C. After washing, the sections were reacted in DAB (1 mg/ml in 0.05 M Tris buffer, pH 7.4, Polysciences) in the presence of 0.1% H2O2 and post-fixed in ferrocyanide-reduced OsO4. After embedding in Epon (Polybed 812, Polysciences, Inc., Warrington, PA) 1 μm thick sections were cut with a glass knife and counterstained with methylene blue-azure II for identification of areas of interest. Selected areas were then sectioned with a diamond knife and stained with lead citrate. Grids were viewed on a JEOL-CX 100 electron microscope.
Immunocytochemistry on isolated bone cells
Cells were isolated from long bones of rat pups or 10-day-old rab-bits as previously described (Ali et al., 1984) and allowed to settle on cortical bone slices or on glass coverslips. After 30 minutes non-adherent cells were discarded by shaking the slices vigor-ously. The remaining cells were cultured in fresh α-MEM medium supplemented with 10% heat-inactivated fetal calf serum. After 18 hours of culture, the specimens were processed for immuno-cytochemistry at room temperature. They were fixed for 10 min-utes in 3.7% formaldehyde, washed in PBS, incubated for 30 min-utes in PBS containing 0.5% BSA and 0.05% saponin (to permeabilize the cells), and thereafter for 90 minutes in the pri-mary antibody diluted in the latter buffer at a concentration of about 40 μg/ml. This was followed by a washing in the same buffer (30 minutes) and an incubation (1 hour) in the same buffer containing rhodamine-labeled rabbit anti-goat IgG for the mouse collagenase primary antibody (Cappel-Organon Teknika Corp., West Chester, PA) and FITC-labeled goat anti-sheep IgG for the rabbit collagenase primary (Boehringer, Mannheim, Germany) at a dilution of 1:100. The specimens were washed, mounted and observed by epifluorescence, either on a Zeiss Axiophot micro-scope (with a 546 nm excitation filter and a 590 nm arrest filter) or on a Bio-Rad confocal microscope equipped with an argon-krypton laser beam. Confocal images were collected as the aver-age of 9 to 15 scans of one optical section of 2 to 3 μm for cells viewed at a ×40 magnification and 0.5 to 1 μm for cells viewed at a ×63 magnification. The images are stored on optical disks and viewed on a Gateway 2000 or NEC Multisync computer system. The images are optimized by subtracting the background and by normalizing to the maximum pixel intensity of that image.
RESULTS
Characterization of the antibodies to mouse bone collagenase
Immune IgG, but not preimmune ones, inhibited and pre-cipitated bone collagenase from mouse (Table 1) and rat (not shown). Antibody specificity was checked by Ouchter-lony double diffusion, crossed immunoelectrophoresis, western blot analysis and immunoaffinity chromatography.
The IgG gave a single precipitation line on double diffu-sion with crude preparations of latent mouse procollage-nase, but two lines with activated collagenase (not shown). Western blot analysis of crude latent procollagenase showed that immune IgG, but not preimmune ones, reacted with a characteristic doublet of bands (65 and 59 kDa) (Fig. 1A, lanes T1 and T3; Fig. 2A, lane +), corresponding to different glycoforms of procollagenase (see, e.g., Nagase et al., 1983). A doublet (50 and 45 kDa), corresponding to activated collagenase, and a few bands of lower molecular mass (26 and 20 kDa), corresponding presumably to frag-ments resulting from proteolytic cleavage (as reported in several other studies: see, e.g., Clark and Cawston, 1989), appeared upon activation or storage of (pro)collagenase and in heavily loaded blots (not shown). It was checked that the antibodies eluted from each band recognized all the others (not shown), thereby further establishing that these bands have at least one common epitope and may thus well cor-respond, respectively, to procollagenase, collagenase and degradation products. The IgG eluted from the 59-65 kDa bands did also precipitate collagenase (Table 1).
Characterization of the anti-(pro)collagenase IgG by immunoaffinity chromatography. Anti-collagenase and preimmune immunoaffinity columns (see Materials and Methods) were loaded with total antigen (T) and sequentially eluted with CCTN buffer containing 0.15 M NaCl (fraction 1), 1 M NaCl (fraction 2) and 3 M KSCN (fraction 3). Samples of total antigens were diluted in CCTN buffer containing either 0.15 M NaCl (T1) or 3 M KSCN (T3). Sample T3 and fractions 3 were dialyzed against 0.15 M NaCl in CCTN buffer prior to analysis by immunoblotting (A), gelatin zymography (B) and casein zymography (C). The molecular mass standards are shown in lanes S.
Characterization of the anti-(pro)collagenase IgG by immunoaffinity chromatography. Anti-collagenase and preimmune immunoaffinity columns (see Materials and Methods) were loaded with total antigen (T) and sequentially eluted with CCTN buffer containing 0.15 M NaCl (fraction 1), 1 M NaCl (fraction 2) and 3 M KSCN (fraction 3). Samples of total antigens were diluted in CCTN buffer containing either 0.15 M NaCl (T1) or 3 M KSCN (T3). Sample T3 and fractions 3 were dialyzed against 0.15 M NaCl in CCTN buffer prior to analysis by immunoblotting (A), gelatin zymography (B) and casein zymography (C). The molecular mass standards are shown in lanes S.
Characterization of the anti-(pro)collagenase IgG with conditioned medium of calvaria cultured in the presence or absence of heparin. Newborn mouse calvaria were cultured with (+) or without (−) 0.3 mg/ml heparin (Lenaers-Claeys and Vaes, 1979). Identical amounts of conditioned media (+) and (−) containing, respectively, 10 units and less than 0.1 unit (pro)collagenase/ml were analyzed by immunoblotting (A), gelatin zymography (B) and casein zymography (C), as described in Materials and Methods. The molecular mass standards are shown in lanes S.
Characterization of the anti-(pro)collagenase IgG with conditioned medium of calvaria cultured in the presence or absence of heparin. Newborn mouse calvaria were cultured with (+) or without (−) 0.3 mg/ml heparin (Lenaers-Claeys and Vaes, 1979). Identical amounts of conditioned media (+) and (−) containing, respectively, 10 units and less than 0.1 unit (pro)collagenase/ml were analyzed by immunoblotting (A), gelatin zymography (B) and casein zymography (C), as described in Materials and Methods. The molecular mass standards are shown in lanes S.
The specificity of the antibodies towards (pro)collage-nase was further examined by immunoaffinity chromatog-raphy. Fig. 1 shows that the immobilized immune IgG, con-trary to the preimmune ones, retain the characteristic doublets of procollagenase and of collagenase (A-C) and that they do not retain the 72-65 kDa gelatinase A (MMP-2), the 96 kDa gelatinase B (MMP-9) (see B), or the caseinolytic stromelysin (MMP-3) (approximately 53 kDa, see Fig. 1C). Biochemical assays confirmed that under these conditions 90 to 100% of the (pro)collagenase activity was retained by the immobilized immune IgG but not by the preimmune ones (not shown). The selective binding of the antibodies to (pro)collagenase was also checked (Fig. 2) by comparing conditioned media that contained either high amounts (lanes +) or 100-fold lower amounts (lanes −) of (pro)collagenase but similar (MMP-2 and MMP-9; B) or less different (MMP-3; C) activities of other matrix metal-loproteinases. The absence of signal in lane (−) of the immunoblot (A) demonstrates that the anti-(pro)collagenase IgG do not bind to these other metalloproteinases.
Immunolocalization of collagenase in the extracellular matrix
These results therefore demonstrate that the antiserum raised against mouse bone collagenase recognizes the enzyme but does not recognize closely related metallopro-teinases. It could therefore be used to immunolocalize col-lagenase in the growth plate areas of tibiae from neonatal rats and mice. In the first series of experiments, the sec-tions were incubated with the antibodies without prior digestion with chondroitinase. Using peroxidase-conjugated secondary antibodies, we found the reaction product to be localized and restricted mainly to the interface between osteoclasts and the bone matrix, in the bone-resorbing com-partment (Fig. 3). No labeling was found along other bone surfaces, which are mostly lined with osteoblasts in these growing animals (Fig. 3). The purified IgG fraction from pre-immune serum failed to show any reaction product in tissues processed in the same manner (Figs 3, 5).
Light microscopic immunolocalization of collagenase and collagen type I in mouse bone. (A) Anti-(mouse) collagenase IgG: the peroxidase reaction product (dark band) is found exclusively at the interface (arrowheads) between osteoclasts (oc) and the bone matrix (bm); no reaction product is found in osteoblasts (ob) or along the adjacent bone matrix. (B) Preimmune serum showing the absence of reaction product along the bone matrix at the interface (arrowheads) with osteoclasts (oc). (C) Anti-(rat) collagen type I: the reaction product is found along the whole bone matrix and is not restricted to any interface with a specific cell type. Bar, 20 μm.
Light microscopic immunolocalization of collagenase and collagen type I in mouse bone. (A) Anti-(mouse) collagenase IgG: the peroxidase reaction product (dark band) is found exclusively at the interface (arrowheads) between osteoclasts (oc) and the bone matrix (bm); no reaction product is found in osteoblasts (ob) or along the adjacent bone matrix. (B) Preimmune serum showing the absence of reaction product along the bone matrix at the interface (arrowheads) with osteoclasts (oc). (C) Anti-(rat) collagen type I: the reaction product is found along the whole bone matrix and is not restricted to any interface with a specific cell type. Bar, 20 μm.
To check that this apparently specific distribution was not due to better access of the antibodies to the antigen in these areas, we performed control experiments with anti-bodies to collagen type I, a molecule that is both abundant and ubiquitous in bone matrix. The results clearly showed (Fig. 3) that our procedures allow antibodies to reach all of the interface between the bone matrix and the cells, although the diffusion of the antibodies in the matrix itself is limited (Fig. 3). In addition, these patterns of localiza-tion were altered neither by decreasing the decalcification time to 2 hours nor by including fixative solution in the decalcification buffer, thereby making it unlikely that selec-tive extraction of the enzyme led to the observed localization. An even stronger labeling of the subosteoclastic resorption zones was found when substituting formic acid for EDTA as decalcifying agent (Vermeulen et al., 1989). We interpreted these results to indicate that the antigen recognized by the antiserum was indeed restricted, along bone surfaces, to the subosteoclastic bone-resorbing compartment.
Despite the apparent specificity of the whole antiserum, we further verified the specificity of our localization exper-iments in two different ways. Firstly, we used an IgG fraction purified by repeated adsorption and elution cycles on western blot strips corresponding to the 59-65 kDa anti-genic bands, as explained in Materials and Methods. When these affinity-purified antibodies were used for immunolo-calization under the same conditions as before, we found that they localized to exactly the same regions as the whole antiserum: namely, the bone matrix in the subosteoclastic bone-resorbing compartment (Figs 4, 5, 6). Interestingly, and both with the whole IgG fraction and with the affin-ity-purified antibodies, collagenase was found in close asso-ciation with collagen fibers, whether of type I in bone or type II in cartilage, but only when osteoclasts were in the process of resorbing these matrices (Fig. 5). Second, immunolocalizations were performed with two other well-characterized antisera, directed against rat and rabbit colla-genase, respectively, and prepared using other experimen-tal protocols (Blair et al. 1986b; Brinckerhoff et al., 1990). No difference in localization of the enzyme was found between mouse, rat (Fig. 7A) and rabbit bone (not shown). We concluded from these experiments that the antigen localized in the bone-resorbing compartment is indeed col-lagenase and that it is present in the bone-resorbing com-partment of several animal species.
Immunolocalization of affinity-purified anti-(mouse) collagenase IgG in the bone-resorbing compartment underlying mouse osteoclasts. The reaction product (A, arrows) is found exclusively in the area between the ruffled border of the cell (rb) and the bone matrix (bm); the attachment area (sealing zone, sz) shows no reaction product (curved open arrows); in (B), the area of the ruffled border and of the underlying bone matrix is shown at higher magnification; n, nuclei. Bars: (A) 1.5 μm; (B) 0.5 μm.
Immunolocalization of affinity-purified anti-(mouse) collagenase IgG in the bone-resorbing compartment underlying mouse osteoclasts. The reaction product (A, arrows) is found exclusively in the area between the ruffled border of the cell (rb) and the bone matrix (bm); the attachment area (sealing zone, sz) shows no reaction product (curved open arrows); in (B), the area of the ruffled border and of the underlying bone matrix is shown at higher magnification; n, nuclei. Bars: (A) 1.5 μm; (B) 0.5 μm.
Affinity-purified anti-(mouse) collagenase IgG (A and C), but not preimmune IgG (B and D) localize in the bone-resorbing compartment under the ruffled border area (left) whether the osteoclast is resorbing cartilage collagen type II (A and B) or bone collagen type I (C and D). Bar, 200 nm.
Immunolocalization of affinity-purified anti-(mouse) collagenase IgG in mouse bone sections that have been pre-digested with chondroitinase ABC. Chondroitinase digestion allows intracellular localization of collagenase in osteoclasts (A) and in osteoblasts (B). (A) The staining of the subosteoclastic bone-resorbing compartment is reinforced but is still restricted (arrow), leaving other bone surfaces unstained; in addition, intracellular structures present between the nuclei (n) and the ruffled border are now stained. (B) Strong intracellular staining in the Golgi region (arrows) near the nuclei (n) of osteoblasts is also prominent. Several other connective tissue cells contain collagenase-positive structures. Bar, 10 μm.
Immunolocalization of affinity-purified anti-(mouse) collagenase IgG in mouse bone sections that have been pre-digested with chondroitinase ABC. Chondroitinase digestion allows intracellular localization of collagenase in osteoclasts (A) and in osteoblasts (B). (A) The staining of the subosteoclastic bone-resorbing compartment is reinforced but is still restricted (arrow), leaving other bone surfaces unstained; in addition, intracellular structures present between the nuclei (n) and the ruffled border are now stained. (B) Strong intracellular staining in the Golgi region (arrows) near the nuclei (n) of osteoblasts is also prominent. Several other connective tissue cells contain collagenase-positive structures. Bar, 10 μm.
Immunolocalization of collagenase in rat tissue (A) using antibodies to rat collagenase (Jeffrey et al., 1990) and peroxidase-conjugated secondary antibodies, and in isolated rabbit osteoclasts on glass (B) using antibodies to rabbit collagenase and rhodamine-conjugated secondary antibodies, and confocal microscopy. In both instances, the extracellular (A, arrows) and intracellular (B) localizations of collagenase are identical to those observed using antibodies to mouse collagenase and mouse or rat bone cells (compare with Figs 3 and 8, respectively) (Bars, 15 μm).
Immunolocalization of collagenase in rat tissue (A) using antibodies to rat collagenase (Jeffrey et al., 1990) and peroxidase-conjugated secondary antibodies, and in isolated rabbit osteoclasts on glass (B) using antibodies to rabbit collagenase and rhodamine-conjugated secondary antibodies, and confocal microscopy. In both instances, the extracellular (A, arrows) and intracellular (B) localizations of collagenase are identical to those observed using antibodies to mouse collagenase and mouse or rat bone cells (compare with Figs 3 and 8, respectively) (Bars, 15 μm).
Immunolocalization of collagenase in bone cells
The confinement of collagenase to the resorption zones strongly suggests that the osteoclasts are at the source of this collagenase. However, this would imply its presence in the biosynthetic and secretory pathways of the osteoclasts, i.e. intracellularly. Since we did not find intracellular col-lagenase with the experimental protocol used hereabove, we varied the experimental conditions in order to determine whether collagenase was also present within osteoclasts. These involved confocal immunomicroscopy of isolated osteoclasts cultured on bone slices or on glass coverslips and enzymatic digestion of tissue sections prior to incuba-tion with the anti-collagenase antibodies (Pelletier at al., 1990; Aeschlimann et al., 1993).
Bone cells were isolated from long bones of newborn rats, cultured for 18 hours on devitalized cortical bone slices, fixed and incubated with anti-collagenase antibod-ies. Using rhodamine-conjugated secondary antibodies, we found a prominent fluorescence in the osteoclasts (Fig. 8) and a weaker signal in many uncharacterized mononuclear cells (not shown). In the osteoclasts that were excavating bone, prominent fluorescence was visible intracellularly, in perinuclear areas, towards the bone surface and even in regions located below the level of the surface of the sur-rounding bone (i.e. in the part of the cell present within the resorption pit) (Fig. 9). These observations thus demon-strated the presence of collagenase in resorbing osteoclasts, from the perinuclear area towards the resorption compart-ment. In osteoclasts that were not associated with pits or that were plated on glass, and even after up to 24 hours, collagenase was also visible as intracellular punctate stain-ing particularly prominent in perinuclear areas, in rabbits (Fig. 7B) as well as in rats. This localization was specific for collagenase, since it could be reproduced using affin-ity-purified antibodies but not when using the IgG fraction from preimmune serum (Fig. 8).
Immunolocalization of collagenase in isolated rat bone cells cultured on a bovine cortical bone slice. The specimens were analyzed in epifluorescence on a conventional microscope (A, B) or on a confocal microscope (C, D). Anti-(mouse) collagenase IgG (either affinity-purified IgG in (A) or total IgG in (C) and (D)) but not preimmune IgG (B),localize in the perinuclear area of osteoclasts (in B, small and large arrows indicate, respectively, the outline and the nuclei of an osteoclast). Intracellular localization is further demonstrated in confocal optical sections (C and D): in (C), the section is taken at the level of the bone slice surface and collagenase is found in several osteoclasts, most prominently the one at the bottom of the field; in (D), the same cell is shown at a higher magnification and at a higher level of optical sectioning than in (C), further above the bone surface, to demonstrate intracellular localization at the level of the nuclei. Bars: (A-C) 10 μm; (D) 2 μm.
Immunolocalization of collagenase in isolated rat bone cells cultured on a bovine cortical bone slice. The specimens were analyzed in epifluorescence on a conventional microscope (A, B) or on a confocal microscope (C, D). Anti-(mouse) collagenase IgG (either affinity-purified IgG in (A) or total IgG in (C) and (D)) but not preimmune IgG (B),localize in the perinuclear area of osteoclasts (in B, small and large arrows indicate, respectively, the outline and the nuclei of an osteoclast). Intracellular localization is further demonstrated in confocal optical sections (C and D): in (C), the section is taken at the level of the bone slice surface and collagenase is found in several osteoclasts, most prominently the one at the bottom of the field; in (D), the same cell is shown at a higher magnification and at a higher level of optical sectioning than in (C), further above the bone surface, to demonstrate intracellular localization at the level of the nuclei. Bars: (A-C) 10 μm; (D) 2 μm.
Confocal immunolocalization of collagenase throughout the cytoplasm of isolated rat osteoclasts cultured on a bovine cortical bone slice. The specimens were analyzed under epifluorescence on a confocal microscope. Four optical sections (each 0.6 μm thick) were made every 7 μm and are shown in A (upper section) to D (lower section). (A) The highest section is taken at the level of the bone surface, as illustrated by a neighboring mononuclear cell (arrow), which disappears in (C); (C) and (D) are below the bone surface, within the resorption pit. This series illustrates the fact that collagenase is present in the osteoclast, and between the nuclei and the ruffled border. Bar, 10 μm.
Confocal immunolocalization of collagenase throughout the cytoplasm of isolated rat osteoclasts cultured on a bovine cortical bone slice. The specimens were analyzed under epifluorescence on a confocal microscope. Four optical sections (each 0.6 μm thick) were made every 7 μm and are shown in A (upper section) to D (lower section). (A) The highest section is taken at the level of the bone surface, as illustrated by a neighboring mononuclear cell (arrow), which disappears in (C); (C) and (D) are below the bone surface, within the resorption pit. This series illustrates the fact that collagenase is present in the osteoclast, and between the nuclei and the ruffled border. Bar, 10 μm.
Since these observations clearly established the presence of collagenase in isolated osteoclasts, we then modified our experimental procedures in order to determine whether we could detect intraosteoclastic collagenase in the tissues. When the sugar moieties were digested by treating the sections with chondroitinase ABC prior to incubation with the anti-collagenase antibodies (Pelletier et al., 1990; Aeschli-mann et al., 1993), there was some decrease in the quality of the pictures, but under these conditions, intracellular labeling became apparent not only in the osteoclasts, but also, and as described by others (Blair et al., 1989b), in osteoblasts, chondroblasts and other uncharacterized mononuclear cells (Fig. 6). The intracellular distribution of the label was consistent with that found in the resorbing isolated osteoclasts, with the highest accumulation in the area between the nuclei and the ruffled border. Furthermore, the labeling of the extracellular resorption zone appeared still stronger and staining was still restricted to these areas of the bone matrix.
DISCUSSION
This study demonstrates that collagenase is present in the bone-resorbing compartment underlying the osteoclasts as well as in the osteoclast itself, independently of it being in the process of resorbing bone or not. This specific local-ization was observed in three different species and using three different well-characterized antibodies against the enzyme. Since it has also been shown that bone resorption depends, at least in part, on the activity of collagenase in the bone-resorbing compartment (Everts et al., 1992) and that resorbing odontoclasts, a closely related cell-type, express mRNA for this enzyme (Okamura, 1992; Okamura et al., 1993), it is tempting to speculate that collagenase is secreted by the osteoclast and involved in the process of bone resorption.
The identification of the antigen present in the osteoclasts and in the bone-resorbing compartment as collagenase is strongly established by the extensive characterization of our antibodies, their inability to cross-react with three metallo-proteinases that are closely related to collagenase (72 kDa gelatinase A (MMP-2), 96 kDa gelatinase B (MMP-9) and stromelysin (MMP-3)), and the use of IgGs that were affin-ity-purified by adsorption on 65-59 kDa procollagenase antigens, eliminating potential cross-reactions with matrilysin (MMP-7) or the tissue inhibitors of metallopro-teinases, which all migrate between 28 and 22 kDa. The fact that the antigen is indeed collagenase is further, and in our eyes definitively, established by the observation that antibodies against rat collagenase or rabbit collagenase, raised and characterized independently (Jeffrey et al., 1990; Brinckerhoff et al., 1990), gave identical results in these two species.
The fact that the collagenase specifically recognized by all three antibodies is present in the bone-resorbing com-partment is also well established. All the preparations of anti-collagenase antibodies used in this study provided identical and specific staining of the bone-resorbing compartment in mouse, rat and rabbit bone, leaving other bone surfaces unstained. This was in contrast with the results obtained with antibodies to collagen type I, which stained the bone matrix surface irrespective of its association with osteoclasts. Preimmune preparations were negative in all cases.
Finally, the fact that collagenase is found within the osteoclast is also well established by our study. Total anti-collagenase IgG as well as affinity-purified IgG, but not preimmune IgG, provided the same specific fluorescence of the intracellular areas of the osteoclasts and, here again with all three antibodies and in all three species. The intracellu-lar distribution of the fluorescence observed with the con-focal microscope in resorbing isolated osteoclasts was in good accord with that of the peroxidase reaction product observed with the electron microscope in chondroitinase-treated tissue sections. Since the intracellular signal appears strong in the perinuclear area of the osteoclasts and inde-pendently of whether the cells are resorbing (on bone) or not (on bone or on glass), it cannot be merely explained by the presence of extracellular collagenase in deep invaginations of the ruffled border or by the internalization of extracellular collagenase. Furthermore, we have previously shown that most of the intracellular vesicles in the osteoclast are constituents of the secretory rather than endocytic pathways (Baron et al., 1985, 1988). Indeed, the intracellular distribution of collagenase observed here by confocal microscopy corresponds well to that of secretory enzymes in these cells (Baron et al., 1985, 1988). It is therefore tempting to speculate that newly synthesized collagenase is indeed within the biosynthetic pathway, transported towards the ruffled border and secreted into the bone-resorbing compartment. This hypothesis, i.e. that osteoclasts synthesize and secrete (pro)collagenase, is further supported by recent in situ hybridization data showing collagenase mRNA in resorbing bovine odontoclasts (Okamura, 1992; Okamura et al., 1993), a cell type involved in the resorption of the mineralized matrix of teeth and thought to be identical to osteoclasts. Furthermore, we (unpublished) and others (Case et al., 1989) have found that osteoclasts also contain stromelysin (MMP-3), a metalloproteinase that is closely related to collagenase (but not recognized by our antibodies) and that is implicated in the activation cascade of col-lagenase (Brinckerhoff et al., 1990; Nagase et al., 1991). Hence, given this array of independent observations, the conclusion that osteoclasts are capable of synthesizing and secreting metalloproteinases (both collagenase and stromelysin) seems warranted.
This conclusion, together with the results of Everts et al. (1992) showing that inhibition of collagenase affects colla-gen degradation in the subosteoclastic zone, suggests a role for metalloproteinases in bone resorption. As reviewed else-where (Vaes, 1988; Delaissé and Vaes, 1992), both lyso-somal enzymes and collagenase are involved in this process. The participation of lysosomal enzymes, particularly the collagenolytic cysteine proteinases cathepsin L (EC 3.4.22.15) and cathepsin B (EC 3.4.22.1) (both present in resorbing bone; Delaissé et al., 1991a), is likely to occur at the level of the bone-resorbing compartment and their inhi-bition prevents the resorption of bone collagen by the osteo-clast (Delaissé et al., 1987; Everts et al., 1988, 1992; Rifkin et al., 1991). Specific collagenase inhibitors also inhibit bone resorption, either in vitro, when acting in organ cul-tures of bone (Delaissé et al., 1985), or in vivo (Delaissé et al., 1991b). In view of the inability of others to detect collagenase in osteoclasts, it was proposed that collagenase was produced by osteoblasts and was responsible for the removal of non-mineralized collagen, which, when present, would prevent the adherence and activation of osteoclasts (Chambers et al., 1985). The presence of (pro)collagenase in the osteoclasts and in the underlying bone-resorbing compartment, established by the present study, suggests a much simpler explanation, i.e. that osteoclastic collagenase could also be active in the removal of the mineralized col-lagen during or after demineralization by the osteoclast. This hypothesis is indeed strongly supported by the fact that broad fringes of demineralized collagen are seen under osteoclasts after treatment with collagenase inhibitors (Everts et al., 1992), as they are also after the inhibition of lysosomal cysteine proteinases (Everts et al., 1988, 1992). This indicates that under these conditions demineralization of bone by osteoclasts proceeded up to a certain point whereas matrix degradation was inhibited. Thus collage-nase is not only present in the subosteoclastic resorption zone, as shown in the present study, but is also active at that level, together with cysteine proteinases.
In trying to understand the respective roles of collage-nase and cysteine proteinases in bone resorption, one could envision the following hypothesis. Besides their col-lagenolytic action, lysosomal cysteine proteinases could generate active collagenase from its zymogen, as cathepsin B is known to activate procollagenase (Eeckhout and Vaes, 1977). Collagenase could then co-operate with col-lagenolytic cysteine proteinases in the degradation of the demineralized collagen, a process that is likely to be ren-dered more efficient by the high concentration of Ca2+ present in the bone-resorbing compartment as a conse-quence of bone demineralization (Etherington and Birkedal-Hansen, 1987; Eeckhout, 1990). Because the optimal pH for these two classes of enzymes are very different (6.0 to 7.5 for collagenase (Vaes, 1972), but 4.5 for the cysteine proteases (Delaissé et al., 1991a)), it may be speculated that the combined action of these two classes of enzymes would broaden the pH spectrum at which matrix resorption can occur. Indeed the effective pH at any time and at any point of the resorption zone will depend both on the relative effi-cacy of the proton secretion and on the buffering capacity exerted by the solubilized bone salts. Thus for instance, the lysosomal cysteine proteinases might act predominantly in the immediate vicinity of the ruffled border where protons are secreted, whereas the collagenolytic action of collage-nase might be predominant in the presumably more neutral zone, localized at the interface between the mineralized and demineralized matrix. Also, the activation of collagenase by lysosomal enzymes would be favored in the latter zone, as this process is more efficient around pH 6 than at lower pH (Vaes, 1972; Eeckhout and Vaes, 1977). Moreover col-lagenase could be further required to degrade the fringe of yet undegraded but already demineralized collagen that is left behind by the osteoclast when it detaches to move along the bone surface, thereby rendering the pH neutral in the bone-resorbing lacuna. Such roles for collagenase and cys-teine proteinases would explain both the accumulation of collagen under osteoclasts in the presence of collagenase and cysteine proteinase inhibitors (Everts et al., 1992) and the finding that the process of matrix resorption seems to go on even after the osteoclast has moved away (Gaillard, 1957).
In conclusion, and independently of speculation on its physiological role in bone resorption, our results support the concept that collagenase is present in the biosynthetic pathway of the osteoclast and secreted into the bone-resorb-ing compartment, where it participates in the bone-resorption process.
The immunocytochemical part of the present work was done at Yale University, in part by J.M.D., who was supported by a travelling grant from the National Fund for Scientific Research (Belgium). The authors are most grateful to: Dr J. Madri (Yale University) for providing the antibodies to collagen type I; Dr J. Jeffrey (The Albany Medical College of Union University) for the antibodies to rat collagenase; Dr C. Brinckerhoff (Dartmouth Uni-versity School of Medicine) for the antibodies against rabbit col-lagenase; Dr S. Wilhelm (Miles Research Center, West Haven, CT) for the antibodies to stromelysin; Pierre Courtoy and Nicola Partridge for helpful discussions and suggestions; Oskar Hoffman for providing isolated rabbit osteoclasts; Agnès Deleruelle, Pas-cale Lemoine and Marie-Christine Baelden, who provided their expert technical assistance; Iris Colon and Yves Marchand who typed the manuscript. This work was supported by grants from the NIH (DE-04724) to R.B., from the fund for Medical Scien-tific Research (Belgium) to G.V. and Y.E. and from the Belgian State, Prime Minister’s Office, Science Policy Programming (Interuniversity Poles of Attraction) to the International Institute of Cellular and Molecular Pathology. P.H. was a Research Fellow of the Institut pour l’Encouragement de la Recherche Scientifique dans l’Industrie et l’Agriculture. Part of this work was presented at the 12th Meeting of the American Society for Bone and Min-eral Research, Atlanta, August 28-31, 1990, at the International Conference on Calcium Regulating Hormones, Florence, Italy, April 24-29, 1992, and at the 13th Meeting of the Federation of European Connective Tissue Societies, Davos, Switzerland, July 12-17, 1992.