In this study we investigated the role of bone lining cells in the coordination of bone resorption and formation. Ultrastructural analysis of mouse long bones and calvariae revealed that bone lining cells enwrap and subsequently digest collagen fibrils protruding from Howship's lacunae that are left by osteoclasts. By using selective proteinase inhibitors we show that this digestion depends on matrix metalloproteinases and, to some extent, on serine proteinases. Autoradiography revealed that after the bone lining cells have finished cleaning, they deposit a thin layer of a collagenous matrix along the Howship's lacuna, in close association with an osteopontin-rich cement line. Collagenous matrix deposition was detected only in completely cleaned pits. In bone from pycnodysostotic patients and cathepsin K-deficient mice, conditions in which osteoclastic bone matrix digestion is greatly inhibited, bone matrix leftovers proved to be degraded by bone lining cells, thus indicating that the bone lining cell “rescues” bone remodeling in these anomalies. We conclude that removal of bone collagen left by osteoclasts in Howship's lacunae is an obligatory step in the link between bone resorption and formation, and that bone lining cells and matrix metalloproteinases are essential in this process.
Remodeling of bone is crucial for maintaining a well-adapted skeleton. This process involves two major steps: resorption of existing bone by osteoclasts and formation of new bone by osteoblasts. It is generally taken that cells of the osteoblast lineage play an essential role in orchestrating bone remodeling. These cells deposit new bone matrix (osteoid) and regulate bone-resorbing activity of the osteoclast by producing a wide range of compounds (e.g., prostaglandin E2 [PGE2], interleukin [IL]-1, IL-6, receptor activator of NF-κB ligand [RANKL]).(1–4) Before resorption by osteoclasts, osteoblasts are considered to remove nonmineralized osteoid from the bone surface with the use of matrix metalloproteinases (MMPs).(5, 6) After resorption, formation of new bone is initiated, resulting in a finely tuned remodeling of bone mass.(7) This event, the coupling between bone resorption and bone formation, is also thought to be mediated by the osteoblast.(8–10)
For proper understanding of the regulation of bone formation after resorption, it is essential to know the sequence of events that occur in the resorption pit after the osteoclast has exerted its activity. It has been shown that the pits are filled again with a layer of new bone. Some authors even proposed that formation of new bone preferentially occurs at these sites.(11) Thus, the resorption pit appears to be an important site where various remodeling activities take place. According to a number of authors, osteoblasts are attracted to the Howship's lacuna by signaling molecules such as transforming growth factor-β (TGF-β),(12–14) insulin-like growth factor (IGF),(15) and/or osteopontin.(16) Yet, the processes that actually take place in the resorption pit are poorly described and not well understood. An important question is whether deposition of new matrix occurs directly after withdrawal of the osteoclast or that the resorbed surface has to be cleaned first. Several studies show that resorption pits left by osteoclasts still contain nondigested demineralized bone collagen.(17–21) So far, it is unclear whether these matrix constituents have to be removed before formation of new bone, and if so, by which cell type(s). Tran Van and coworkers(17) suggested that fringes of nondigested collagen are removed by macrophages, whereas Heersche(22) and Rifkin and Heyl(23) considered the fibroblast responsible for cleansing activity at this site. On the other hand, electron microscopic studies of Takahashi and coworkers(24) showed that osteoblast-like cells enwrap and phagocytose nonmineralized collagen from the surface of calvarial bone.
It was the aim of this study to analyze in detail the phenomena taking place in Howship's lacunae after withdrawal of the osteoclast. To this end we analyzed by microscopy resorptive activity along the bone surface of calvariae and long bones in the presence or absence of selective inhibitors of the activity of MMPs, cysteine proteinases, serine proteinases, and aspartic proteinases. Bones from patients suffering from pycnodysostosis and from cathepsin K-deficient mice were used to analyze the events under conditions in which areas of nondigested bone matrix are present in excessive amounts in pits in vivo.(25, 26) Finally, we investigated deposition of new matrix along the bone surface of cultured bone explants by autoradiography.
MATERIALS AND METHODS
Parathyroid hormone (PTH; 69.6 μg PTH/mg solid), calcitonin, pepstatin, aprotinin, E-64, fluorescein isothiocyanate (FITC)-conjugated sheep anti-mouse immunoglobulin G (IgG), and xylenol orange were obtained from Sigma Chemical (St. Louis, MO, USA). Medium 199 and fetal calf serum were from Gibco (Grand Island, NY, USA). The MMP-inhibitor CI-1(27) was a gift from Dr. C. Sear (Searle Research and Development, High Wycombe, UK). A rat antibody specific for mouse monocytes and macrophages (MOMA-2) was obtained from Serotec (Oxford, UK). FITC-labeled goat anti-rat IgG was obtained from Boehringer-Mannheim Biochemicals (Mannheim, Germany). Goat anti-rat osteopontin was obtained from Dr. W. Butler (Houston, TX, USA), mouse anti-human osteocalcin was a kind gift from Dr. P. Cloos of OsteoMeter (Herlev, Denmark), goat anti-human type I collagen was from Southern Biotechnology Associates (Birmingham, AL, USA), and goat anti-mouse ICAM-1 (CD54) from Santa Cruz Biotechnology (Santa Cruz, CA, USA). For the detection of osteocalcin the Histomouse-SP Kit of Zymed Laboratories (San Francisco, CA, USA) was used. Protein A-gold conjugate (10 nm), streptavidin-gold conjugate (10 nm), bovine serum albumin (BSA)-C, and silver enhancement kit were from Aurion (Wageningen, The Netherlands). All other reagents were of analytical grade.
Calvariae (frontal and parietal bones) and meta-tarsals of 5-day-old NMRI mice were aseptically removed and the calvariae were cultured on top of a filter paper floating on 400 μl medium 199, containing 2.5% fetal calf serum and with or without 0.1 μM PTH. The protease inhibitors were used in concentrations that have a maximal effect on resorption: E-64, 42 μM; CI-1, 40 μM; aprotinin, 30 μM; and pepstatin, 35 μM.(28) The explants were cultured for up to 72 h. To analyze mineral deposition in bone tissue, xylenol orange (125 μg/explant) was added to some explants. The explants were processed for microscopic examination after culture.
Calvarial and long bone explants were processed for light and electron microscopic examination as described previously.(29)
Semithin transverse sections (1-2 μm) were cut and stained with methylene blue or with a modified Goldner Trichrome staining method. Ultrathin sections were cut with a diamond knife, stained with uranyl and lead, and examined in a Zeiss EM-10 electron microscope (Zeiss, Oberkochen, Germany).
Morphometric analyses of the resorption areas were performed as described previously.(29) Data were expressed as mean square micrometer demineralized area per osteoclast or per calvarial section ± SEM of five or six explants.
Collagen enwrapment by bone lining cells was quantified by assessing the total number of bone lining cell profiles associated with the concave side of the calvariae and by establishing the percentage that had enwrapped collagen in the plane of the section. Bone lining cells were characterized by their long, slender, and flattened appearance; and their association with the bone surface at sites where a thin nonmineralized collagen layer was present. Each calvarial explant was represented by one ultrathin section and analyzed by a person not aware of the experimental condition. The data were expressed as mean% ± SD of five or six explants.
Calvarial bone explants were cultured for various periods of time in the presence or absence of proteinase inhibitors. A number of explants cultured for 24 h in the presence of a proteinase inhibitor were cultured for another 24 h in the absence of inhibitor. To all explants, [3H]proline was added in a final concentration of 1.25 μCi/well at 24 h before fixation. The explants were extensively washed in medium lacking labeled proline, and then fixed and embedded in epoxy resin.
Semithin sections of 2 μm were made and collected on gelatin-coated glass slides and dipped in L-4 emulsion (Ilford Ltd., Essex, UK). The slides were dried and stored at 4°C in the dark and after an exposure time of 1 week, autoradiographs were developed with D-19 developer (Kodak, Rochester, NY, USA). After fixation and rinsing, sections were stained with methylene blue.
Quantitative analysis of the autoradiographs was performed by counting the number of (1) osteoclasts, (2) resorption pits with or without nondigested bone matrix, (3) resorption pits covered with bone lining cells, and (4) labeled resorption pits.
For the localization of monocytes/macrophages with the marker MOMA-2 and for the localization of ICAM-1, bone tissue was isolated from mice, embedded in 8% gelatin, and frozen in liquid nitrogen. Cryosections were collected, fixed for 10 minutes with 4% paraformaldehyde, thoroughly washed in phosphate buffered saline (PBS) containing 1% BSA, and incubated with rat anti-mouse MOMA-2 (diluted 1:25) in PBS/BSA or goat anti-mouse ICAM-1 (diluted 1:200) in PBS/BSA for 60 minutes. The sections were washed in PBS/BSA and incubated with FITC-conjugated anti-rat IgG or FITC-conjugated anti-goat IgG for 45 minutes to detect MOMA-2 or ICAM-1, respectively. After washing, the sections were covered and examined with a light microscope equipped with epifluorescence. Control sections were incubated with nonimmune IgG of the same species.
Osteopontin, osteocalcin, and type I collagen were immunolocalized in semithin sections of bone explants embedded in LR-White.(30) The bones were fixed for 4 h in 4% paraformaldhyde in 0.1 M sodium cacodylate buffer, washed, decalcified in 0.1 M EDTA, dehydrated, and embedded in LR-White resin. Semithin sections were collected on glass slides, washed with PBS/BSA-C, and incubated overnight with anti-osteopontin (1:50), anti-osteocalcin (1:400), or anti-type I collagen (1:200). The sections were extensively washed and incubated with gold-labeled protein A or streptavidin-gold (both diluted 1:100) for 120 minutes. The sections incubated with anti-osteocalcin, before streptavidin labeling, were incubated with the Histomouse-SP kit. After washing, silver enhancement was performed on the sections according to the description provided by the manufacturer.
Alkaline phosphatase activity
Activity of alkaline phosphatase was localized with the method described by Van Noorden and Jonges(31) as modified by Groeneveld et al.(32)
Data were statistically analyzed using the Kruskal-Wallis nonparametric analysis of variance (ANOVA) test followed by Tukey-Kramer's multiple comparison test. Effects were considered statistically significant at p < 0.05 (two-tailed).
Characterization of cells covering the bone surface
Light and electron microscopic examination of calvariae and long bones showed at least two distinct populations of mononuclear cells covering the bone, each of which was distributed in different areas along the surface. On the apposition side of bones, the surface was lined by cuboidal osteoblasts. A thick layer of osteoid separated the cells from the bone surface. In resorptive areas, apart from osteoclasts, we found numerous cells with long and slender cytoplasmic extensions that covered the bone surface. These cells had a flattened and elongated appearance and were characterized by a moderately developed rough endoplasmic reticulum and Golgi apparatus (Fig. 1A; Table 1). Neighboring cells were in contact with each other by gap junctions (Fig. 1B). In addition, in numerous occasions we found a close spatial relationship of these cells with osteoclasts attached to bone. The plasma membranes of both cell types were separated only by a very narrow gap (Fig. 1C). Occasionally, vesicles were found in the cytoplasm of the flattened cells, preferentially in the area of close apposition with osteoclasts. Assessment of the percentage of osteoclasts attached to bone that expressed a close apposition with one or more of these cells revealed that this was the case for at least 90% of the osteoclasts. Between the flattened cells and the bone surface, a thin layer of nonmineralized fibrillar collagen was present. The density of this collagen fringe was low compared with that of osteoid. The flattened cells were morphologically comparable to bone lining cells, as described by Miller and Jee.(33)
Table Table 1. Characteristics of Osteoblasts and Bone Lining Cells
Both mature osteoid-producing osteoblasts and bone lining cells expressed a high level of alkaline phosphatase activity (not shown). Neither cell type stained with MOMA-2, a selective marker for cells of the monocyte lineage, whereas positive mononuclear cells were present at some distance from the bone surface, in marrow spaces, and periosteum (not shown).
Immunolocalization of type I collagen revealed, in addition to a very strong labeling of bone, intracellular labeling in cuboidal osteoblasts on the appositional side of the bone (Fig. 1D). Labeling was absent in bone lining cells (Fig. 1E). Osteocalcin proved to be present in most areas of the bone tissue. A high level of osteocalcin, however, was observed in bone adjacent to sites where osteoblasts were active in osteoid matrix deposition (Fig. 1F). Bone covered by bone lining cells contained a low level of labeled osteocalcin. Because Tanaka et al.(34) recently showed that ICAM-1 positive osteoblasts play a role in osteoclastogenesis, we immunolocalized this intercellular adhesion molecule in bone sections. Bone lining cells proved to stain positive for this protein (Fig. 1G). No labeling was observed in cuboidal osteoblasts. Control sections incubated with irrelevant species-specific IgGs proved to be negative. The findings are summarized in Table 1.
Enwrapment of collagen protruding from the bone surface
A remarkable feature was that bone lining cells but not cuboidal osteoblasts enwrapped collagen fibrils that protruded from the bone surface. In most instances, cells had enwrapped more than one collagen fibril in the plane of sectioning (Fig. 2). Enwrapment was observed at two different sites along the bone surface: (1) at sites characterized by a sharp demarcation between mineralized bone and nonmineralized collagen fibrils protruding from its surface, indicating that osteoclasts had not been active (Fig. 2A); and (2) in areas left by osteoclasts, in the Howship's lacunae (Figs. 2B and 2C). The latter sites were characterized by a gradual decrease in the amount of mineral toward the bone surface as a result of partial demineralization of the bone by the osteoclasts (Fig. 2C).(35) Slender extensions of bone lining cells were found frequently encircling nonmineralized fibrils in pits that were still partly occupied by an osteoclast. Electron-dense vacuoles containing cross-banded collagen fibrils were found occasionally in the cytoplasm of bone lining cells. However, the number of such collagen-containing vacuoles was much less than that in adjacent periosteal fibroblasts.
The morphological phenomena described here were found in calvariae and long bones, both in noncultured and cultured tissue samples.
Morphometric analysis of collagen enwrapment
Collagen enwrapment at sites previously not occupied by osteoclasts
Morphometric analysis of cultured calvarial bone explants revealed that in the plane of sectioning, 24 ± 7% of the bone lining cells had enwrapped nonmineralized collagen protruding from the bone surface at sites where the osteoclast had not been active. This number increased significantly when explants were cultured in the presence of PTH (38 ± 3%, mean ± SD, n = 12 bone explants, p < 0.01).
Collagen enwrapment in Howship's lacunae
Bone lining cells covered virtually all resorption pits left by osteoclasts and many of them enwrapped nondigested bone collagen. Inhibition of bone resorption by calcitonin resulted in withdrawal of osteoclasts from the bone surface but had no effect on collagen enwrapment by bone lining cells.
To investigate whether collagen left over by the osteoclast was actually digested by bone lining cells, calvarial explants were cultured for 24 h in the presence of the cysteine proteinase inhibitor E-64. This resulted in wide fringes of demineralized bone collagen adjacent to osteoclasts because of their inability to digest bone matrix.(29) Bone lining cells enwrapped bone collagen fibrils protruding from such pits that were left by the osteoclasts (Fig. 3A). During a subsequent culture period in the presence of calcitonin and without E-64, osteoclasts retracted from the Howship's lacunae. Their place was occupied by bone lining cells, which started to resorb the exposed bone collagen fringes. Morphometric analysis revealed that these cells almost completely removed the nondigested bone matrix within 24 h (Fig. 3B).
Enzymes involved in resorption of nonmineralized bone collagen
To determine which proteolytic enzymes are involved in resorption of bone collagen by bone lining cells, calvarial explants were incubated in the presence or absence of selective inhibitors of MMPs, cysteine proteinases, serine proteinases, and aspartic proteinases. The explants were analyzed for the enwrapment activity of bone lining cells and the removal of areas of nondigested demineralized bone matrix by these cells.
MMPs and cysteine proteinases
Enwrapment of collagen protruding from the bottom of Howship's lacunae was affected only by the MMP-inhibitor. The number of cells involved in this activity increased significantly (Table 2). Inhibition of cysteine proteinases had no effect on the volume density of intracellular vacuoles containing collagen fibrils. The protease inhibitors had a similar effect on bone lining cells that were involved in collagen enwrapment outside Howship's lacunae.
Table Table 2. Bone Lining Cells Demonstrating Collagen Enwrapment
To analyze removal of nondigested demineralized areas, explants were cultured in the presence of proteinase inhibitors for 24 h followed by a subsequent period of culture with the same inhibitors, but in the presence of calcitonin. After the initial 24-h culture period not only the cysteine proteinase inhibitor but also inhibition of MMPs had induced large areas of nondigested demineralized bone matrix (Figs. 3 and 4).(29) During subsequent culture in the presence of calcitonin, an almost complete digestion of demineralized bone matrix occurred, despite the presence of the cysteine proteinase inhibitor. Inhibition of the activity of MMPs, however, completely prevented digestion (Fig. 4).
The effect of the MMP-inhibitor proved to be reversible because 48 h after its withdrawal the demineralized matrix was digested. The amount of demineralized area per osteoclast after 24 h in the presence of the MMP-inhibitor was 198 ± 40 μm2 and, after a culture period of 48 h in the absence of the inhibitor, it was reduced to 28 ± 7 μm2 (p < 0.01; n = 6 calvarial explants for either condition).
We subsequently tested whether similar effects occurred when explants were first cultured with the cysteine proteinase inhibitor (E-64) and then cultured for a second period with the MMP-inhibitor. Morphometric analysis of these explants showed that also under these conditions digestion of demineralized areas was prevented by inhibiting MMP activity (16-h culture in the presence of the cysteine proteinase inhibitor: 119 ± 22 μm2 demineralized area; followed by a 12-h culture period in the presence of calcitonin, with MMP-inhibitor: 75 ± 10 μm2 [NS] and without MMP-inhibitor: 26 ± 7 μm2 [p < 0.01; n = 6 calvarial explants for either condition]).
Serine and aspartic proteinases
Inhibition of serine proteinases also affected resorption of nondigested matrix. In the presence of the serine proteinase inhibitor aprotinin digestion was inhibited by approximately 50% (Fig. 5). Inhibition of aspartic proteinases by pepstatin had no effect on digestion of bone matrix.
Remineralization of demineralized matrix in resorption pits?
Because the absence of demineralized areas, as shown in the previous paragraphs, could also be explained in terms of remineralization of the bone matrix, we assessed whether deposition of mineral occurred in these sites. To this end calvarial bone explants were cultured for 24 h in the presence of the cysteine proteinase inhibitor E-64. During a subsequent second culture period (24 h) in the presence or absence of E-64 and calcitonin, the mineral-binding compound xylenol orange was added to the culture medium. In addition a number of the E-64 precultured explants were devitalized and then cultured for a second period, as indicated above. Microscopic evaluation of the explants revealed the presence of fluorescent label at appositional sites. No label was found in resorption pits, either in vital or in devitalized explants (not shown), thus indicating the absence of mineral deposition in demineralized areas of Howship's lacunae.
Deposition of new matrix in resorption pits
To analyze whether deposition of new matrix occurred in Howship's lacunae, explants were cultured for 24 h in the presence of radiolabeled proline. Autoradiographs of these explants showed high levels of label at the convex side, in osteoid layers, and also in resorption pits (Fig. 6A).
Because all resorption pits that were lined by tritium-labeled material appeared to be devoid of nondigested bone matrix, we investigated whether cleaning of the pits was a prerequisite for deposition of tritium-labeled material. Explants were cultured in the absence or presence of the cysteine proteinase inhibitor E-64 and subsequently cultured for a second 24-h time interval without this inhibitor. Because of the presence of the inhibitor, pits were left with a substantial amount of nondigested demineralized bone matrix. During the second 24-h culture period pits were cleaned, thus giving us the opportunity to analyze depositional activity in both cleaned and noncleaned pits (Fig. 6B).
Microscopic examination and quantitative analysis of the autoradiographs showed that high levels of label were present only in cleaned resorption pits. None of the resorption pits that contained nondigested bone matrix proved to be labeled (Figs. 6A and 6B). When tritium-labeled material was present in pits, they were covered by bone lining cells and not by osteoclasts.
To establish the nature of newly deposited matrix in the resorption pits in more detail, ultrathin sections were made of some of the labeled pits in the autoradiographs. Ultrastructural examination of these sections revealed that labeled material consisted of thin collagen fibrils associated with the bone surface (Fig. 6C). Both the density and the thickness of collagen fibrils were much less than that in osteoid underneath cuboidal osteoblasts.
Characteristically for these pits was a sharp demarcation between the fibrils and the mineralized bone. This demarcation was formed by an electron-dense layer, the cement line. As shown by light microscopic immunolocalization, the cement line proved to contain a higher level of osteopontin than that of the adjacent bone (Fig. 7).
Resorption of nondigested bone collagen in bone of pycnodysostosis patients and cathepsin K-deficient mice
Patients suffering from the rare osteopetrosis-like bone disease pycnodysostosis(25) and mice deficient in cathepsin K(26, 36) show a strongly reduced bone digestion by osteoclasts, resulting from a lack of activity of the cysteine proteinase cathepsin K.(26, 36, 37) In both modalities, lack of cathepsin K activity results in large areas of nondigested bone matrix adjacent to osteoclasts. Yet, bones of pycnodysostosis patients and cathepsin K-deficient mice show relatively mild osteopetrotic features, suggesting removal of bone matrix.
In bone samples of pycnodysostosis patients bone lining cells were frequently found at sites where osteoclasts were retracting (Fig. 8). Areas of nondigested bone matrix left by the osteoclasts were occupied by bone lining cells expressing enwrapment activity.
Metacarpal bones of mice lacking cathepsin K were analyzed for the presence of nondigested bone matrix at two sites: in resorption pits covered by osteoclasts (Fig. 9A) and in Howship's lacunae left by osteoclasts. Electron microscopic examination revealed the presence of bone lining cells in close proximity to the osteoclasts and in resorption pits left by osteoclasts. Moreover, in these bones the cells enwrapped nondigested bone matrix. Morphometric analysis showed that almost all nondigested demineralized bone matrix was removed from Howship's lacunae vacated by osteoclasts (Fig. 9B).
The data presented here indicate for the first time that bone lining cells exert a series of activities crucial for remodeling of bone: after withdrawal of the osteoclast from the resorption pit, bone lining cells enter the lacuna and clean its bottom from bone matrix leftovers. This cleaning proves to be a prerequisite for the subsequent deposition of a first layer of (collagenous) proteins in the resorption pits.
Several studies have shown that mononuclear cells are involved in phagocytosis of collagen at the bone surface and particularly at sites adjacent to osteoclasts.(17, 23, 38-40) Some authors suggested that these cells removed demineralized collagen that was not digested by osteoclasts. Our study is the first to show that this is indeed the case. On the basis of the morphological features of the cells involved in this process, we propose that the cells responsible for this activity are bone lining cells, a cell type that is not yet functionally well defined. Our findings strongly suggest that these cells, although not being osteoblasts in the sense that they produce an osteoid layer, belong to the same lineage as osteoblasts for the following reasons: they are alkaline phosphatase positive, respond to PTH, and are associated with the bone surface. Yet, important morphological and functional differences between these cell types were noted in the present study (summarized in Table 1). Cuboidal osteoblasts are characterized by a pronounced synthetic activity and deposit high amounts of type I collagen (osteoid matrix) and noncollagenous proteins, including osteocalcin. One of the characteristics of the bone lining cell, however, is its ablility to enwrap and resorb collagen protruding from the bone surface. Because heterogeneity of the osteoblastic phenotype has been well established(34, 41, 42) we propose that the bone lining cells constitute a subpopulation of the osteoblast family. In this respect it is of interest that Tanaka et al.(34) showed that ICAM-1 positive osteoblastic cells modulate osteoclastogenesis, whereas ICAM-1 negative osteoblastic cells had no effect on this parameter. In the present study we have shown that it is the bone lining cells that are positive for ICAM-1, whereas cuboidal osteoblasts were not.
Matrix metalloproteinases are required for resorption of nonmineralized collagen
Resorption of nonmineralized collagen protruding from the bone surface was observed at two sites: (1) along surfaces that were not lined by osteoblasts and were without indications of prior osteoclastic activity, and (2) in Howship's lacunae left by osteoclasts. Because in both situations inhibition of MMPs resulted in increased amounts of enwrapped collagen in bone lining cells, we propose that digestion of this collagenous material requires the activity of MMPs. Direct proof for this assumption was obtained for the sites where osteoclasts had exerted their activity, within the resorption pits. Inhibition of the activity of MMPs completely prevented degradation of demineralized bone collagen that was left by osteoclasts.
Digestion of collagen occurred at sites where the plasma membrane of bone lining cells enwrapped collagen, probably by membrane-bound MMPs. Although the type(s) of MMP involved in this process have yet to be determined, possible candidates are gelatinase A (MMP-2) and/or one of the membrane type MMPs (MT-MMPs), given that these enzymes are membrane-bound and have the capacity to digest collagen.(43, 44) Alternatively, an MMP associated with collagen protruding from the bone surface may be involved. Immunolocalization studies have shown the association of MMP-13 to collagen at such sites.(45) One or more of the MMPs involved in this process may be activated by the serine proteinase plasmin.(46) The partial inhibition of resorption by the serine proteinase inhibitor aprotinin supports this view.
Resorption of collagen fibrils as described here differs from digestion of fibrillar collagen of soft connective tissue by fibroblasts. The latter cell type can partially digest collagen in segregated areas formed by cytoplasmic extensions and this digestion depends on MMPs, although the bulk of collagen is digested intracellularly in the lysosomal apparatus by cysteine proteinases.(47) This lysosomal degradation is not an essential step in the resorption by bone lining cells. Inhibition of cysteine proteinase activity, which greatly enhances the amount of nondigested phagocytosed collagen in fibroblasts,(48–50) does not have such an effect on bone lining cells and does not affect resorption of nonmineralized bone collagen by these cells. A feasible explanation for digestion at the plasma membrane and not intracellularly is that the majority of fibrils are still embedded in bone, thus making uptake physically difficult, if not impossible. A pathway of collagen digestion with similarities to the one described in the present study has been noted in fibroblast-like cells.(23, 39) These authors showed the presence of these cells adjacent to osteoclasts with high amounts of engulfed/enwrapped collagen fibrils. In addition, here electron-dense collagen-containing vacuoles were virtually absent. Partially digested (bone) collagen freed by osteoclasts was assumed to be resorbed by these cells.(22)
The finding that cysteine proteinases are not essential for digestion of collagen protruding from the bone surface coincides with data presented by Jilka and Hamilton(51, 52) and Delaissé and Vaes.(28) These authors showed that nonosteoclastic resorption of bone-associated collagen was not inhibited by cysteine proteinase inhibitors. They concluded that another cell type was involved in the digestion of this collagen. PTH, but not IL-1 or LPS, stimulated the collagen-resorbing activity of these cells.(52) The findings of these authors can be explained by the presently described activity of bone lining cells.
It is interesting to note that collagen enwrapment by bone lining cells occurred only along the cell surface directed toward the bone. No enwrapment was noted along the opposite cell surface, the surface in contact with periosteal collagen. Thus it seems that the cells somehow recognize and select collagen protruding from bone rather than collagen associated with the periosteal soft connective tissue. It remains to be elucidated which mechanisms are involved in this polarized activity.
Bone lining cells digest bone matrix that is not resorbed by osteoclasts in pycnodysostosis and cathepsin K-deficient mice
The present data are the first to provide a plausible explanation for the relatively mild osteopetrotic features of patients suffering from pycnodysostosis and mice lacking cathepsin K activity. Both conditions are characterized by osteoclasts that are incapable of efficiently digesting bone matrix. The cells demineralize bone but, because of the lack of active cathepsin K, large areas of nondigested bone matrix are formed by the osteoclasts (pycnodysostosis(25); cathepsin K knockout mice(26, 36)). When removal of this nondigested matrix does not occur, formation of new bone would have been strongly hampered. We now show that bone lining cells enter the demineralized nondigested areas as soon as the osteoclasts leave the Howship's lacunae and resorb the demineralized matrix. In fact, in cathepsin K-deficient mice the areas of nondigested demineralized bone not covered by osteoclasts were small (less than 12% of the total area of demineralized bone matrix), thus suggesting that this cleaning process occurs rather rapidly. We propose that the bone lining cells by their cleaning properties rescue the process of bone remodeling.
It is only in cleaned resorption pits that bone lining cells form a cement line and deposit a thin layer of collagen
Our autoradiographic data show that bone lining cells deposit a layer of proline-rich protein at the bottom of cleaned resorption pits. Because this activity was observed only in pits completely devoid of remnants left by the osteoclast, the data indicate that cleaning is a prerequisite for deposition to occur. The electron-dense layer covering the bottom of the resorption pits contained osteopontin and represents the so-called lamina limitans or cement line, which demarcates sites where formation of new bone is initiated.(30) Thin collagen fibrils were found to be inserted at this site. Such an initial formation in resorption pits of relatively thin collagen fibrils was previously shown by Shen and coworkers.(53) These authors analyzed the sequential deposition of extracellular matrix components by osteoblasts at different stages of development and found that the first layer consisted of thin collagen type III fibrils. At a later stage thicker type I collagen fibrils were deposited.
Do bone lining cells orchestrate bone remodeling?
A finding that may have implications for our understanding how bone lining cells interact with osteoclasts is that virtually all osteoclasts attached to bone were in close contact with these cells. We noted at these sites small vesicles in cytoplasmic extensions of bone lining cells. This observation may suggest transduction of signals from the bone lining cell to the osteoclast or vice versa. It is generally taken that cells from the osteoblast lineage are essential for the modulation of osteoclastogenesis and activation of these cells.(2, 54) Our present findings may indicate that not mature osteoblasts but bone lining cells are important for modulating osteoclast activity. In line with this assumption is our finding that bone lining cells express ICAM-1. Osteoblast-like cells positive for this protein were shown to modulate osteoclastogenesis.(34)
We found at sites previously not occupied by osteoclasts, that bone lining cells enwrap collagen protruding from the surface, and we propose that these cells predispose (clean) the surface for osteoclastic resorption. Because the amount of enwrapped collagen increased after inhibition of MMP activity, we assume that this cleaning activity depends on MMPs. This finding coincides with the view of Chambers and coworkers,(6) who suggested that osteoblasts clean the bone surface before osteoclastic attack, an activity depending on MMPs. Our data strongly suggest that not the mature osteoblast but the bone lining cell is essential for this activity. In this respect it is of interest to note that the amount of enwrapped collagen increased by PTH, a hormone known to activate bone resorption and to enhance the release of MMPs by cells of the osteoblast lineage.(55–57)
Proposed sequence of coupling bone resorption and bone formation
The present findings suggest the following sequence of activities related to the coordinated link between bone resorption and bone deposition (Fig. 10):
(1) Before osteoclastic attachment and resorption, bone lining cells digest nonmineralized collagen protruding from the bone surface. This digestion occurs in segregated areas at the plasma membrane of bone lining cells and depends on activity of MMPs.
(2) Osteoclasts attach to these sites and resorb bone. This activity depends on cysteine proteinases (long bone and calvaria) and MMPs (calvaria).(58) The resorption by osteoclasts is not complete; after withdrawal they leave behind remnants of demineralized nondigested bone collagen.
(3) Bone lining cells enter resorption lacunae and digest the collagen left by osteoclasts. Again, this activity depends primarily on MMPs.
(4) Bone lining cells form a cement line and deposit a thin layer of fibrillar collagen on the cleaned surfaces.
Finally, osteoid deposition and actual bone formation by osteoblasts is taken to occur at these sites.(11)
The authors are grateful for Dr. C. J. F. van Noorden's invaluable comments, J. A. Niehof's excellent technical assistance, and C. E. Gravemeijer's careful handling of the photographic material. This study was supported by a grant of the Netherlands Institute of Dental Sciences (IOT).