Bone formation throughout skeletal growth and remodeling always entails deposition of new bone onto a pre-existing mineralized surface. In contrast, the initial deposition of bone in development requires the formation, ex novo, of the first mineralized structure in a nonmineralized tissue. We investigated the cellular events associated with this initial bone formation, with specific reference to the respective role of cartilage and bone cells in bones which form via a cartilage model. The cellular architecture of initial osteogenic sites was investigated by light, confocal, and electron microscopy (EM) in the membranous ossification of fetal calvarial bones (not forming via a cartilage model) and in the membranous ossification of the bony collars of endochondral bones. Bone sialoprotein (BSP), which is expressed during early phases of bone deposition and has been proposed to be involved in the control of both mineral formation and bone cell–matrix interactions, was used as a marker of initial bone formation. We found that at all sites, BSP-producing cells (as identified by intracellular immunoreactivity) are arranged in a characteristic vis-à-vis (face to face) pattern prior to the appearance of the first mineralizing BSP-immunoreactive extracellular matrix. In perichondral osteogenesis, the vis-à-vis pattern comprises osteoblasts differentiating from the perichondrium/periosteum and early hypertrophic chondrocytes located at the lateral aspects of the rudiment. By EM, the first mineral and the first BSP-immunoreactive sites coincide temporally and spatially in the extracellular matrix at the boundary between cartilage and periosteum. We further showed that in an in vitro avian model of chondrocyte differentiation in vitro to osteoblast-like cells, early hypertrophic chondrocytes replated as adherent cells turned on the expression of high levels of BSP in conjunction with the switch to collagen type I synthesis and matrix mineralization. We propose a model for the priming of bone deposition, i.e., the formation of the first bone structure, in which the architectural layout of cells competent to deposit a mineralizing matrix (the vis-à-vis pattern) determines the polarized deposition of bone. For bones forming via a cartilage model, the priming of bone deposition involves and requires cells that differentiate from early hypertrophic chondrocytes.
Development and growth of bones formed via a cartilage model involves two distinct patterns of bone formation. Endochondral bone formation is the process whereby cartilage matrix in the interior of growing metaphyses calcifies, undergoes partial resorption, and provides mineralized cores onto which newly recruited osteoblasts deposit new bone. Perichondral bone formation, in contrast, is the process whereby bone is formed outside, and around, cartilage, and leads to the formation of a primitive cortex and of its associated periosteum. In embryonic cartilage rudiments, perichondral ossification provides the bony collars that form around the midshaft of the cartilage rudiments prior to the onset of vascular invasion and endochondral ossification.(1–5) During prenatal and postnatal bone growth, perichondral bone formation continues at the leading edge of the growing bone cortex, around the upper zone of cartilage hypertrophy. Because it does not depend on prior mineralization and resorption of cartilage (but precedes both), perichondral ossification can be regarded, like periosteal bone formation which it anticipates and heralds, as a membranous type of bone formation.(1,2,4)
Despite the dual processes by which bone formation occurs during the development of long bones, the role of cartilage cells in bone development is generally equated to their role in endochondral bone formation. Here, cell proliferation and matrix synthesis provide the basis for longitudinal growth, before apoptosis (occurring precisely in the “degeneration zone” of classical histology)(6) of hypertrophic chondrocytes sets the stage for cartilage calcification and subsequent resorption.
To learn more about the potential role of cartilage cells in the other major osteogenic process (perichondral bone formation) associated with development and growth of long bones, we analyzed the details of the formation of the bony collar during development by light, confocal, and electron microscopy (EM), and compared them with early events occurring at sites of cartilage-free membranous ossification. Bone sialoprotein (BSP), a cell adhesion-promoting noncollagenous bone matrix protein,(7–9) known to be specifically associated with de novo bone formation(10,11) and putatively involved in the control of bone mineralization,(9,12,13) was used as a tissue marker of early events in bone formation. We provide evidence here that initial perichondral bone formation involves the participation of early hypertrophic chondrocytes and osteoblasts in the deposition of a BSP-enriched mineralized matrix. This duplicates, in the specific setting of perichondral bone formation, a previously unrecognized pattern of deposition of bone between facing rows of cells, which recurs whenever bone is to be formed in the absence of a pre-existing mineralized substrate (available, in contrast, in endochondral bone formation and in bone remodeling).
MATERIALS AND METHODS
Histology and immunohistology
Pregnant albino Wistar rats were sacrificed by ether inhalation under institutionally approved animal protocols. Fetuses were collected at 15, 16, 17, and 18 days of gestation, cut sagittally, and fixed in 4% formaldehyde (freshly made from paraformaldehyde) in 0.1 M phosphate buffer, pH 7.2, for 2 h at 4°C. From some fetuses, ribs were dissected and embedded separately to obtain both longitudinal and transverse sections. Additional samples were prepared for low-temperature glycol methacrylate embedding and alkaline phosphatase (ALP) cytochemistry on undecalcified plastic sections as described.(10) A polyclonal antiserum to BSP generated against human BSP purified from bone matrix (LF6), and cross-reacting with the rat protein,(14,15) was a gift from Dr. L.W. Fisher, National Institute of Dental Research, National Institutes of Health, Bethesda, MD. Light microscopy immunolocalization by an immunoperoxidase protocol was performed as described previously.(10)
Tissue sections of variable thickness (4–20 μm), immunostained for BSP, were studied by confocal laser scanning microscopy in the reflectance mode, that allows for confocal imaging of the diaminobenzidine reaction product.(16) A Phoibos Sarastro system equipped with an argon ion (l = 488 nm) laser was used. The laser beam was attenuated (30%) with a neutral density (ND) filter. Stacks of confocal images were displayed and photographed as uncorrected, pseudo–color-coded digital images.
EM localization of BSP
Low-temperature Lowicryl K4M (Sigma, St. Louis, MO, U.S.A.) embedding and EM immunolocalization of BSP were done using the LF6 antiserum and an immunogold protocol, as described.(13) Briefly, aldehyde-fixed samples from ribs of day 15 rat fetuses were processed for low-temperature (−35°C) Lowicryl K4M embedding. Polymerization was induced by UV irradiation at −35°C overnight, followed by 2 days of UV irradiation at room temperature. Semithin sections were stained with methylene Blue or Giemsa for morphology and for selecting relevant fields (first subperichondral bone formation) for EM. Additional semithin sections were stained with the Von Kòssa method for demonstration of mineralization at the light microscopy level. After exposure to undiluted normal goat serum, thin sections were incubated overnight with a 1:10 dilution (in phosphate-buffered saline [PBS], 0.1% bovine serum albumin) of primary antibody, and, after appropriate washing, with gold-labeled (particle size 10 nm) goat anti-rabbit immunoglobulin G (BioCell, Cardiff, U.K.) diluted 1:50 in PBS, 0.1% bovine serum albumin, for 30 minutes at room temperature. Controls included omission of the BSP antiserum, its replacement with normal rabbit serum (NRS), and the use of similar dilutions of rabbit antisera of unrelated specificity.
Cell cultures were performed according to published procedures.(17–19) Briefly, chondrocytes were enzymatically dissociated from 6-day-old chick embryo tibiae and cultured in adhesion for 3 weeks in Coon's modified culture medium containing 10% fetal calf serum. After their expansion as dedifferentiated cells expressing type I collagen, the cells were transferred into suspension cultures on 1% agarose coated dishes for additional 3 weeks. In this condition, cells redifferentiated to proliferating and hypertrophic chondrocytes expressing, at first, type II collagen and then type X collagen. Eventually, a homogeneous population of single isolated hypertrophic chondrocytes was obtained. Cell differentiation was assessed by metabolic labeling and by peroxidase staining with antibodies directed against type X collagen. Hypertrophic chondrocytes were filtered through a nylon filter Nytex (42 μm mesh), digested with hyaluronidase (1 mg/ml), and replated as adherent cells (1.8 × 106 in 100-mm dish) in the same culture medium. After 3 days, the medium was supplemented with 100 μg/ml of ascorbate, 10 mM β-glycerophosphate, and 500 nM retinoic acid. Culture was continued for 2–3 weeks. The medium was changed every day without cell passaging.(18,19)
Cell labeling and protein analysis
Cells were labeled with [35S]methionine as described.(20) Briefly, a confluent culture in a 100-mm dish was starved in medium lacking methionine for 2 h and labeled for 2 additional h with fresh medium containing 100 μCi/ml [35S]methionine. Culture medium was collected and cells were washed with PBS and scraped in 0.01% SDS in PBS. Samples of culture media and cell lysates were run for protein analysis on SDS-PAGE under reducing conditions. Determination of the differentiation stage of cultured cells was based on the profile of labeled protein synthesized.
Western blot analysis
Production of BSP by cultures of hypertrophic chondrocytes replated as adherent cells (as described in the cell culture section) was studied by Western blot analysis. Cell lysates were prepared by adding 0.1% SDS in PBS to the cells. Cell layer and culture media, containing 100 μg of protein, were loaded onto a 10% SDS-PAGE. Electrophoresis was performed under reducing conditions, the gel was blotted to a BA85 nitrocellulose membrane (Schleicher & Schuell GmbH, Dassel, Germany) according to the procedure described by Towbin et al.(17) The blot was saturated for 2 h with 5% nonfat cow milk in TTBS buffer (20 mM Tris HCl, pH 7.5, 500 mM NaCl, 0.05% Tween 20), incubated with a polyclonal antiserum recognizing chick BSP (a gift of Dr. L.W. Fisher, NIDR, NIH: LF119) (1:2000) for 16 h at room temperature. After additional washes, the detection was performed by an ALP-conjugated anti-rabbit immunoglobulin G (H + l; Boehringer-Mannheim, Mannheim, Germany) using Western Blue stabilized substrate for ALP (Promega Co., Madison, WI, U.S.A.).
The vis-à-vis pattern of early bone formation
The early cellular events in the formation of the bony collar in bones ossifying via cartilage models were studied in the ribs of rat fetuses and compared with the early events in the membranous ossification of calvaria.
The initiation and subsequent progression of the formation of a bony collar in developing ribs were analyzed in sections of 15-day and 16-day/17-day fetuses, respectively. Morphology and ALP cytochemistry were used for proper identification of osteogenic cells and hypertrophic chondrocytes (Figs. 1 and 2), whereas BSP was used as an established marker of actual early deposition of bone. Compared with ALP activity, a known marker of both osteogenic cells and hypertrophic chondrocytes, BSP immunoreactivity was restricted to spatially defined subsets of either cell population. In sections of ribs from 15-day fetuses, intracellular BSP immunoreactivity was found to appear simultaneously (i.e., at the same site along the rudiment axis) both in some osteoblasts within the osteogenic perichondrium and in some chondrocytes located at the edge of the cartilage rudiment. BSP-positive osteoblasts and BSP-positive chondrocytes were arranged in facing rows of cells (vis-à-vis). The earliest, BSP-enriched matrix of the forming bony collar was deposited between these rows of BSP-producing osteoblasts and chondrocytes and appeared as a thin, continuous film of reaction product. In ribs of 16-day and 17-day fetuses, the primitive bony collar appeared as a continuous shell of bone around the hypertrophic cartilage. The primitive haversian canals were instead just beginning to form and appeared as incomplete arcades of new bone located at the external aspect of the primitive bony collar (Fig. 2c). In sections immunostained for BSP, only the osteoblasts associated with the forming portions of the haversian arcades (de novo, initial bone formation) displayed intracellular immunoreactivity for BSP. Osteoblasts associated with already formed bone were devoid of BSP immunoreactivity, as were also the underlying hypertrophic chondrocytes. Thus, the expression of BSP, which appeared to be turned on simultaneously in vis-à-vis chondrocytes and osteoblasts at the stage of initial formation of the bony collar, also appeared to be simultaneously turned off in the same cell types at a later stage.
In developing calvaria of 15-day fetuses, studied for comparison, the earliest bone matrix, detected as thin ribbons of BSP-enriched matrix, was localized between facing rows of BSP-immunoreactive osteoblasts (vis-à-vis). As a result, the first membranous bony trabeculae, like the developing bony collars of cartilaginous rudiments, were observed to be rimmed by vis-à-vis rows of BSP-immunoreactive cells (Fig. 3c).
To investigate further the architecture of the earliest osteogenic site, we recorded stacks of serial optical sections from the same specimens used for immunohistochemical analysis by light microscopy. Figure 4 shows one such stack, corresponding to the section and field shown in Fig. 3a, and depicting the (distinct) sites of appearance of the first BSP-immunoreactive cells and extracellular matrix in the forming bony collar. At the site where cells marked by intracellular BSP immunostaining first appear, but the extracellular matrix is devoid of BSP, serial optical sectioning showed that the first chondrocytes and osteoblasts immunoreactive for BSP are misaligned in the z axis, as well as in the xy planes. The leading edge of the first BSP-immunoreactive subperiosteal matrix appeared to coincide with the site of appearance of immunoreactive chondrocytes and osteoblasts aligned in the same focal plane, indicating that BSP is only deposited between two cells that are producing it. Close association of cells with the extracellular reaction product was seen on both the osteoblastic and chondrocytic sides. Minor amounts of extracellular reaction product, not detected in transmitted light and haphazardly distributed, were demostrated by confocal microscopy within the matrix of hypertrophic cartilage deep to the cartilage/bone junction.
Deposition of BSP and of the first mineral between vis-à-vis cells
We next compared the pattern of mineral and BSP deposition within the first subperiosteal bone. Figure 5 shows mineral distribution as seen at the light level using the von Kòssa stain. As seen in semithin Lowicryl K4M sections, mineral deposition appears as a thin ribbon of staining separating the vis-à-vis rows of osteoblasts and chondrocytes and is virtually identical to the pattern of extracellular BSP deposition. When the same fields were examined in thin sections at the EM level, early, discrete mineral clusters were easily identified at the boundary between a typical cartilage matrix (thin, type II collagen fibers and proteoglycan “granules”) and a collagenous matrix made of periodically banded, thick type I collagen fibers (Fig. 6). In thin sections used for BSP immunolabeling, the sites of mineral deposition (which appear decalcified due to the prolonged exposure to aqueous solution during the immunolabeling procedure)(13,18) were coincident with the sites of BSP localization (Fig. 6). Identical structures, imaged both in confocal reflectance microscopy (as discrete granules) and in EM of thin Lowicryl sections, were identified as the sites of BSP immunoreactivity in the bone matrix at sites of membranous bone formation (Fig. 7).
Adherent hypertrophic chondrocytes produce BSP in vitro during their further differentiation to osteoblast-like cells
As immunolocalization data suggested that hyperthrophic chondrocytes (or a subset thereof) participate in the deposition of the earliest, BSP-enriched layer of bone in vivo, we decided to look at BSP production in a defined in vitro system in which chick hypertrophic chondrocytes further differentiate into osteoblast-like cells(19,20) upon culture as adherent cells. Culture medium and cell layer were harvested from cultures of hypertrophic chondrocytes grown as adherent cells in the presence of retinoic acid, and production of BSP was assessed by Western blot analysis. Figure 8A shows that BSP is detectable in the culture medium starting at day 7, which is immediately before the onset of mineralization (days 9–10). Interestingly, the appearance of BSP in the culture medium is temporally coincident with the induction and release of two proteins (arrows in Fig. 9) recognized by antibodies raised against type I collagen chains (data not shown). When equivalent fractions of cell layers were analyzed, BSP was detectable as early as day 4 (Fig. 8B). The production and secretion of BSP by mineralizing osteoblast-like cells derived from hypertrophic chondrocytes cultured in the absence of retinoic acid was also observed in temporal association with the onset of mineralization (data not shown).
Throughout bone growth and remodeling, deposition of bone regularly occurs by addition of new matrix onto a pre-existing mineralized substrate consisting of either mineralized cartilage (in the metaphyseal growth plates) or pre-existing mineralized bone. Likely because a mineralized surface is capable of binding and immobilizing osteoblast-produced matrix protein, deposition of bone under these circumstances is a highly polarized histologic event, and new bone is deposited in an orderly fashion in direct continuity with the pre-existing mineralized structure. The initial formation of bone in development (membranous bone formation, perichondral bone formation), in contrast, involves the deposition of new bone before a preformed mineralized substrate is available, which sets a noticeable difference in the local mechanisms of matrix deposition and accumulation. Using BSP as a marker of early bone formation, we have identified the deposition of a mineralizing matrix between cells organized in opposing rows (which we called the vis-à-vis pattern) as a recurring histologic theme of de novo bone formation both in calvarial and in perichondral ossification. Consistent with a postulated role of BSP in the control of matrix mineralization,(12,13) early mineral aggregates and BSP immunoreactive sites codistribute within this early matrix. Therefore, the early mineral itself is formed between vis-à-vis rows of (BSP-producing) cells. We propose that this multicellular organization represents a critical step for the priming of osteogenesis. Assuming that secretion of matrix proteins is an instance of nonpolarized secretion (as is secretion of most proteins by most connective tissue cells), the vis-à-vis arrangement of cells secreting bone matrix proteins would generate, by default, a local high concentration of secreted proteins (which are enriched in BSP and other cell-binding proteins in the early phases of matrix production)(10,11) where cells come to face each other. Deposition of mineral would then enforce the direction of subsequent deposition of bone by providing adsorption sites for secreted proteins, thus generating a mechanism of polarized deposition of matrix produced by (initially) nonpolarized secretory cells. The local matrix stoichiometry may also direct the subsequent polarization of osteogenic cells, provided that one or more adsorbed proteins display, as BSP and other bone proteins do, cell-binding properties. Sorting of integrins along the plasma membrane as dictated by the physical distribution of matrix ligands has been invoked as one mechanism for establishing cell polarity(21) and might be envisioned, albeit as a testable hypothesis, as operating in bone cells as well (Fig. 10).
For the vis-à-vis pattern of bone deposition to occur in perichondral ossification, and for the earliest mineralized matrix to be deposited directly onto the cartilaginous surface of the rudiment (underneath the developing periosteum), one of the two rows of matrix-producing cells must be provided by chondrocytes. We have shown here that intracellular BSP appears simultaneously in osteoblasts and in vis-à-vis early hypertrophic chondrocytes prior to the appearance of extracellular BSP and mineral at the rudiment surface. Expression of BSP in (early) hypertrophic chondrocytes located in the central regions of cartilage rudiments (growth plates) has been reported previously. Characteristically, in previous studies(10,22) and in the present one, late hypertrophic chondrocytes associated with matrix mineralization and displaying the typical “degenerative” appearance (and undergoing apoptosis)(6) do not express BSP or its mRNA. The priming of the formation of the bony collars thus occurs in spatial and temporal association with early hypertrophy of chondrocytes.
Chondrocytes expressing BSP and participating in the priming of perichondral bone formation thus correspond to a stage of differentiation that precedes their entering the apoptotic pathway at the stage of cell “degeneration,” i.e., “late” hypertrophy.(6) Spatially, early hypertrophic cells are located not only in the uppermost part of the zone of hypertrophy of the growth plates, but also at the lateral edges of the rudiments (vis-à-vis the developing perichondral bone). Expression of ALP in these lateral, flattened chondrocytes has been shown in the present study, in agreement with previous findings in the chick embryo, in which coexpression of collagen type X in the same subset of cells was also shown.(23)
The very possibility to replate hypertrophic chondrocytes as adherent cells and observe them resuming cell proliferation indicate that the apoptotic fate is not yet irreversibly determined at the time when a chondrocyte enters the stage of differentiation marked by the expression of collagen type X. Cells that can be replated efficiently as adherent cells in our system are necessarily nonapoptotic, and yet they produce collagen type X at the time of replating (early hypertrophic chondrocytes).
We have shown here that replated adherent (early) hypertrophic chondrocytes, which resume cell proliferation and deposit and mineralize a collagen type I–rich matrix,(20) simultaneously express BSP. Taken together, our in vivo and in vitro observations suggest that once a chondrocyte has reached the stage of early hypertrophy, it begins to express phenotypic traits of the osteogenic lineage. In vitro, the complete development of the osteogenic phenotype is dependent on adhesion to a substrate. We believe that this in vitro condition mimicks the fate of early hypertrophic chondrocytes exposed to the specific matrix and humoral environment found at the lateral aspects of cartilage rudiments in vivo, as well as to the potential signals from osteogenic cells in the periosteum/perichondrium. We also speculate that early hypertrophic chondrocytes that have already started to express some traits of the osteogenic lineage but are not exposed to a permissive microenvironment—as in the case of cells centrally located in the cartilage rudiment—become exposed to conflicting signals and enter an apoptotic pathway.
In vivo, Indian hedgehog (IHH), a major regulator of skeletal development, is expressed in early hypertrophic chondrocytes,(24,25) and the binding subunit of its receptor, patched, is expressed in the primitive osteogenic cells associated with the formation of the primitive bony collar.(26) Vis-à-vis early hypertrophic chondrocytes and osteogenic cells are thus ideally positioned for efficient cellular cross-talk involving IHH and other signaling molecules downstream of IHH.
The role exterted by chondrocytes in the control of endochondral bone formation (growth plate physiology) resides mainly in the rates of cell growth, matrix synthesis, and apoptosis along the longitudinal axis of the growth plate. The Indian hedgehog parathyroid hormone–related protein (PTHrP) signaling pathway has emerged as a key regulator of these processes.(24,26–28) PTHrP, which is induced by IHH in the periarticular perichondrium in vivo,(25,28) induces chondrocyte proliferation in vitro(29) and increases the expression of the antiapoptotic protein, Bcl-2.(6) Disruption of this signaling pathway by ablation of the PTHrP gene in mice results in severe abnormalities of longitudinal bone growth.(30,31) However, abnormalities in perichondral ossification are also obvious in the PTHrP-depleted mice.(30,31) Formation of the bony collars is accelerated in these animals prior to rudiment invasion and formation of the growth plates. Of note, formation of the bony collar is strictly an instance of membranous bone formation, but membranous bone formation is not disturbed in the cranial bones in PTHrP-depleted mice, suggesting that chondrocyte–osteogenic cell interactions at the level of the developing bony collar are also altered by disruption of the IHH-PTHrP pathway.
Endochondral ossification and perichondral ossification are quite different events. The former is the replacement of mineralized cartilage with new bone, while the latter consists in de novo osteogenesis. The former requires resorption of mineralized cartilage prior to deposition of bone, while the latter involves deposition of bone prior to resorption. The former involves formation of bone on a mineralized scaffold, while the latter involves the formation of the scaffold itself. Endochondral bone formation ultimately leads to replacement of cartilage with a marrow cavity (a localized loss of mineralized tissue), whereas perichondral/periosteal bone formation leads to the building of a bony shaft (a localized gain of mineralized tissue). From these studies, it is clear that chondrocytes participate in different ways in the two different processes, depending on the local microenvironment in which they are located.
We thank Giuliana Silvestrini for help with electron microscopy. This work was supported in part by grants from the Italian Space Agency (ASI) and the MURST, Italy (P.B.).