Formation of a chondro-osseous rudiment in micromass cultures of human bone-marrow stromal cells

Probing the differentiation potential of human bone-marrow stromal stem cells (BMSCs) employs a variety of experimental approaches. Most commonly,single or multiple differentiation capabilities (e.g. osteo-, chondro- or adipogenic) are ascribed to either clonal or non-clonal stromal cell populations based on the assessment of tissue-specific differentiation markers in culture. Nonetheless, several considerations limit the value and significance of such assays, without detracting from their empirical value or from their desirable amenability to routine use. For example, the expression of an in vitro phenotype reminiscent of osteogenic cells does not necessarily predict the ability of a given cell strain to generate bone upon in vivo transplantation (Kuznetsov et al.,1997; Satomura et al.,2000); likewise, marked phenotypic variability is observed over time in clonal cell strains in culture(Deryugina et al., 1994; Muraglia et al., 2000). The use of in vivo transplantation assays(Friedenstein et al., 1987; Goshima et al., 1991; Gundle et al., 1995; Haynesworth et al., 1992; Krebsbach et al., 1997; Martin et al., 1997; Quarto et al., 1995) has become a valuable standard for assessing the osteogenic potential of marrow stromal cells. Not only do these assays probe the physiological osteogenic function in vivo, they also provide the simplest and most convincing readout(the formation of histologically proved true bone tissue). The micromass culture [`pellet culture' (Johnstone et al., 1998)] of human BMSCs provides an assay of similar significance for chondrogenesis. Culturing BMSCs after inducing an artificial condensation event and exposing them to a chondrogenic cytokine milieu results in the generation of a three-dimensional structure that is directly reminiscent of true hyaline cartilage(Johnstone et al., 1998). Recent work has shown that the system can be advantageously used with human BMSCs (Johnstone et al., 1998; Mackay et al., 1998; Mastrogiacomo et al., 2001; Sekiya et al., 2001; Sekiya et al., 2002; Yoo et al., 1998) as well as with BMSCs from other species (Johnstone et al., 1998), and has emphasized the value of the system for dissecting molecular determinants of chondrogenesis in a relatively simple and defined set of experimental conditions(Sekiya et al., 2002).

Starting from previous in vivo and in vitro observations indicating that chondrogenic differentiation does not preclude the further development of an osteoblast-like phenotype in vitro (Bianco et al., 1998; Galotto et al.,1994; Gentili et al.,1993; Jimenez et al.,2001), we asked whether this could be effectively investigated taking advantage of the histological dimension and the three-dimensional nature of the pellet culture system of marrow-derived skeletal progenitor cells. We report here that, by appropriate manipulation of the system, not only can true bone formation be obtained but also specific spatial and temporal patterns of chondro- and osteogenesis are obtained in vitro that culminate in the generation of an in vitro formed `organoid' that directly mimics the formation of the bony collar around cartilage anlagen that occurs during embryonic bone development.

Recombinant human fibroblast growth factor 2 (rh FGF-2) and recombinant human transforming growth factor β1 (rhTGFβ1) were from Austral Biologicals (San Ramon, CA). Foetal calf serum (FCS) was purchased from Kallergen (Settimo Milanese, Italy). All other chemicals were from Sigma (St Louis, MO).

BMSCs were obtained from iliac crest marrow aspirates of healthy donors(age range 1-60 years). All the procedures were approved by an institutional ethical review committee. After washing in PBS, mononucleated cells were stained with 0.1% methyl violet in 0.1 M citric acid and counted. The cells were then suspended in Coon's modified Ham's F12 medium supplemented with 10%FCS and 1 ng ml^-1 rhFGF-2 and plated at between 2×10⁶ and 5×10⁶ cells per 100 mm dish. Medium was changed 3 days after plating and then twice a week thereafter.

Cells from confluent cultures (20-25 days in culture, passage 0,corresponding to 12-15 doublings) were released by 0.05% trypsin in 0.01%EDTA, counted and used to generate pellet cultures conducive for chondrogenesis in vitro, essentially as previously described(Johnstone et al., 1998). Briefly, 2.5×10⁵ cells were centrifuged at 500 g in 15 ml polypropylene conical tubes and the resulting pellets were cultured for 4-7 weeks. Control cultures were grown in a serum-free chemically defined medium consisting of Coon's modified Ham's F12 medium supplemented with 10^-6 M bovine insulin,8×10^-8 M human apo-transferrin, 8×10^-8 M bovine serum albumin, 4×10^-6 M linoleic acid, 10^-3M sodium pyruvate (control medium). To induce chondrogenic differentiation,the control medium was supplemented with 10 ng ml^-1 rhTGFβ1,10^-7 M dexamethasone and 2.5×10^-4 M ascorbic acid.

Cultures were incubated for 4 weeks at 37°C in an atmosphere containing 5% CO₂; the medium was changed every 4-5 days and ascorbic acid added three times a week. To monitor chondrogenesis, cultures were harvested at 1-4 weeks and processed for histology (see below). At the end of the 4-week culture, some cultures were incubated for a further 2-3 weeks in a medium conducive for in vitro mineralization (control medium containing 7.0×10^-3 M β-glycerophosphate, 10^-8 M dexamethasone and 2.5×10^-4 M ascorbic acid), then fixed and processed for histology.

The cell aggregates were fixed with 4% formaldehyde in PBS for 10-15 minutes and routinely embedded in paraffin. Paraffin sections were stained with haematoxylin-eosin, toluidine blue, alcian blue and alizarin red S, and viewed in transmitted and polarized light microscopy.

Monoclonal antibodies against type I and type II collagen (SP1D8 and CIICI,respectively) were obtained from the Developmental Studies Hybridoma Bank(Department of Biological Sciences, University of Iowa). Supernatants from hybridoma cultures were used undiluted (SP1D8) or concentrated ten times(CIICI). A monoclonal antibody to human recombinant type X collagen (X53) was kindly provided by K. von der Mark (Institute of Experimental Medicine,Friedrich Alexander University of Erlangen, Germany); supernatant from hybridoma culture was used undiluted. A rabbit antiserum raised against human bone sialoprotein [BSP, LF6 - (Fisher et al., 1995)] was kindly provided by L. W. Fisher (NIDCR, NIH,Bethesda, MD) and was used at a dilution of 1:100 in PBS, 0.1% bovine serum albumin (BSA). A rabbit antiserum raised against bovine osteocalcin(cross-reactive with the human protein) was kindly provided by S. Robins(Rowett Research Institute, Aberdeen, UK) and was used at a dilution of 1:500 in PBS, 10% goat serum.

Deparaffinized and rehydrated 5 μm sections were incubated with 3%hydrogen peroxide in methanol for 30 minutes to inhibit endogenous peroxidase activity. Some sections were subjected to digestion with 1 mg ml^-1hyaluronidase in PBS, pH 6.0 for 15 minutes at 37°C prior to use. Sections were exposed to normal goat or pig serum (Dako, Glostrup, Denmark) diluted 1:10 in PBS, 0.1% BSA for 30 minutes before incubation with the primary antibodies. Slides were then washed with PBS, 0.01% Triton X-100 (Sigma, St Louis, MO) (four times for 5 minutes each), incubated with the secondary biotinylated antibodies (1:200 or 1:500 in PBS, 0.1% BSA) for 30 minutes,rinsed in PBS, 0.01% Triton X-100 (four times for 5 minutes each) and incubated with peroxidase-conjugated ExtrAvidin (1:50 in PBS, 0.1% BSA) for 30 minutes. The peroxidase reaction was developed using either 3-amino-9-ethylcarbazole (AEC) or 3,3′-diaminobenzidine tetrahydrochloride (DAB) as chromogens. All incubations were performed at room temperature. After rinsing in distilled water, sections were dehydrated in ascending ethanol solutions, cleared in xylene and mounted.

In vitro generated tissues were decalcified in neutral buffered 10% EDTA or left undecalcified. After washing in PBS, samples were postfixed for 1 hour at 4°C in 1% osmium tetroxide in cacodylate buffer, rinsed in water,dehydrated through graded ethanol solutions, transferred in propylene oxide,and embedded in epoxy resin (Araldite™). Semithin sections were stained with Azur II-Methylene Blue to select appropriate fields; ultrathin sections were cut with diamond knives, placed on uncoated grids, contrasted with uranyl acetate and lead cytrate and examined with a CM 10 Philips transmission electron microscope.

BMSC pellets cultured in the presence of TGFβ1 generated a solid three-dimensional tissue structure that could be harvested and processed intact for histology. By day 7, a clear-cut chondroid morphology (substantial amounts of basophilic, alcianophilic, metachromatic matrix containing cells encased in chondrocytic lacunae) was formed(Fig. 1). Strong diffuse immunoreactivity for type II collagen was seen as early as day 7(Fig. 2). Interestingly, a thin peripheral rim of matrix remained essentially unlabelled.

The in vitro generated cartilage `beads' grew in size progressively,reaching a plateau at day 14 (Fig. 2). Alcianophilia and metachromasia, two histochemical features of proteoglycan content in cartilage, were uniform throughout the cross-sectional areas of the cartilage `beads' up until day 14, when the beads reached their maximum size. After day 14, both alcianophilia and metachromasia continued to increase in intensity in the central region of the beads. After day 21, they began to disappear from a progressively wider peripheral zone. By day 28, a clear-cut zonal pattern had formed, with a non-basophilic collar of cellular tissue, up to 300 μm thick, encasing a central core of histologically and histochemically well-defined hyaline cartilage.

Immunoreactivity for type X collagen was not seen until day 21, at which point it became distinctive in individual cells and throughout the extracellular matrix in the central cartilaginous region of the tissue bead(data not shown), indicating progression to hypertrophy of chondrocytes differentiated in pellet culture.

BMSC pellets cultured in the absence of TGF-β1 (control cultures) did not grow in size, nor did they generate any tissue structure reminiscent of cartilage (data not shown).

BMSC pellets which had been cultured in the presence of TGF-β1 for 28 days were transferred to mineralization medium and cultured for an additional 1-3 week period. Within one week of culture under conditions conducive to in vitro mineralization, the peripheral collar of tissue had turned into a tissue histologically reminiscent of bone (Fig. 3). The tissue was fully mineralized, whereas the central core of hyaline cartilage had remained uncalcified, as demonstrated by alizarin red S or von Kossa staining. Apparently, viable cells were encased in lacunar spaces within the fully calcified bone-like matrix, reminiscent of osteocytes. The bone-like matrix appeared as woven bone in polarized light microscopy. In essence, a bony collar had formed around a cartilage core, closely mimicking the events occurring during the early phases of endochondral ossification.

Electron microscopic analysis (Fig. 4) confirmed the occurrence of mineralization and demonstrated features typical of genuine cartilage and bone matrix. Central areas of the cartilage beads, exhibiting histological features of hyaline cartilage,contained widely spaced, thin, non-banded collagen fibrils and abundant proteoglycan granules. At the transition between the central areas and the outer collar of non-chondroid tissue, mineralization nodules appeared in the context of a typical cartilaginous matrix structure. More peripherally, thick and periodically banded collagen fibrils became predominant. In the outer portion of the peripheral collar, dense bundles of banded fibrils and extensive calcification were observed.

Because both bone- and cartilage-like matrix formed in our system, we investigated the expression of characteristic bone matrix proteins during the process leading to the in vitro generation of an osteochondral `organoid'. To investigate the expression of type I collagen, we used a monoclonal antibody recognizing the N-propeptide of type I procollagen, which is cleaved extracellularly following secretion of procollagen molecules. Thus,intracellular immunoreactivity could be taken as evidence of procollagen synthesis and extracellular immunoreactivity as indicative of sites of initial deposition associated with freshly secreted procollagen molecules. Expression and initial deposition of type I collagen were spatially restricted to an outer layer of tissue that progressively increased in thickness over time(Fig. 5). Extracellular immunoreactivity was restricted at all times to a thin rim of matrix marking the outer boundary of the hyaline cartilage region. Intracellular immunoreactivity for type I collagen N propeptide was detected in a progressively thicker region of peripheral tissue. From day 14 onward,extracellular immunoreactivity for type I collagen N-propeptide was lost in the same region, probably as a result of matrix maturation, whereas immunolabelling of procollagen producing cells remained distinctly detectable.

Bone sialoprotein is specifically expressed during early stages of embryonic bone formation, and so we also analysed the production of BSP during in vitro perichondral osteogenesis in the pellet culture system(Fig. 6). BSP-producing cells were detected as early as day 7 and increased in number thereafter. At all time points, BSP-producing cells were spatially confined to the outer portion of the cartilage `beads'. At day 14, they formed a distinct peripheral region of BSP production. Between days 21 and 28, when physical growth of the beads had ceased, production of BSP shifted from the outermost portion of the beads to an adjacent, more central region. Overall, the spatial and temporal pattern of cellular immunolabelling for BSP and type I collagen N propeptide were closely similar to one another. Extracellular immunoreactivity for BSP only appeared in cultures exposed to mineralization-conducive conditions (day 28-42).

No immunoreactivity for osteocalcin was observed at any time during the chondrogenic period (days 1-28). Abundant extracellular immunoreactivity for osteocalcin was, by contrast, seen in cultures harvested during the mineralization period (days 35-49) within the peripheral rind of bone-like tissue (Fig. 6).

Direct evidence of osteogenesis in pellet cultures of BMSCs(Johnstone et al., 1998; Mackay et al., 1998; Mastrogiacomo et al., 2001; Sekiya et al., 2001; Yoo et al., 1998) has never been reported before. Likewise, so-called `mineralization nodules'(Beresford et al., 1993; Jaiswal et al., 1997; Maniatopoulos et al., 1988; Martin et al., 1997) but not histologically proved, three-dimensional bony structures form in cultures of adherent osteogenic cells (reviewed by Bianco and Robey, 1999; Bianco and Gehron Robey, 2000)and in vivo transplantation assays(Beresford, 1989; Goshima et al., 1991; Krebsbach et al., 1997; Kuznetsov et al., 1997; Martin et al., 1997) are ultimately required to prove the osteogenic capacity of a test cell strain. We have provided evidence that a fully mineralized tissue with the histological,ultrastructural and immunohistochemical characteristics of bone (as observed in vivo) does form in pellet cultures of BMSCs exposed sequentially to chondrogenic and mineralization-conducive conditions. Thus, our modification of the pellet-culture system provides a relatively simple in vitro assay for testing the osteogenic capacity of putative osteogenic cells strains.

Considering that the pellet-derived tissue beads stop growing in size around day 14, the subsequent deposition of peripheral rind of bone-like matrix must occur at the expenses of pre-existing cartilage. Thus, an internal remodelling of cartilage to bone matrix occurs once the pellet-derived tissue has reached its maximum size. Assuming a roughly spherical shape for the tissue bead formed in vitro, approximately one-third of the volume of unmineralized cartilage existing at day 14 is physically replaced by an equal volume of bone-like tissue. Given the very different water contents of unmineralized hyaline cartilage and mineralized bone, the actual matrix remodelling taking place is even greater. The internal remodelling of cartilage into bone in our system might reflect either a direct phenotypic conversion of chondrogenic cells to an osteogenic phenotype or the local selection over time of an osteogenic population replacing a chondrogenic one. The former possibility would be consistent with prior in vitro and in vivo evidence (Berry et al., 1992; Gentili et al., 1993; Galotto et al., 1994),suggesting an eventual osteogenic fate of hypertrophic chondrocytes. The non-clonal nature of our cell population prevents inferences about a direct conversion of chondrocytes to osteoblasts. For the same reason, whether chondrogenesis and osteogenesis in our cultures reflect the dual differentiation potential, the existence of a single stromal progenitor cannot be conclusively stated. The specific point of interest in our present results rather rests upon the precise spatial and temporal pattern that chondro- and osteogenesis obey in the pellet-culture system.

In vivo, embryonic bone formation occurs around cartilaginous anlagen(rudiments) prior to the onset of endochondral ossification proper. This peculiar spatial arrangement is closely mimicked in our culture system, in which a rind of bone-like tissue forms around a core of hyaline and mineralizing cartilage. Our system thus directly models the precise spatial determinants of cell differentiation operating in development. Interestingly,whereas cartilage only forms occasionally in open transplants of BMSCs(Krebsbach et al., 1997; Kuznetsov et al., 1997; Martin et al., 1997),cartilage and bone tissues that do form in closed transplantation systems(diffusion chambers) (Ashton et al.,1980; Bab et al.,1984; Gundle et al.,1995) are organized in a spatial pattern similar to the one we observe in vitro (cartilage in the interior, bone in the periphery). It is plausible that a gradient of oxygen tension(Ashton et al., 1980; Scott, 1992) might be directly involved in the generation of the spatial pattern observed in both systems.

Although chondrogenesis and osteogenesis are spatially segregated in our system, they partially overlap temporally. Initiation of osteogenesis occurs simultaneously with chondrogenic differentiation but can only be completed upon switching to mineralization-conducive conditions. During the chondrogenesis phase, type I collagen and BSP are actively synthesized and the proteoglycan content is reduced in the region that is to become mineralized,bone-like tissue. By contrast, it is the exposure to a mineralization-conducive environment that induces the production and deposition of osteocalcin, and the deposition (but not the production) of BSP. In vivo, production of BSP by differentiating osteoblasts in the presumptive bony collar and in the adjacent outermost chondrocytes occurs simultaneously and `primes' the subsequent deposition of the bony collar(Bianco et al., 1991; Bianco et al., 1993; Bianco et al., 1998; Riminucci et al., 1998). Just as our system models the formation of the bony collar in vitro, so our data on the localization of BSP recapitulate the specific spatial and temporal pattern of expression of BSP associated with the in vivo events.

The pellet culture of human BMSCs is generally taken as a good in vitro model of chondrogenesis. By showing the occurrence of bone deposition in the same system, our data highlight a hitherto unrecognized experimental benefit of the system. That is, the direct transition, in vitro, from a cell culture to an organ culture dimension. A number of specific, temporally defined events of bone morphogenesis, such as endochondral bone formation proper or ontogeny of the bone marrow, could in principle be modelled and dissected in terms of their molecular determinants by further exploring the experimental flexibility of the system.

This work was supported by grants from MIUR, Ministero della Sanita' and funds from the Italian (ASI) and European (ESA-ERISTO) Space Agencies.