Osteoclast formation in bone is supported by osteoblasts expressing receptor activator of NF-κB ligand (RANKL) and macrophage colony-stimulating factor (M-CSF) expression. Numerous osteotropic factors regulate expression levels of RANKL and the RANKL decoy receptor osteoprotegerin (OPG) in osteoblasts, thereby affecting osteoclast differentiation. However, not only is RANKL widely expressed in soft tissues, but osteoclasts have been noted in extraskeletal lesions. We found that cultured skin fibroblastic cells express RANKL, M-CSF, and OPG messenger (mRNA). Stimulation by 1α,25 dihydroxyvitamin D3 [1,25(OH)2D3] plus dexamethasone (Dex) augmented RANKL and diminished OPG mRNA expression in fibroblastic cells and caused the formation of numerous osteoclasts in cocultures of skin fibroblastic cells with hemopoietic cells or monocytes. The osteoclasts thus formed expressed tartrate-resistant acid phosphatase (TRAP) and calcitonin (CT) receptors and formed resorption pits in cortical bone. Osteoclast formation also was stimulated (in the presence of Dex) by prostaglandin E2 (PGE2), interleukin-11 (IL-11), IL-1, tumor necrosis factor-α (TNF-α), and parathyroid hormone-related protein (PTHrP), factors which also stimulate osteoclast formation supported by osteoblasts. In addition, granulocyte-macrophage-CSF (GM-CSF), transforming growth factor-β (TGF-β), and OPG inhibited osteoclast formation in skin fibroblastic cell-hemopoietic cell cocultures; CT reduced only osteoclast nuclearity. Fibroblastic stromal cells from other tissues (lung, respiratory diaphragm, spleen, and tumor) also supported osteoclast formation. Thus, RANKL-positive fibroblastic cells in extraskeletal tissues can support osteoclastogenesis if osteolytic factors and osteoclast precursors are present. Such mesenchymally derived cells may play a role in pathological osteolysis and may be involved in osteoclast formation in extraskeletal tissues.
OSTEOCLAST DIFFERENTIATION, a critical determinant of bone resorption in vivo, is stimulated by a range of osteolytic factors, including 1α,25 dihydroxyvitamin D3 [1,25(OH)2D3], parathyroid hormone (PTH) and PTH-related protein (PTHrP), interleukin-1 (IL-1), tumor necrosis factor α (TNF-α) and TNF-β, prostaglandins (PGs), and gp130 receptor-mediated cytokines such as IL-11.(1) These factors act indirectly, by stimulating expression of the TNF-related membrane bound molecule receptor activator of NF-κB ligand (RANKL) in osteoblasts and bone marrow stromal cells, and by reducing their secretion of the RANKL decoy receptor osteoprotegrin (OPG).(2–4) Indeed, it is probably the ratio of RANKL to OPG expression rather than RANKL expression levels per se that determines the degree of osteoclastogenic stimulus. RANKL is essential for osteoclast differentiation(5) and acts in concert with macrophage colony-stimulating factor (M-CSF) on osteoclast precursors, which are mononuclear phagocytes found chiefly in the bone marrow, spleen, and the circulation but which also are detected in most soft tissues including inflammatory and neoplastic lesions.(6–9)
RANKL expression is not confined to bone,(3,10) reflecting both its known involvement in the immune system and the likelihood that hitherto unrecognized roles for RANKL exist in other physiological systems. We have previously described the expression of RANKL protein and messenger RNA (mRNA) mouse extraskeletal tissues, such as skin.(11) Osteoclasts are not normally found in such tissues, but a number of soft tissue lesions are associated with osteoclasts or osteoclast-like cells, including giant cell tumors of tendon sheath, pilar tumor of scalp, and, occasionally, breast carcinomas.(12–16) It is possible that, as in bone, it is the mesenchymally derived stromal cell elements in these lesions that support osteoclasts formation and that, if so, such stromal cells may respond to osteolytic stimuli differently than to osteoblasts. Extraskeletal osteoclastogenic stromal cells also may play a role in pathological osteolysis associated with more common neoplastic or chronic inflammatory lesions. We therefore investigated the ability of murine fibroblastic stromal cells from skin and other tissues to support osteoclast differentiation in vitro.
MATERIALS AND METHODS
Reagents and cells
Dexamethasone (Dex) and prostaglandin E2 (PGE2) were obtained from Sigma Chemical Co. (St. Louis, MO, U.S.A.). 1,25(OH)2D3 was obtained from the Wako Pure Chemical Co. (Osaka, Japan). Human recombinant IL-1α, IL-11, transforming growth factor β (TGF-β), TNF-α, and mouse recombinant granulocyte-macrophage-CSF (GM-CSF) were obtained from R & D Systems (Minneapolis, MN, U.S.A.). Recombinant human PTHrP(1-34) and salmon calcitonin (CT) were synthesized on an Applied Biosystems 433A Synthesizer (Applied Biosystems, Inc., Foster City, CA, U.S.A.). Recombinant human M-CSF was a kind gift of the Genetics Institute (Cambridge, MA, U.S.A.); recombinant human OPG was a kind gift from Amgen, Inc. (Thousand Oaks, CA, U.S.A.). Cells were cultured in α-minimal essential medium (α-MEM; Life Technologies, Grand Island, NY, U.S.A.) containing 10% fetal calf serum (CSL Biosciences, Parkville, Australia; MEM/fetal bovine serum [FBS]) supplemented with penicillin and streptomycin (CSL Biosciences). All tissues used in this study were obtained from freshly killed c57BL/6J mice (Monash University, Clayton, Australia) except melanoma tumor tissue, which was obtained from transgenic UE-Ty-SV40 mice, which were derived from mouse strain FVB/n(17) and were kindly provided by Dr. W. Murphy of the Peter Macallum Institute of Cancer Research (Melbourne, Australia); these mice spontaneously developed melanoma primary and secondary tumors.
Preparation of cells
Mouse spleen hemopoietic cell suspensions were prepared by crushing adult (6-week old) mouse spleens through a fine (100 μm) metal sieve; this cell suspension was used as a source of hemopoietic osteoclast precursors without further purification. Peripheral blood mononuclear cells (PBMCs) were obtained by diluting mouse blood (1:4 in Hanks' balanced salt solution; Life Technologies, Grand Island, NY, U.S.A.) layering this over Ficoll-Paque solution (Pharmacia Biotech, Uppsala, Sweden) and then centrifuging (693g), washing, and resuspending in MEM/FBS. Osteoblasts were prepared by the sequential digestion method from newborn mouse calvaria.(18) Mouse skin fibroblastic cells were prepared from approximately 1 cm2 of dorsal skin from either the newborn mice or the shaved dorsal skin from freshly killed adult mice. The skin was cut into 2 mm2 pieces, rinsed in phosphate-buffered saline (PBS), digested in 2 mg/ml collagenase (type 2; Worthington Corp., Freehold, NJ, U.S.A.),at 37°C for 1 h, and the resulting dispersed cells were centrifuged and resuspended in MEM/FBS and incubated in a 100-mm diameter petri dish until confluent. Explanted (approximately 2 mm2) pieces of newborn mouse lung, respiratory diaphragm, and spontaneous UE-Ty-SV40 melanoma secondary tissue were incubated in a 100-mm-diameter petri dish until proliferating fibroblast-like cells around the explants reached confluence, where upon the tissue fragments were removed and remaining adherent fibroblastic cells were dispersed by trypsinization. Spleen fibroblastic cells were obtained by incubating a suspension of total spleen cell (prepared as described above; 106 cells/cm2) on plastic for approximately 2 weeks in MEM/FBS and proliferating fibroblastic cells removed for use by trypsinization. Before use in cocultures or for reverse-transcription polymerase chain reaction (RT-PCR) analysis, osteoblastic and fibroblastic cells obtained as described above were grown to confluence in 60-mm petri dishes.
Detection of RANKL, OPG, and M-CSF expression in cultured fibroblastic cells
RNA was extracted from confluent fibroblastic cell cultures stimulated by 20 nM 1,25(OH)2D3 and 100 nM Dex for the periods indicated. RNA extraction, complementary DNA (cDNA) synthesis, PCRs, and Southern transfer of PCR products were performed as previously described.(4) Oligonucleotide sequences employed to detect and verify M-CSF expression, including all splice variants, were as follows: primer M-CSF-1 (TGT TCT ACA AGT GGA AGT GGA, nucleotides 1719–1743) and primer M-CSF-2 (TCA AGG AAG ACA ACC GTC CCAC, nucleotides 1960–1981) for 28 cycles, annealing temperature 58°C, and products verified as described(4) with oligonucleotide M-CSF-3 (CAG AGG GAC ATT GAC AAA CG, nucleotides 1856–1875).
Fibroblastic or osteoblastic cells were added to 10-mm plastic culture wells (4 × 104 cells/well). Spleen hemopoietic cells were then added to the wells (5 × 105 cells/well)(1,2,4) and the cells were incubated for 7 or 11 days in the presence of added mediators. In some cocultures, monocytes were cocultured with stromal cells instead of spleen hemopoietic cells; PBMCs were added to wells (5 × 105 cells/well), incubated for 2 h, nonadherent cells were removed by vigorous rinsing, and stromal cells were added. In all cultures, medium and mediators were removed and replenished every 3 days. Cultures were formaldehyde fixed at endpoint and histochemically stained for tartrate-resistant acid phosphatase (TRAP) as previously described;(19) the total number of TRAP-positive cells and the number of TRAP-positive multinucleated cells (MNCs) per well was then determined. In some wells, cells were cocultured on 6-mm glass coverslips, which were removed at 7 days to detect CT receptors by radiolabeled ligand binding and autoradiography as previously described,(20)and then stained for TRAP. Alkaline phosphatase expression also was investigated by histochemical staining(11) in some cultures.
Bone resorption assays
Spleen hemopoietic cells and skin fibroblastic cells (106 cells/cm2 and 105 cells/cm2, respectively) were settled on a plastic surface (12 cm2) coated with collagen gel (Cellmatrix 1A; Nitta Gelatin, Inc., Osaka, Japan) and incubated for 11 days in the presence of 20 nM 1,25(OH)2D3 and 100 nM Dex as previously described.(21) Cells were then harvested using collagenase (2 mg/ml), centrifuged, and the cells resuspended were then added to tissue culture wells containing bovine cortical bone slices, as previously described.(22) After 3 days of incubation, bone slices were removed and stripped of cells using 0.25 M NH4OH and the surface was examined for resorption pits on a scanning electron microscope.(22) In addition, similar spleen cell/skin fibroblastic cell cocultures were set up directly on cortical bone slices, incubated for 11 days, and processed for scanning electron microscopy.
Data were analyzed using Student's 2-tail t-test where indicated.
Response of stromal cells to treatment with 1,25(OH)2D3 and Dex
RANKL, OPG, and M-CSF mRNA were expressed by cultured skin fibroblastic cells. RANKL mRNA expression was increased by 1,25(OH)2D3/Dex by 8 h compared with control, and this increase was sustained until the 24-h time point (Fig. 1A). OPG levels were, in contrast, greatly reduced after 4 h of 1,25(OH)2D3/Dex treatment. An increase in the ratio of RANKL mRNA to OPG mRNA expression was similarly noted in lung and respiratory diaphragm fibroblastic cells (Fig. 1B). M-CSF mRNA levels were not affected by treatment with 1,25(OH)2D3/Dex.
Cultured fibroblastic stromal cells from skin were of typical spindle-shaped morphology, and cells of epithelial morphology or adipocyte appearance were not observed. In contrast to cultured calvarial osteoblasts, which were uniformly strongly alkaline phosphatase positive, among skin fibroblastic cell populations alkaline phosphatase-positive cells were sparse (<50/10-mm-diameter well) and indeed were often absent entirely. Coculture of the fibroblastic stromal cells with spleen hemopoietic cells (total spleen cell population) or monocytes in the presence of 1,25(OH)2D3 and Dex resulted in numerous TRAP-positive and CT receptor mononucleated cells and MNCs (Fig. 2) after 7 or 11 days of incubation. No increase in alkaline phosphatase expression was seen. All TRAP-positive MNCs were also CT receptor positive, showing that they were bona fide osteoclasts. This was confirmed with cells extracted from spleen cell/skin fibroblastic cell cocultures incubated on collagen gels, which formed numerous bone resorption pits on cortical bone (Fig. 2); resorption pits were also noted in cocultures incubated directly on bone.
Hemopoietic spleen cells cocultured for 11 days with calvarial osteoblasts or with fibroblastic cells derived from skin, lung, spleen, respiratory diaphragm, and secondary melanoma similarly formed TRAP-positive MNCs when stimulated by 1,25(OH)2D3 and Dex (Figs. 2 and 3A), although considerable variation was noted in the osteoclastogenic capacity of fibroblastic cells from different tissues. Skin fibroblastic cells from both neonatal and adult skin supported osteoclast formation (Fig. 3B). Hemopoietic cell/skin fibroblastic cell cocultures stimulated by 1,25(OH)2D3, TNF-α, or IL-1α did not result in TRAP-positive MNC formation (Fig. 4); PTHrP, PGE2, and IL-11 stimulation resulted in very small numbers of the TRAP-positive MNCs (<0.2 per well; Fig. 4). In contrast, when these factors were added in the presence of 100 nM Dex, numerous TRAP-positive cells were noted (Fig. 4). Addition of M-CSF to hemopoietic cell/skin fibroblastic cell cocultures had no significant effect on osteoclast numbers (data not shown). Osteoclast formation in hemopoietic cell/skin fibroblastic cell cocultures [stimulated for 7 days by 1,25(OH)2D3 and Dex] was strongly inhibited by 100 ng/ml OPG (Fig. 5A), thereby showing dependence on RANKL stimulation. GM-CSF (10 ng/ml) or 10 ng/ml TGF-β (10 ng/ml) inhibited osteoclast formation and no TRAP-positive cells were found in these cultures at all (Fig. 5B). In contrast, CT greatly reduced TRAP-positive MNC formation but did not reduce total TRAP-positive cell numbers (Fig. 5B).
Osteoclasts are multinucleated bone resorbing cells derived from mononuclear phagocyte precursors found in the circulation and in hemopoietic tissues such as bone marrow and spleen. RANKL causes osteoclast formation both in vivo and in vitro.(2,5) Our previous findings of strong RANKL expression by cells (including mesenchymally derived cells) in extraskeletal tissues and of osteoclast formation supported in vitro by activated RANKL expressing T cells(11,23) raised the possibility that mesenchymally derived cells other than osteoblasts and bone marrow stromal cells also support osteoclast formation. Our data confirm this is the case; fibroblastic cells derived from skin (both adult and newborn) were as efficient as calvarial osteoblasts in supporting osteoclast formation from hemopoietic cell precursors when stimulated by 1,25(OH)2D3 and Dex. All TRAP-positive MNCs expressed CT receptors and, when extracted from the cocultures, formed numerous bone resorption pits on cortical bone, verifying their phenotype as that of functional osteoclasts. Similar stimulated fibroblastic stromal cells from lung, spleen, respiratory diaphragm, and secondary melanoma tumor cocultured with hemopoietic cells also supported osteoclast formation. Considerable variability was noted in the osteoclastogenic ability of fibroblastic cells from different tissues. Whether this variation represents inherent phenotypic differences in the cells or is a reflection of their proliferation and differentiation response in culture is unclear, but it must be emphasized that although RANKL and M-CSF play a critical role in initiating osteoclast formation, many other factors influence osteoclast differentiation and survival. These include mesenchymally derived factors such as PGs, GM-CSF, IL-1, and IL-18.(1,19,24)
The role of RANKL in osteoclastogenesis in these cocultures was confirmed by its abolition when high levels of OPG were added to the cultures. Furthermore, we found that 1,25(OH)2D3/Dex costimulation elevated RANKL mRNA levels in skin, lung, and respiratory diaphragm fibroblastic cells after 8–24 h, whereas OPG levels were diminished. This pattern of RANKL and OPG regulation in skin fibroblastic cells is strikingly similar to that previously found in mouse calvarial osteoblasts(4) and, if reflected in protein expression levels, would lead to an increase of bioavailable RANKL. Because M-CSF, the other factor essential for osteoclast formation, was not regulated in these cells and was not limiting osteoclast formation in the cocultures, osteoclast formation probably is determined primarily via RANKL/OPG regulation in these systems. However, although osteoclastogenesis supported by fibroblastic cells was strikingly similar in so many respects to that supported by osteoblasts, a notable exception was its dependence on the presence of Dex. The presence of Dex is not essential for osteoblast-supported osteoclastogenesis but does markedly increase it, probably by further suppressing OPG levels.(25) The requirement for Dex in fibroblastic cell-supported cocultures is unclear. It may reflect differences between osteoblasts and fibroblasts in their regulation of RANKL and OPG or in their regulation of stromal cell-associated stimulators (e.g., soluble IL-6 receptor) or inhibitors (e.g., IL-18 and GM-CSF) of osteoclastogenesis. The development of sufficiently sensitive assays for RANKL and OPG protein would enable this to be further investigated.
The normal function of RANKL in extraskeletal connective tissues is unclear. RANKL is known to play a role in the immune system, specifically increasing dendritic cell survival, naive T cell proliferation, and dendritic cell-T cell interaction. On the basis of studies in which the gene for RANKL or its receptor (RANK) were ablated, roles in lymph node organogenesis, B cell development, and control of hemopoiesis levels have been suggested.(3,5,26) RANK has been detected in many tissues but the precise cellular distribution has not been described, although expression by dendritic cells and mononuclear phagocytes is known.(10) The physiological target of the RANKL expressed in connective tissue fibroblasts is therefore unclear, but a role in some of the above immune-related functions is possible. However, despite the presence of RANKL, it is clear that osteoclasts are not normally found in skin or other soft tissues. This may be caused by a lack of local osteoclast precursors or of osteolytic factors or by the requirement of local fibroblasts for Dex or other additional stimulus. Nevertheless, our results indicate that fibroblasts present in sclerotic, granulomatous, or neoplastic lesions could contribute to increased osteolysis if large numbers of mononuclear phagocytes are present. Supporting this notion, we have found that fibroblastic cells grown from a spontaneous melanoma supported osteoclastogenesis. Tumors that invade bone often replace local bone marrow; however, because tumor infiltrating macrophage populations include many cells that are osteoclast precursors,(8,9) these results show that invading tumors carry with them both the necessary cellular components for osteolysis to proceed. Fibroblastic cell involvement in the initiation of bone resorption also may occur in inflammatory lesions in bone. For example, in rheumatoid arthritis bone destruction frequently is associated with local pannocyte invasion. Fibrous membrane tissue also is found at the bone surface around failed joint prosthesis implants that have undergone aseptic loosening, a process in which pathological bone resorption plays a significant role in undermining prosthesis adhesion. Furthermore, osteoclast-like MNCs have been noted in a range of soft tissue lesions.(12–16) Although osteoclast precursors are present in macrophage populations in neoplastic and inflammatory lesions, whether significant RANKL levels are found in the above lesions is yet to be investigated. Our results nevertheless indicate that the precise phenotype of the local mesenchymally derived cells does not in itself determine whether an osteoclastogenic stimulus is present.
The observation that a small stromal cell population in the spleen is capable (when greatly expanded) of supporting osteoclastogenesis raised the possibility that when spleen hemopoietic cells are used as the source of osteoclast precursors, it is the spleen stromal cells therein that cause osteoclasts to form. However, spleen cells cultured alone and stimulated by 1,25(OH)2D3 and Dex do not form osteoclasts. We also found that skin fibroblastic cells (like osteoblastic cells and preadipocytic marrow stromal cells (1,27)) support osteoclast formation from monocytes, proving that the skin cells themselves supply the stimulus. Spleen stromal cell support of osteoclastogenesis also has been noted by Owens et al.,(28) the only previous report of extraskeletal osteoclastogenic stromal cells. Curiously, osteoclasts formed by their culture method were not functional (i.e., did not resorb bone) without further addition of stimulated osteoblasts. The reason for this is unclear, although our data indicates that the osteoclastogenic stimulus supplied by cultured spleen stromal cells is weak. It is possible that RANKL levels were insufficient in the culture system of Owens et al. to cause osteoclast activation (like IL-1, RANKL activates osteoclasts (29)) or that the thick extracellular matrix elaborated by the stromal cells provides a barrier between osteoclasts and the bone surface, a barrier that stimulated calvarial osteoblasts can remove.(30)
Another feature previously noted in osteoblast-based cocultures is that OPG, TGF-β, and GM-CSF inhibited not only TRAP-positive MNC formation but the formation of TRAP-positive mononuclear cells,(2,19,31) which is consistent with the status of the TRAP-positive mononuclear cells (some of which express CT receptors) as osteoclast precursors and mononuclear or “prefusion” osteoclasts. In contrast to these inhibitors, CT inhibited MNC formation without reducing total TRAP-positive cell numbers. The effects of CT on osteoclast differentiation are controversial because CT receptor expression occurs as a late event in the process, but our work suggests that rather than affecting osteoclast differentiation, CT reduces osteoclast fusion, an effect previously noted in other coculture systems.(32) However, CT effects on osteoclast differentiation cannot be clarified easily as the most specific osteoclast phenotypic markers, CT receptor expression and bone resorption, cannot be used; CT both strongly inhibits its own receptor expression and directly inhibits osteoclastic bone resorption.(32)
In conclusion, we have found that connective tissue fibroblastic cells present in extraskeletal tissues support osteoclastogenesis, a capability previously showed in osteoblasts and preadipocytic bone marrow stromal cells that are also mesenchymally derived. Expression of functional RANKL by such stromal cells (when stimulated by osteolytic factors) could play a role in pathological osteoclast formation and osteolysis or in other RANKL-dependent cell interactions. Fibroblastic cells such as those derived from skin also could be used as a surrogate for osteoblasts in in vitro studies of osteoclastogenesis in which osteoblasts are not easily obtainable, as is the case with studies employing certain transgenic mouse strains.
J. Quinn is a University of Melbourne C.R. Roper Fellow, and M.T. Gillespie is an National Health & Medical Research Council, Australia (NHMRC) Senior Research Fellow. N. Horwood is the recipient of an NHMRC Dora Lush Scholarship. This work also has been supported by a NHMRC Program Grant (963211, M.T. Gillespie and T.J. Martin) and grants from Aza Research Pty Ltd., Australia, and from Chugai Pharmaceutical Co. Ltd., Japan.