The Role of Osteoblasts in the Hematopoietic Microenvironment


  • Dr. Russell S. Taichman,

    Corresponding author
    1. Department of Periodontics/Prevention/Geriatrics, University of Michigan Dental School, Ann Arbor, Michigan, USA
    • Department of Periodontics/Prevention/Geriatrics, University of Michigan School of Dentistry, 1011 North University Avenue, Ann Arbor, Michigan 48109-1078, USA
    Search for more papers by this author
  • Stephen G. Emerson

    1. Departments of Internal Medicine & Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, USA
    Search for more papers by this author


Hematopoietic stem cell differentiation occurs in direct proximity to osteoblasts within the bone marrow cavity. Despite this striking affiliation, surprisingly little is known about the precise cellular and molecular impact of osteoblasts on the bone marrow microenvironment. Recently, it has been proposed that human osteoblasts support the growth of primitive human hematopoietic cells in vitro and possibly in vivo. Evidence to support this hypothesis is reviewed as follows: the influence of osteoblasts on osteoclast development; the participation of osteoblasts in long-term bone marrow cultures; the production of positive hematopoietic regulatory molecules by osteoblasts; the production of cell-cycle inhibitory factors by osteoblasts, and cell-cell interactions between early hematopoietic cells and osteoblasts.

Introduction and Historical Notions: Do Osteoblasts Participate in the Regulation of Hematopoiesis in the Bone Marrow Cavity?

After birth, blood cells are produced in the bone marrow. The development of the bone marrow cavity is a coordinated process in which blood precursors migrate and colonize spaces carved out of embryonic bone and cartilage. Very early in life, an intimate physical association between blood cells and bone cells is established in this microenvironment. This paper will explore whether this spatial relationship between bone and blood also mirrors a functional interdependence between the two tissues: in particular, do osteoblasts or osteoblast-derived factors play a role in hematopoiesis?

It has long been appreciated that bone marrow stromal cells (BMSCs) provide the structural scaffolding for hematopoiesis [1-5]. More recent investigations in animals and humans have shown that BMSCs (reticular fibroblasts, macrophages, adipocytes, and endothelial cells) produce several factors critical for the development of blood cells. In brief, these studies have demonstrated that direct stromal cell-to-blood cell contact, stromal cell production of the extracellular bone marrow matrix, and cytokine synthesis by BMSCs are implicated in the formation of various blood cells [6-10]. Although osteoblasts are also part of the stromal cell support system in the bone marrow and may be derived from a common precursor, little is known about their relationship to primitive hematopoietic cells. It is relevant, however, that primitive hematopoietic cells are closely approximated with endosteal surfaces, rather than being randomly distributed throughout the marrow cavity [1-4] (Fig. 1). Thus, there are good reasons to suspect that osteoblasts influence hematopoiesis [11, 12] Evidence to further support this hypothesis will be reviewed as follows: the influence of osteoblasts on osteoclast development; the participation of osteoblasts in long-term bone marrow cultures; the production of positive hematopoietic regulatory molecules by osteoblasts; the production of cell-cycle inhibitory factors by osteoblasts, and cell-cell interactions between early hematopoietic cells and osteoblasts.

Figure Figure 1..

Bone marrow architecture.

Osteoblast-Osteoclast Interactions as a Paradigm for Osteoblast-Hematopoietic Cell Interactions

In developing a rationale for the hypothesis that osteoblasts are involved in the regulation of hematopoiesis, it may be useful to consider some aspects of the relationship between osteoblasts and osteoclasts. Although clearly not primitive, osteoclasts are derived from CD34+ hematopoietic cells responsible for the resorption of mineralized tissues. Consequently, there may be similarities in the way that osteoblasts affect the behavior of osteoclasts and primitive hematopoietic cells. It has not been entirely resolved whether osteoclasts represent a distinct lineage from monocyte/macrophages or originate from the fusion of monocyte precursors. Nonetheless, bone marrow transplantation and chick-quail chimera investigations clearly demonstrate that these cells are of hematopoietic origin [13-16]. Functionally, osteoclasts form calcitonin-inhibitable resorption pits when cultured on mineralized tissues, strongly express tartrate-resistant acid phosphatase, form multicellular syncytium, which may contain up to 25 nuclei per cell, and membrane specializations called ruffled borders, which presumably localize the resorptive activity [15-17].

Recent studies indicate that osteoblasts may regulate bone resorption by inducing the expansion, maturation, and activation of osteoclast precursors. While the specific nature of the osteoblast-derived activities which modulate and/or expand the osteoclast precursor pool is not known, osteoblasts interact with hematopoietic precursors to orchestrate the production of osteoclasts. In osteosclerotic (oc/oc) mice, for instance, numerous tartrate-resistant acid phosphatase positive cells expressing calcitonin receptors are formed in response to 1α,25-dihydroxyvitamin D3, macrophage colony-stimulating factor (M-CSF) and osteoblasts [17]. Furthermore, rat hematopoietic progenitors differentiate into mononuclear preosteoclasts capable of resorbing bone in the presence of osteoblasts [18]. Alternatively, there are data that suggest that both the formation and activation of osteoclasts may be “osteoblast-independent” events in some model systems [18, 19].

In addition to influencing osteoclast development, osteoblasts are involved in osteoclast activation [20]. Receptors for cytokines and hormones that trigger mineralized tissue resorption are rarely detected on osteoclasts. In many cases, osteoblasts appear to be the direct or primary targets of bone-resorbing stimuli such as PTH, prostaglandins, and 1α,25-dihydroxyvitamin D3 by expressing receptors for these agents [13]. Osteoblasts may provide the “secondary” resorption signal(s) to osteoclasts to activate and regulate bone removal. These molecules may include one or more of the following activities: activation of osteoclastic activities by small molecular weight peptides or prostaglandins; secretion of osteoclastic chemotactic signals by osteoblasts that attracts osteoclasts to mineralized surfaces; osteoblast production of proteolytic enzymes necessary for the enzymatic removal of osteoid, as osteoid covered bone is poorly resorbed; retraction of osteoblasts from bone surfaces, facilitating osteoclast access to the underlying mineral, or osteoblast expression of de novo adhesion molecules utilized by osteoclasts to localize to a particular resorption site [13, 20-22].

From the foregoing, it is clear that both hematopoietic precursors and osteoblasts have established communications for the production of osteoclasts and/or the activation of resorption. However, osteoblast participation in hematopoiesis beyond that of the osteoclast has not yet been fully addressed. In fact, this interaction with osteoclasts probably reflects only a small fraction of the total possible interactions between osteoblasts and blood cells. For example, by making the following assumptions, we can predict that the average daily output of osteoclasts reflects only 0.2% of the total daily hematopoietic output: the total rate of blood cell production per day is, on average, 383 billion cells [23], the average number of osteoclasts/mm2 of endosteal surface is 0.2 cells [24], and the total endosteal bone surface area is 1,000-5,000 m2 [25]. Admittedly, these calculations are simplistic. At a minimum, they fail to consider that the life span of an average osteoclast is likely to be longer than one day, that not all osteoclasts are associated with endosteal surfaces, and they are based on resting conditions. Nevertheless, even if these calculations are in error by as much as a log, they represent the fraction of the hematopoietic/microenvironment interaction that has yet to be investigated from the perspective of an osteoblast. Thus, our present knowledge as to osteoblast-blood cell interactions is rudimentary.

The Ontogeny of Osteoblasts in the Hematopoietic Microenvironment

Few would dispute that osteoblasts are primarily concerned with synthesizing the extracellular matrix of bone. Similarly, BMSCs are largely involved in supporting hematopoiesis. Where the lines of lineage commitment are drawn or whether there is some degree of plasty between the lineages is not known. Conceivably, a spectrum of hematopoietic-supporting activities could be displayed by osteoblasts based upon their stage of development and/or their anatomic/physiologic location, and the model system examined (Table 1). Furthermore, the maturational state of the target cells themselves may alter the functional behavior of osteoblasts [26]. These questions have been difficult to address due to the lack of appropriate lineage markers required to discriminate between immature osteoblasts and other BMSCs. Recent evidence linking hematopoietic-supportive BMSCs with osteoblasts has focused on the source of osteoprogenitor cells within the bone marrow. Several in vitro studies reveal that both primary and transformed BMSCs can acquire an osteoblastic phenotype to form bone-like tissues [27-29]. Furthermore, when implanted in in vivo diffusion chambers, bone tissue is formed [27-29]. Other in vitro studies carried out under serum-free conditions have shown that the nonadherent low-density bone marrow cells may develop into “osteoblast-like” cells which may mineralize their extracellular matrix [30, 31]. While these data clearly establish bone marrow stroma as a source of osteoprogenitor cells, the data do not directly address whether osteoblasts support hematopoiesis.

Table Table 1.. Human osteoblast-like culture systems available for investigations of osteoblast-hematopoietic cell interactions
Cell systemOrigin of systemMineralizationSupport for hematopoietic progenitorsReferences
  1. a

    Legend: HOB = primary human osteoblast-like cells (enzymatically released HOBs: ND); ND = No data.

HOBExplant outgrowths++[112]
MG-63Osteosarcoma[113, 114]
hFOBTransformed fetal HOB+ND[113, 115]
HOS TE 85Osteosarcoma[113, 116]
Marrow stromal cellMarrow stromal cell induced to osteogenic phenotype+ND[117]
SaOS-2Osteosarcoma+[113, 118]
U2 OSOsteosarcoma+[113, 119]

Long-term in vitro bone marrow culture systems which support limited myelopoiesis (Dexter cultures) or lymphopoiesis (Whitlock-Witte cultures) are dependent upon the formation of stromal cell layers. These adherent stromal cells produce several hematopoietic growth factors, albeit at low or subliminal levels, that support limited hematopoiesis [8, 10, 32-38]. What the role of osteoblasts is in these systems is not clear [7, 10]. It is clear, however, that not all bone marrow stromal cell elements support hematopoiesis in in vitro assays. From these observations, it is suggested that discrete cellular elements in the bone marrow have distinct hematopoietic supporting functions. These distinctions are probably based on expression of hematopoietic-supporting cytokines and/or receptors.

As previously stated, osteoblast-like cells have been observed within bone marrow stromal cell layers that support hematopoiesis and share several phenotypic characteristics with stromal cell lines [39-43]. For example, the murine bone marrow stromal cell lines BMS2 and +/+2.4 express high levels of alkaline phosphatase, collagen (I), and bone sialoprotein [43]. In addition, murine RNA for osteocalcin, an osteoblast-specific protein, is detected in BMS2 cells [43]. In a series of experiments using several stromal cell lines, Benayahu et al. found that all cell types examined (MBA-1 fibroblasts, MBA-2 endothelial-like, MBA-13 fibroendothelial, 13F1.1 cloned preadipocyte, MBA-15 osteoblastic) possess some osteoblastic features but differ in their levels of expression [39, 40]. Thies et al. found that recombinant human bone morphogenic protein-2 induces osteoblastic differentiation in the W-20-17 murine stromal cell line [42]. Additionally, ectopic marrow transplantation experiments demonstrate that newly formed bone marrow stroma and bone are derived from the donor, while blood cells are of host origin, thus supporting the possibility of a common precursor [44-46]. However, the data do not directly address whether osteoblasts support hematopoiesis.

Osteoblasts Synthesize Cytokine-Like Molecules Which Stimulate Hematopoietic Cell Proliferation

In order to prove that osteoblasts affect the development of hematopoietic cells, it is important to catalog the spectrum of hematopoietic growth-promoting cytokines elaborated by osteoblasts. In this context, primary murine osteoblasts have been shown to produce G-CSF [47], M-CSF [48, 49], GM-CSF [48, 49], interleukin 1 (IL-1) [50], and IL-6 [50-52], while transformed murine osteoblasts produce G-CSF [53], M-CSF [39, 48, 49], G-CSF [39, 54], M-CSF [48, 50], IL-1 [50] and IL-6 [39, 51]. Resting primary murine osteoblasts produce relatively low levels of some of these proteins, but production can be enhanced following stimulation with IL-1, tumor necrosis factor (TNF), and lipopolysaccharide [48-54]. By contrast, the rat ROS 17/2.8 osteoblast-like osteosarcoma cell line constitutively produces G-CSF [54]. Primary human osteoblast-like cells (HOBs) are not as well characterized, but ongoing reports indicate that they express RNA messages for G-CSF [9], GM-CSF [9], IL-1α [56], IL-1β [56-58], IL-6 [9, 56-59], transforming growth factor-beta (TGF-β) [55-57], and TNF-α [9, 58], but not IL-3 [9], IL-4 [57], or IL-8 [57]. At the protein level, human osteoblasts produce G-CSF [9], GM-CSF [9], IL-1β [60], IL-6 [26, 60], leukemia inhibitory factor (LIF) [61-63], TNF-α [64], and vascular endothelial growth factor [65]. While this list is not complete, it is important to keep in mind that none of these cytokines, alone or in combination, are likely to fully account for the hematopoietic promoting activities of osteoblasts [9], (Taichman et al., submitted for publication).

Production of Hematopoietic Inhibitory Factors by Osteoblasts

In addition to generating positive growth signals, osteoblasts may conceivably limit hematopoietic cell replication. In fact, these positive and negative activities may not be mutually exclusive. For example, positive signals produced by osteoblasts may ensure the survival of early hematopoietic cells. They may not, however, be sufficient in either quantity or quality to induce hematopoietic cell proliferation in the presence of negative regulatory signals produced locally by osteoblasts. Thus, by producing both inhibitory and competence/progression factors, osteoblasts might maintain hematopoietic stem cells as stem cells.

As to specific hematopoietic inhibitory molecules produced by osteoblasts, little is known. Most osteoblasts basally produce TGF-β1, LIF, and, to varying degrees, TNF-α and TNF-β (lymphotoxin), but fail to produce macrophage inhibitory protein-1α unless stimulated (Taichman et al., submitted) [61-64, 66]. Whether these activities impact hematopoiesis is also not known. Several reports suggest that osteoblasts inhibit hematopoiesis, based upon the observations that osteosarcomas fail to support hematopoietic progenitor cell colony formation in methylcellulose assays [39, 67]. While human hematopoiesis is certainly different from that of the mouse, perhaps the data should be interpreted differently. Perhaps the role of osteoblasts is to limit hematopoietic cell proliferation. This could be accomplished by the combined production of both cell-cycle inhibitory cytokines and competence/progression factors. As such, the ability of osteoblasts to maintain hematopoietic stem/progenitor cells as “stem cells” would indicate a substantial role for osteoblasts in hematopoiesis.

Adhesion of CD34+ Bone Marrow Cells to BMSCs

In the bone marrow, hematopoietic stem cells are closely approximated with the endosteal surfaces rather than randomly distributed throughout the marrow cavity [1, 68]. In vitro long-term bone marrow cultures (LTBMCs) also appear to require an immediate approximation to bone marrow cells. If the two tissues are separated by more than a few millimeters, a precipitous decline in stem cell populations ensues [11, 69]. Moreover, in LTBMCs, discrete stromal elements seem to support specific hematopoietic lineages. For instance, clones of lymphocytes and granulocytes as well as other hematopoietic cell populations occupy discrete cellular “niches” [10, 11, 70]. Together, these two observations strongly suggest that cell-cell adhesion plays an important role in hematopoiesis [71].

Multiple receptor-ligand adhesion molecules (CAMs), including the cadherins, immunoglobulins, integrins, and selectins, mediate blood cell adhesion to BMSCs. Of these, β1 integrins expressed on the CD34+ cells and VCAM-1 expressed by bone marrow stromal cells have received the most attention [72-76]. However, other heterologous interactions are also possible. Most notably, CD34+ cells express very late activation antigen-4 (VLA-4), VLA-5, and leukocyte function-associated antigen-1 (LFA-1) receptors. ICAM-1, ICAM-3, CD44, LFA-3, and PECAM-1 are also constitutively expressed by these cells [77, 78]. Receptor expression and density may also vary according to the maturational status of the CD34+ cell and may play a major role in the release of these cells into the circulation. By example, colony-forming CD34+ cells are detected in the α4 and the α5 marrow fraction (α4β1 and α5β1) whereas during myeloid differentiation, α5β1 is lost at the myelocytic-metamyelocytic stage, before the loss of α4β1 at the band stage [79, 80]. In addition, CD34+ cell ligand affinity varies with maturation and with engagement of other cell surface receptors. Here, VLA-4-mediated adhesion of CD34+ cells to VCAM-1 is enhanced by antibodies to the PECAM-1 receptor [77, 81, 82]. Stromal cell-derived extracellular matrix molecules (ECMs) including glycosaminoglycans (heparan and heparan sulfate), thrombospondin, fibronectin, and hemonectin (and their respective receptors, where known) also mediate stem cell adhesion [72-74, 78]. Adding yet another level of complexity, growth factors and their receptors may also position stem cells to their microenvironment. Membrane-bound growth factor/cytokines include: IL-1α, M-CSF, c-kit ligand [83-85], and ECM-bound factors; IL-3, GM-CSF, and TGF-β1 [86-88], which themselves may serve as CAMs. Thus, while the population of CD34+ cells in the marrow is limited, based upon CD34 and CAM receptor density and/or affinity, many options exist to localize CD34+ cells to a particular microenvironment [73, 89-94]. In this regard, marked heterogeneity in the adherence of CD34+ cells has been observed in patients with myeloproliferative disorders [81, 95-97]. All of these examples serve to illustrate that adhesion of blood cells to BMSCs may be crucial in many clinical settings, including those relating to the regeneration of bone and marrow [14, 95, 97, 98].

Adhesion of CD34+ Bone Marrow Cells to Osteoblasts

As stated earlier, the presence of hematopoietic stem cells near endosteal surfaces may reflect a requirement for osteoblast-derived products critical for hematopoietic stem cell survival and self-renewal. If this is the case, then many questions remain. For example, what type of cell adhesion molecule(s) mediates hematopoietic cell adhesion to osteoblasts? Does adhesion to osteoblasts vary with osteoblast maturation? How are these adhesive interactions regulated to facilitate stem cell exit from the bone marrow, such as during peripheral blood mobilization procedures? Lastly, are these associations altered during development or in inflammatory or neoplastic states?

Much information is available on the identity of those receptors utilized by osteoblasts in their adherence to extracellular matrix proteins (i.e., collagen, fibronectin, proteoglycans, osteonectin, osteopontin, vitronectin, laminin, and bone sialoproteins) [10, 82, 99-107]. Many of these molecules could potentially be utilized during blood cell/osteoblast adhesions. Those most relevant to the present investigations relate to osteoblast/osteoclast adhesions. Here LFA-1 and ICAM-1 [82], αvβ3 (classical vitronectin receptor), α2β1 (collagen/laminin) and αvβ1 (vitronectin receptors) expression have received the most attention [104]. Osteoclasts also express β1 and α2, α5, αv and αvβ3 integrins (but may not constitutively express α46LM and β2) [101, 102]. Although there are little direct data on receptor interactions between osteoblasts and osteoclasts, many of these molecules might be involved. Those interactions known to occur between osteoblasts and osteoclasts involve VCAM-1 [103]. One clue as to how these interactions might occur comes from data that illustrate that by cross-linking VCAM-1 and LFA-3 receptors on osteoblasts (with antibodies or T-cells), IL-6 secretion is increased. This is very similar to our recent findings that in the presence of CD34+ cells, osteoblasts produce elevated IL-6 levels [26, 100, 108]. It may also be important to consider that cell-associated ECM molecules (i.e., collagen, fibronectin, osteocalcin, osteopontin) may facilitate these interactions [28, 109-111].

Conclusions and Future Directions

In spite of the significant voids in our knowledge, there are good reasons to suspect that osteoblast-derived factors play a central role in hematopoietic development in vivo. In the marrow, osteoblasts are in a biologically relevant site to transmit information to the developing hematopoietic lineages. Moreover, osteoblasts produce factors that influence blood cell development, particularly towards the granulocytic lineages in vitro [9]. Further identifying the function of osteoblasts with regard to hematopoiesis, whether restrictive and/or stimulatory, will undoubtedly yield significant insights into the functional relationships of this complex issue. These findings may be useful to re-engineer the marrow organ in patients with myeloproliferative disorders. Clearly, further investigations are needed.


This review was supported in part by National Institutes of Health Grants R29-DE11283. Dr. Emerson is supported by a Scholar Award from the Leukemia Society of America. The authors are indebted to M.J. Reilly, R.S. Verma, L.B. and N.S. Taichman, L.K. McCauley, P.H. Krebsbach, R. Franceschi, and M. Somerman for helpful discussions.