The hematopoietic stem cell niche

2.1. Osteoblasts

In adult humans, steady-state hematopoiesis occurs in the bone marrow, which is situated within the bone cavity. Rather than being an inert structure, bone tissue undergoes a constant process of remodeling via a tight coupling between bone formation from osteoblasts (derived from mesenchymal stem cells) and bone resorption by osteoclasts (which are hematopoietic in origin; Martin and Sims, 2005). A number of different cell types of the osteoblast lineage are present in bone and bone marrow including (from most primitive to most mature) mesenchymal stem cells, osteoprogenitor cells, osteoblasts and osteocytes (Mackie, 2003).

Osteoblasts are usually found in a layer along the endosteum at the interface between bone and marrow and periosteum, which comprise the internal and external surfaces of bone, respectively. The major functions of osteoblasts in bone remodeling are the secretion of unmineralized bone matrix proteins (collectively termed osteoid) and cells of the osteoblast lineage regulate osteoclast differentiation (Lian, 2003; Martin and Sims, 2005).

Two complementary and simultaneously published studies identified that osteoblasts play an active role as part of the regulating microenvironment or niche for HSCs (Calvi et al., 2003; Zhang et al., 2003). In one report, the bone morphogenetic protein receptor 1A (BMPR1A) was conditionally deleted from mice using Mx1-Cre (Zhang et al., 2003). This resulted in ectopic formation of a trabecular bone-like area (TBLA), and a significant increase in the number of N-cadherin-positive osteoblasts (SNO cells) in the TBLA (Zhang et al., 2003). The increased TBLA correlated with increased numbers of HSCs (measured by functional as well as phenotypical properties), and label-retaining cells thought to be HSC were found to be attached to the SNO cells in association with N-cadherin (Zhang et al., 2003). It should be noted however, that the promoter driving the Cre recombinase and therefore the excision of BMPR1a was not specific for osteoblasts therefore other cell types might also have been involved.

In contrast, Calvi et al. examined the effects of the constitutively activated parathyroid hormone/parathyroid hormone-related peptide receptor (PPR) under the control of the osteoblast-specific 2.3kB α1(1) collagen promoter (Calvi et al., 2003). The PPR transgenic mice had significantly increased HSCs in conjunction with increased trabecular bone and elevated numbers of trabecular osteoblasts which expressed the Notch ligand Jagged 1. HSC-enriched cells from PPR transgenic mice had more activated Notch 1 than wildtype HSCs. Moreover, contact between the HSCs and Jagged 1-expressing osteoblasts were required for the HSC-potentiating effects of the osteoblasts, and the addition of a γ secretase inhibitor (which inhibits Notch activation) to stromal cell cultures prevented this effect (Calvi et al., 2003). Finally, treatment of wildtype stromal cultures or wildtype mice with parathyroid hormone (PTH) recapitulated the phenotypes observed in PPR transgenic mice (Calvi et al., 2003). PTH has also been shown to have therapeutic potential for both HSC mobilization and hematopoietic recovery post-transplantation via regulation of the size of the osteoblast HSC niche in mouse models (Adams et al., 2007; Calvi et al., 2003).

It should be noted that these in vivo models were initiated largely because of prior in vitro work in which it was noted that osteoblasts or osteoblastic cell lines could support hematopoietic cells in culture. Human osteoblasts have been shown to support HSCs in ex vivo culture systems (Taichman and Emerson, 1994; Taichman and Emerson, 1998; Taichman et al., 1996; Taichman et al., 2000). While studies of osteoblast expansion have demonstrated an increase in HSCs, studies of osteoblast decrease have not consistently shown a reduction in HSCs. Visnjic et al. showed that deletion of osteoblasts by gancyclovir in mice transgenic for thymidine kinase expressed in osteoblasts caused reduced marrow and increased extrameduallary hematopoiesis (Visnjic et al., 2004). This was accompanied by a reduction in absolute numbers of primitive hematopoietic cells in the bone marrow and increased numbers in spleen and liver (Visnjic et al., 2004). However, osteoblast dysfunction in the biglycan deficient mouse did not result in a change in marrow HSC (Kiel et al., 2007). Therefore, either specific subsets of osteoblasts contribute to niche or osteoblasts are not required for niche function and may be supplanted by other cell types under conditions of more gradual decline in osteoblast number. Products of osteoblasts that have been shown to be positive regulators of HSCs include angiopoietin-1, thrombopoietin and Jagged-1 (under conditions of PPR activation; Arai et al., 2004; Calvi et al., 2003; Wilson et al., 2004; Yoshihara et al., 2007), whereas osteoblast-associated negative regulators of the HSC niche include osteopontin and dikkopf1 (Fleming et al., 2008; Nilsson et al., 2005; Stier et al., 2005; reviewed in detail in (Adams and Scadden, 2006; Kiel and Morrison, 2008; Yin and Li, 2006)). While Jagged-1 appears to be critical for PPR activated osteoblast increase in HSC, it does not participate in HSC homeostasis (Mancini et al., 2005). Osteoblasts also abundantly express CXCL12, a chemokine that has been implicated to have roles in chemotaxis, homing, survival of HSCs and in the retention of HSCs in the bone marrow (Broxmeyer et al., 2005). It should be noted, however, that other non-osteoblast “reticular” cells in the marrow also express high levels of CXCL12 and may participate in the niche (Papayannopoulou and Scadden, 2008; Sugiyama et al., 2006). While osteoblasts are contributors to HSC regulation in the marrow, it remains unclear whether direct contact with HSCs is required. What is clear is that other cell types besides the osteoblast participate in HSC regulation in the marrow.

2.2. Endothelial cells

Endothelial cells line all blood vessels in the body. In the bone marrow they form a barrier between the developing hematopoietic cells and the blood. They are therefore the initial site of entrance of all blood cells into the bone marrow from the circulation and also the final place whereby blood cells leave the bone marrow to enter the bloodstream (Sipkins et al., 2005; Winkler and Levesque, 2006). The possibility of a perivascular zone serving as a regulatory niche for stem cells is derived from two studies. In vivo imaging of primitive hematopoietic cells in animals over time revealed that they localized to specific subsections of the marrow microvasculature where cells persisted or increased in number over a 70 day interval (Sipkins et al., 2005). These data suggested the perivascular site as a potential niche, but the hematopoietic populations studied were not highly enriched for HSCs.

It was the discovery of SLAM antigens marking HSCs that enabled histologic assessment of where the HSCs resided in the marrow microenvironment (Kiel et al., 2005). These studies indicated that the majority of HSCs were in the perivascular region with only a minority (∼16%) at the periendosteal region. These data are consistent with the perivascular regions serving as a niche, but several caveats must be kept in mind. The most important is that HSCs traffic into and out of the vasculature, so it remains formally possible that the cells accumulate around vessels because it is an impedance point in their trafficking. As such, this site would not meet the criteria for a niche (a site where self-renewal occurs). Also, in the sponge-like trabecular region where most HSCs reside, it is difficult to discern all anatomic relationships from two-dimensional images. Finally, to date no modification in endothelial function in vivo has been shown to affect HSCs other than what might be expected from peturbed trafficking. Therefore, whether the perivascular zone represents a true niche in the marrow still requires experimental definition

However, developmental changes in hematopoiesis would suggest that perivascular sites are likely to serve as niches and there are in vitro studies indicating that endothelial cells derived from various tissues can support HSCs in culture systems (Chute et al., 2002; Li et al., 2004; Rafii et al., 1995). In embryonic development, HSCs originate in the yolk sac and from hemogenic endothelium in the dorsal aorta in the aorta, gonad, mesonephros (AGM) region. The placenta may also serve as a site of HSC production with the fetal liver becoming the site where HSCs reside, expand and produce blood during the second trimester of gestation. Hematopoiesis establishes in bone marrow only later in the second trimester of embryonic development (Cumano and Godin, 2007; Orkin and Zon, 2008; Orkin and Zon, 2008). With the exception of the bone marrow, all of the sites where self-renewing HSCs can be isolated during embryonic development are devoid of osteoblasts, but contain endothelial cells, which have been closely associated with the generation of the HSCs (Cumano and Godin, 2007; Orkin and Zon, 2008; Orkin and Zon, 2008).

3.1. One or more HSC niches?

The localization studies of HSCs in their niche in vivo have predominantly relied on visualization of cells in bone marrow sections (Arai et al., 2004; Kiel et al., 2005; Zhang et al., 2003). These two-dimensional sections are not true representations of the three-dimensional structure of the bone marrow microenvironment. Such studies may therefore miss the correct anatomical structure of the HSC niche which may consist of a variety of different cell types rather than being restricted to one cell type. It is highly likely, for example, that the osteoblast-containing HSC niche is in close connection with endothelial cells. Evidence to support this comes from studies demonstrating that the bone marrow (including the endosteal surface) is highly vascularized (de Saint-Georges and Miller, 1992; Narayan et al., 1994). Indeed, vascularization of the cartilage template is a crucial process during the formation of embryonic bone by endochondral ossification, and impaired angiogenesis during this process results in significantly reduced bone formation (Brandi and Collin-Osdoby, 2006; Maes et al., 2002; Maes et al., 2007). Furthermore, the vasculature is highly important in the ongoing process of bone remodeling (Brandi and Collin-Osdoby, 2006). The osteoblast-containing HSC niche is therefore likely also in close proximity to endothelial cells, and these two cell types potentially collectively form the trabecular HSC niche.

There are, however, distinct regions of the bone marrow in which the endothelial cells are not located near osteoblasts, particularly in areas devoid of trabecular bone in the central region of the diaphysis (de Saint-Georges and Miller, 1992; and Dr Natalie Sims, personal communication). It remains to be determined if these endothelial cells also co-localize with HSCs or if they are different to those identified as the perivascular HSC niche. Some recent studies, however, have proposed the existence of two distinct HSC niches that support quiescent or activated HSCs.

3.2. Quiescent versus activated HSC niches

Some studies have noted that the osteoblast-containing niche represents a niche in which HSCs remain quiescent. To date, at least three molecules found on the surface of osteoblasts (N-cadherin, angiopoietin-1 and thrombopoietin) have been shown to regulate HSC quiescence via interaction with their receptors (N-cadherin, Tie-2 or Mpl, respectively), expressed on HSCs (Arai et al., 2004; Haug et al., 2008; Wilson et al., 2004; Wilson et al., 2007; Yoshihara et al., 2007; Zhang et al., 2003). Moreover, osteoblast-specific overexpression of the canonical wingless (Wnt) inhibitor, Dikkopf1 (Dkk1) resulted in loss of HSC quiescence accompanied by reduced serial transplant potential of HSCs (Fleming et al., 2008). These data suggest that Wnt signaling in the niche is important for sustaining HSC quiescence and self-renewal.

In contrast, the perivascular niche has been proposed to represent a niche in which the HSCs are in a more activated state. Such activated HSCs include cycling HSCs or HSCs that have been subjected to stress, such as by treatment with chemotherapeutic drugs (for example, 5-fluorouracil or cyclophosphamide) or cytokines (such as G-CSF; Morrison et al., 1997; Randall and Weissman, 1997; Wilson et al., 2007; Zhang and Li, 2008). It is currently not clear if such a niche may be a transient niche for HSCs, if it represents a stable HSC niche for a particular population of HSCs or if the HSCs at the perivascular niche were merely in transit when they were visualized.

3.3. Different niches for distinct hierarchies of HSCs?

It is now well-recognised that HSCs are not a homogeneous population of stem cells, but are hierarchically organized based on their functional potential, with the most immature HSC capable of sustaining hematopoiesis through serial transplantation (Purton and Scadden, 2007). Some recent studies comparing properties of the HSC populations initially used to identify the prospective HSC niches suggest they may represent different populations of HSCs. Though these studies are controversial, studies identifying the importance of the osteoblast in the HSC niche have commonly utilized the lineage negative, c-Kit-positive, Sca-1-positive (LKS+) population of HSCs (Ikuta and Weissman, 1992; Okada et al., 1992), which is a heterogeneous population of short-term repopulating HSCs, long-term repopulating HSCs and multipotent progenitor cells (Adolfsson et al., 2001; Osawa et al., 1996; Yang et al., 2005). However, further purification of such HSCs was also performed in these studies using the Hoescht 33342 “side population” protocol (Arai et al., 2004; Goodell et al., 1996) or BrdU label retaining population, which has been shown to select for the slowly cycling “quiescent” HSCs (Bradford et al., 1997; Cheshier et al., 1999; Zhang et al., 2003).

One study reported that more than 70% of HSCs identified based on the SLAM markers (CD48-CD150+LKS+) express CD34 (Wilson et al., 2007), a marker that is normally associated with activated (non-quiescent) or short-term repopulating HSCs (Ogawa et al., 2001; Yang et al., 2005). The majority of HSCs identified using SLAM markers are not BrdU label-retaining cells (Kiel and Morrison, 2006) and SLAM-associated HSCs do not express N-cadherin (Haug et al., 2008; Kiel et al., 2007), an adhesion molecule that some have reported maintains quiescent HSCs in the osteoblast niche (Haug et al., 2008; Wilson et al., 2004; Zhang et al., 2003).

Interestingly, during embryonic development HSCs (which are closely associated with vasculature) are actively cycling cells that express CD34 (Ito et al., 2000; Matsuoka et al., 2001; Sanchez et al., 1996). Likewise, mobilized murine HSCs, which are perivascular (Yilmaz et al., 2006), express CD34 (Tajima et al., 2000), whereas in the steady-state condition murine HSCs are CD34 low/negative (Osawa et al., 1996). Hence it is currently unclear if the HSC populations used to initially identify the HSC niches represent two distinct populations of HSCs or if they have the same functional potential. It will be experimentally difficult, but ultimately important to test populations from distinct locations using stringent tests of HSC self-renewal, such as the serial transplant assay (Purton and Scadden, 2007) to determine if there are functional differences in distinctly localized cell populations.

3.4. Do all osteoblasts or endothelial cells support HSCs?

Another current unknown is whether all osteoblasts and/or endothelial cells have the same HSC support potential, or if specialized osteoblasts/endothelial cells or specific locations of osteoblasts/endothelial cells have different roles in regulating HSCs. For example, the trabecular osteoblasts were initially identified as being part of the HSC niche (Calvi et al., 2003; Zhang et al., 2003), however, we commonly refer to the HSC niche as being the endosteal surface (which applies to both cortical and trabecular bone). Of note, cortical and trabecular bone differ in their anatomical location, structure and function. Cortical (compact) bone is located in the diaphyseal region of bone, is thick and dense (80–90% of the bone is calcified) and considered to have predominantly mechanical and protective functions. In contrast, trabecular (cancellous) bone is found in the metaphyseal region of bone, is less dense (15–25% is calcified) and has more of a metabolic function (Baron, 2003). It still remains to be shown whether or not osteoblasts in different anatomical locations have distinct roles with respect to regulating HSCs.

Likewise, endothelial cells are specialized cells that have different functions in distinct tissues and organs (Belloni and Tressler, 1990). There are also differences in properties and functions of endothelial cells derived from different sized blood vessels, eg arteries or capillaries (Belloni and Tressler, 1990). Moreover, HSCs have been shown to home to specific CXCL12-expressing microdomains in the bone marrow vasculature, suggesting specific regions of endothelial cells may function as the perivascular HSC niche (Sipkins et al., 2005).

4.1. Do other cell types also play important roles as components or regulators of the HSC niche?

It is highly likely that the HSC niche is more complex than only one cell type. Various studies have revealed that the sympathetic nervous system (SNS) also plays an important role in regulating the HSC niche (Afan et al., 1997; Katayama et al., 2006; Mendez-Ferrer et al., 2008). Anatomical studies have revealed that bone marrow is highly innervated with both myelinated and nonmyelinated nerve fibers (Calvo and Forteza-Vila, 1969; Calvo and Forteza-Vila, 1970; Yamazaki and Allen, 1990), most of which are located near arterioles in the bone marrow (Yamazaki and Allen, 1990). Surgical severance of the femoral nerve resulted in reduced bone marrow cellularity accompanied by significant mobilization of progenitor cells (Afan et al., 1997). In addition, treatment of mice with the neurotoxin 6-hydroxydopamine (which resulted in blockade of neurotransmitter synthesis in adrenergic and dopaminergic nerve fibers) resulted in decreased bone marrow cellularity (Afan et al., 1997).

More recently, the SNS neurotransmittor norepinephrine (NE) was shown to control both the suppression of osteoblasts and downregulation of CXCL12 expression in bone cells that occurs in response to G-CSF treatment (Katayama et al., 2006). When the SNS was disrupted (such as in UDP-galactose ceramide galactosyltransferase-deficient mice, which display defects in nerve conduction due to a lack of myelin), mobilization of HSCs from their niche did not occur. In contrast, when a β2-adrenergic agonist was administered to both control and NE-deficient mice, HSC mobilization was enhanced (Katayama et al., 2006). Additional studies in the mouse model revealed that the cyclical release of HSCs and expression of CXCL12 in the bone marrow microenvironment was regulated by circadian NE secretion by the SNS (Mendez-Ferrer et al., 2008). Neurotransmittors were also recently shown to have roles in human HSC mobilization, proliferation and differentiation (Spiegel et al., 2007).

Other cell types that have potential roles in regulating the HSC niche include chondrocytes (Jacenko et al., 2002), adipocytes (Walkley et al., 2007) and CXCL12-abundant reticular cells (Sugiyama et al., 2006). Further studies will be required to fully elucidate how each of these cell types contributes to the HSC niche.

Hematopoietic cells may also play a role in regulating the HSC niches. For example, monocytes have been shown to express osteopontin, which significantly decreased Notch1 signaling in HSCs that were co-cultured with a supportive bone marrow stromal cell line that expresses Jagged-1 (Iwata et al., 2004; Li et al., 1998). In addition, the monocyte-derived osteoclasts play an important role not only in the regulation of osteoblast activity and hence potentially osteoblast HSC niche size (Martin and Sims, 2005), but elevated Ca²⁺ concentration released during osteoclast-mediated bone resorption may be important in directing HSCs (which express the calcium-sensing receptor) to the osteoblast-containing HSC niche (Adams et al., 2006). Indeed osteoclasts have been suggested to induce HSC mobilization from the bone marrow microenvironment (Kollet et al., 2006).

4.2. Do the HSC niche components influence each other?

It is currently unknown whether there is cross-talk between the osteoblast and endothelial cells in HSC niches, although there is evidence that endothelial cells and osteoblasts can regulate each other. Angiopoietin-1 (which is expressed by the osteoblast lineage), interacts with Tie-2 expressed by endothelial cells to increase angiogenesis (the formation of new blood vessels) and reduce vascular permeability (Fukuhara et al., 2008; Saharinen et al., 2008). Likewise, osteoblasts secrete vascular endothelial growth factor (VEGF), which also modulates vascularization and permeability of endothelial cells (Ferrara and Davis-Smyth, 1997), and has a central role in bone morphogenesis (Zelzer and Olsen, 2005). Other known interactions between endothelial cells and osteoblasts have been reviewed in detail by Kanczler and Oreffo (Kanczler and Oreffo, 2008).

There have also been a series of reports describing that the perivascular niche is also the niche for mesenchymal stem cells (MSCs), the precursors of the osteoblast (Sacchetti et al., 2007; Shi and Gronthos, 2003). It is currently unclear if the MSC perivascular niche has the same anatomical location and function as the HSC perivascular niche. It may be speculated that, in humans MSCs may be identical to the CXCL12-abundant reticular cells that are also associated with both the perivascular and osteoblast HSC niches (Sacchetti et al., 2007; Sugiyama et al., 2006). Determining whether MSC or the MSC niche participate in altering the HSC niche is an area of great interest that is expected to receive considerable attention in the future.