The bone marrow microenvironment in health and disease at a glance

The adult hematopoietic system, predominantly located in the bone marrow, is the most widely studied adult stem cell system (see poster). Our understanding of this system has changed immensely during the past two decades with increasing evidence, suggesting interdependency of hematopoietic stem cells (HSCs) and their surrounding bone marrow microenvironment (BMM). The BMM itself is an anatomical location within the bone that contains many different cell types such as osteolineage cells, mesenchymal stem cells, arteriolar and sinusoidal endothelial cells and neurons, and is influenced by cytokines and factors, such as oxygen tension and shear forces, all of which govern the fate of HSCs (Krause and Scadden, 2015) (see poster). Leukemia cells or leukemia stem cells (LSCs) are influenced by the BMM (Krause et al., 2013), although in a different manner to HSCs. LSCs can modulate the BMM (Frisch et al., 2012; Schepers et al., 2013; Welner et al., 2015), leading to the establishment of a bone marrow niche that is most beneficial for the maintenance of LSCs, thereby ‘sheltering’ them from various therapies. Hence, this niche also constitutes a site of origin for leukemic relapse and progression (Ishikawa et al., 2007). Plasma cell myeloma is another hematological malignancy that is characterized by highly specific and very distinct interactions with the BMM (Kumar et al., 2017) (Box 1). Owing to these complexities, a comprehensive understanding of the interactions between HSCs and the BMM and how these interactions are altered in pathological states can aid the design of therapeutic strategies (Krause and Scadden, 2015; Passegue et al., 2003). This Cell Science at a Glance article and the accompanying poster provide a brief summary of the current knowledge on the BMM and its associated elements, as well as of the molecular networks that regulate or influence hematopoiesis in health and disease. We will also focus on discussing the current and future strategies to target these networks in an effort to augment existing treatment strategies (see poster).

Bone marrow, the spongy tissue inside bones, represents the production site of the various multilineage cellular constituents of our blood, the ‘progeny’ of HSCs. In health, the hematopoietic system, with HSCs at the top of the hierarchy, works throughout the lifespan of an individual, thereby providing a constant output of oxygen-carrying red blood cells, as well as white blood cells and platelets, to maintain homeostasis (see poster). The concept of a hematopoietic hierarchy was widely held for many years, but, recently, this view has been challenged. An array of studies has shown that lineage commitment might not be as strict as previously thought (Naik et al., 2013; Notta et al., 2016; Yamamoto et al., 2013). Furthermore, a recent study has shown that the fate of stem cells is actually governed at the molecular level of a single cell (Velten et al., 2017).

In adult animals, hematopoietic stem and progenitor cells (HSPCs) reside and function inside the bone marrow, with their location inside the niche appearing to depend on their maturation state (Lo Celso et al., 2008) and/or their activity (Morrison and Scadden, 2014). Controversies exist about the nature and function of the endosteal (Adams and Scadden, 2006; Calvi et al., 2003; Zhang et al., 2003) and vascular niches (Ding and Morrison, 2013; Ding et al., 2012; Kiel et al., 2005; Pitt et al., 2015; Wang et al., 2007) in the bone marrow, and how exactly the location of HSPCs in these niches may correlate with their function remains to be elucidated. However, emerging evidence suggests that HSCs exist close to arterioles and that hematopoietic ‘stress’ leads to HSC proliferation and their distribution away from arterioles (Kiel et al., 2005; Koechlein et al., 2016; Kunisaki et al., 2013) (see poster). Using two-photon microscopy in the calvarium of live mice, it was shown that the oxygen tension differed in different regions of the bone marrow cavity with the endosteal niche being less hypoxic than the deep perisinusoidal regions (Spencer et al., 2014). Low permeability of arterioles and low concomitant reactive oxygen species (ROS) in their vicinity appear to result in the maintenance of HSC quiescence, whereas the high permeability of sinusoids and a high local concentration of ROS leads to differentiation and migration of HSCs (Itkin et al., 2016; Ludin et al., 2014). Similarly, increased nitric oxide (NO) in the bone marrow vasculature has been linked to HSC egress, while lower levels of NO mediate decreased motility and increased adhesion of HSC via the endothelial protein C receptor (EPCR) (Gur-Cohen et al., 2015).

Mesenchymal stromal cells (MSCs) are a heterogeneous self-renewing population of cells defined by a set of different markers, such as nestin (Méndez-Ferrer et al., 2010), neural-glial antigen (NG)-2 (Kunisaki et al., 2013), leptin receptor (Ding et al., 2012) or paired related homeobox (Prx-1) (Ding and Morrison, 2013) and are enriched close to the vasculature in the BMM. MSCs give rise to different lineages, including osteoblastic cells, chondrocytes and adipocytes (Kfoury and Scadden, 2015). Different progeny of MSCs associate with HSCs and, in general, most of them are known to secrete HSC-supporting factors, such as C-X-C motif chemokine ligand 12 (CXCL12), angiopoietin, stem cell factor (SCF/Kit ligand) and others, but differences according to MSC type and location, i.e. arteriolar or sinusoidal, have been revealed (Asada et al., 2017). Finally, endothelial cells themselves have been shown to support HSC maintenance by providing factors, such as CXCL12, SCF, angiopoietin, fibroblast growth factor (FGF) 2, and Delta-like 1 (Crane et al., 2017; Morrison and Scadden, 2014). Furthermore, deletion of E-selectin from endothelial cells increased HSC quiescence and self-renewal, confirming that E-selectin also supports HSC function (Winkler et al., 2012).

The endosteum lies at the interface between bone and bone marrow and is predominantly composed of osteoblastic cells, which can also regulate the number and function of HSCs (Adams et al., 2007; Bowers et al., 2015; Calvi et al., 2003; Taichman et al., 1996; Zhang et al., 2003). Furthermore, bone-degrading osteoclasts (Kollet et al., 2006; Lymperi et al., 2011; Yokota et al., 2010) and osteocytes (Asada et al., 2011; Fulzele et al., 2013) also affect hematopoiesis (see poster).

However, the existence and potential contributions of the osteoblastic niche to HSC maintenance has been viewed critically, partly as a result of the heterogeneity amongst the osteoblastic cell population and their varying differentiation status (Nakamura et al., 2010). Recently, the complexity within osteolineage cells and their influence on HSCs was elucidated by the use of a proximity-based, differential, single-cell approach uncovering the molecular heterogeneity among osteolineage cells. Thereby, the RNase angiogenin, interleukin (IL)-18 and embigin were identified as regulators of HSPC quiescence (Silberstein et al., 2016).

With accumulating data supporting the involvement of the BMM in promoting normal hematopoiesis, attention also focused on its role in abnormal or malignant hematopoiesis. The first report demonstrated that mice deficient for retinoic acid receptor (RAR)-γ had a myeloproliferative phenotype, which was present even after the transplantation of wild-type bone marrow into an RAR-γ-deficient BMM (Walkley et al., 2007a). In addition, loss of the retinoblastoma gene in both niche and hematopoietic cells (Walkley et al., 2007b) led to a myeloproliferation. Knockdown of the mRNA-processing enzyme Dicer1 in osteoprogenitors led to a myelodysplasia-like syndrome, with a few mice developing acute myeloid leukemia (AML) (Raaijmakers et al., 2010). Furthermore, a mutation in the canonical Wnt pathway that leads to constitutively active β-catenin in osteoblasts resulted in acute myeloid leukemia, with increased Notch and Wnt activation being found in 38% of patients with MDS or AML (Kode et al., 2014). More recently, an MSC- and osteoprogenitor-cell-specific activating mutation in Ptpn11, a gene that encodes the tyrosine phosphatase SHP2 and is found in 50% of patients with Noonan Syndrome, led to a predisposition to myeloproliferative neoplasia in children and adolescents, resulted in increased levels of chemokine (C-C motif) ligand 3 (CCL3) and other pro-inflammatory cytokines, leading to leukemogenesis. (Dong et al., 2016) (see poster).

Identification of oncogenes and the underlying molecular events in any cancer has been a long-term goal in research, as in many, but not all cases, these pathways have been proven to be worthy targets for therapy. However, for hematopoietic cancers, the focus of current research efforts has shifted towards understanding the specific interactions between malignant hematopoietic cells and the BMM, which may be exploited as novel adjuvant treatments along with available therapeutic regimens.

CD44 is a cell-surface glycoprotein that interacts with components of the extracellular matrix, such as hyaluronan, osteopontin and E-selectin (Ponta et al., 2003). In murine chronic myeloid leukemia (CML), higher CD44 expression on leukemia cells correlated with a requirement of CD44 for their efficient homing and engraftment, but its role in B-cell acute lymphoblastic leukemia was dispensable (Krause et al., 2006). Similarly, in acute myeloid leukemia (AML), CD44 was found to be an essential mediator of interactions with the niche and required for homing of leukemic stem cell-supportive niches (Jin et al., 2006). E-selectin, which is expressed on endothelium and known to mediate the engraftment of leukemia-initiating cells in CML (Krause et al., 2014), is upregulated five- to ten-fold in AML and mediates chemoresistance in the vascular niche (Winkler et al., 2014). Specific endothelial locations that are particularly enriched in E-selectin and stromal-derived factor (SDF)-1 were shown to be sites of engraftment for a B-acute lymphoblastic leukemia (B-ALL) cell line (Sipkins et al., 2005). Integrins belong to the large family of cell-adhesion molecules, which under homeostatic conditions, regulate HSC homing and function. β1 integrins are involved in adhesion of leukemia cells in CML, as for instance, increased adhesion to fibronectin and decreased proliferation of leukemia cells was observed if an activating antibody against α5β1 integrin was used (Lundell et al., 1997), whereas interferon-α restored β1-integrin-mediated adhesion of CML cells (Bhatia et al., 1996). β1 and β2 integrins on endothelial cells are also mediators of chemoresistance in chronic lymphocytic leukaemia (CLL) (Maffei et al., 2011). Moreover, different in vivo studies have identified β3 integrin as an integral part of leukemogenesis in AML, although it is dispensable for normal hematopoiesis (Miller et al., 2013). Integrin β3 was recently shown to enhance Wnt signaling in AML cells by mediating resistance to the kinase inhibitor sorafenib (Yi et al., 2016). In addition, the integrin-binding glycoprotein CD98 was shown to promote AML proliferation and maintenance of leukemic stem cells through an increased engagement with the BMM, while anti-CD98 blockade inhibited AML growth (Bajaj et al., 2016). In a murine model of chronic lymphocytic leukemia (CLL), transplantation of CLL cells into mice deficient for the Lyn kinase or Bruton's tyrosine kinase (BTK) led to a decelerated leukemia progression and prolonged survival, suggesting that Lyn and BTK in the BMM are essential for leukemic progression in CLL. Interestingly, Lyn deficiency in macrophages, which exert a ‘nursing’ function for CLL cells through direct cell–cell contact, were implicated in the observed phenotype (Nguyen et al., 2016) (see poster).

Modification of the BMM by malignant hematopoietic growth is a well-characterized phenomenon that affects the osteoblastic and vascular components of the BMM. However, this appears to be highly specific for oncogenic events in leukemia cells (Jacamo et al., 2017; Krause and Scadden, 2015). Vascular endothelial growth factor (VEGF) and other factors such as angiopoietin 2, which are secreted by leukemia cells, have been implicated in the proliferation of leukemia and endothelial cells in B-ALL (Veiga et al., 2006), AML (Hatfield et al., 2009) and CLL (Maffei et al., 2010). CCL3 (also known as MIP-1α) is secreted by AML cells and leads to a reduction of osteoblastic cells and mineralized bone in the osteolineage compartment of the BMM, thereby inhibiting osteoblastic cell function (Frisch et al., 2012). Another report found that AML cells appear to induce osteogenic differentiation, but inhibit adipocytic differentiation of MSCs. Here, induction of osteogenic differentiation was thought to be due to activation of Smad-1/5 signaling in MSCs by bone morphogenetic proteins (BMPs) derived from AML cells, leading to an increase in pre-osteoblastic cells in the leukemic niche (Battula et al., 2017). In myeloproliferative neoplasia (MPN), the endosteal niche is remodeled into a leukemic niche through the stimulation of mesenchymal stromal cells and the generation of functionally abnormal osteoblastic cells; here, thrombopoietin, CCL3 and direct cell–cell interactions were shown to promote the expansion of osteoblastic cells, and TGF-β, Notch and inflammatory signaling were shown to be involved in niche remodeling (Schepers et al., 2013). Similarly, in T-cell acute lymphoblastic leukemia (T-ALL), rapid remodeling of the endosteal space, but not the perivascular space occurs, leading to a loss of osteoblastic cells and impairment of normal HSCs (Hawkins et al., 2016). Even adipocytes were found to support the growth of AML cells (Shafat et al., 2017).

Tunneling nanotubes (TNTs), a means of communication between leukemia cells and their BMM, and in particular MSCs, have been described in B-ALL. This mode of intercellular signaling leads to the secretion of prosurvival cytokines and chemotactic proteins, as well as the development of resistance to prednisolone, a drug frequently used for the treatment of B-ALL. Accordingly, the blockade of TNT inhibits leukemic progression and prednisolone resistance that is mediated by the stroma (de Rooij et al., 2017; Polak et al., 2015) (see poster). Furthermore, a mitochondrial transfer to leukemic cells (AML cells) but not non-malignant CD34 cells by stromal cells leads to increased chemoresistance through the increased production of mitochondrial adenosine triphosphate and decreased propensity to depolarization of mitochondria after chemotherapy (Moschoi et al., 2016). Additionally, transfer of mitochondria from bone marrow stroma to AML cells has been shown to occur, with this transfer to AML cells being greater than to normal CD34⁺ HSCs. This transfer is mediated through NADPH oxidase 2 (NOX2)-derived superoxide and nanotubes, as inhibition of NOX2 prevented mitochondrial transfer, increased the apoptosis of AML cells and prolonged survival in a xenotransplantation model (Marlein et al., 2017).

Cytokines are well described modulators of leukemic progression. The involvement of various growth factors and their respective receptors has been extensively studied and therapeutically targeted in different leukemias. For example, interleukins are well described modulators of growth and differentiation. The pathway downstream of CXCL12 (also known as SDF-1α) and its receptor has been well studied in the context of its relevance for normal hematopoiesis, as CXCL12–CXCR4 is responsible for the retention of HSCs in the bone marrow and for the maintenance of HSC quiescence by various niche cells (Morrison and Scadden, 2014). Different types of leukemia cells also utilize this axis for engraftment, but, additionally, usurp CXCL12–CXCR4 to gain advantage over normal HSCs. C-X-C chemokine receptor type 2 (CXCR2), the receptor for IL-8, was shown to be overexpressed on AML blasts and considered to be an adverse prognostic factor, whereas inhibition of CXCR2 specifically inhibited the proliferation of AML and MDS cells (Schinke et al., 2015). Axl, a member of the receptor tyrosine kinase family and a prognostic marker in AML, induces the expression of its ligand Gas6 (growth arrest-specific gene 6) from the BMM and increases the proliferation of leukemic cells (Janning et al., 2015). The same axis has been reported to regulate the self-renewal of CML cells (Jin et al., 2017). Additionally, secreted vesicles, called exosomes or microvesicles, are emerging factors in the evolution and progression of different cancer types, including leukemia (Krause and Scadden, 2015) (see poster and Box 2).

In the past, insensitivity towards available chemotherapeutic regimens or disease relapse has primarily been attributed to acquired chemoresistance of LSCs. However, more recently, contributions of the BMM to chemoresistance have become evident. The interaction of the CML cell line K562 with the extracellular matrix component fibronectin, which is mediated by β1 integrin, downregulated the proapototic factor Bim1, thereby conferring a phenotype of resistance against chemotherapeutic agents (Hazlehurst et al., 2007). Furthermore, through their integrin α4β1 [also known as very late antigen 4 (VLA4) receptor], leukemic cells interact with vascular cell adhesion molecule (VCAM-1) on stromal cells, leading to the activation of the NF-κB pathway in stromal cells and chemoresistance (Jacamo et al., 2014). Additionally, the VLA4 (integrin α4β1) and CD44 axes have been shown to regulate drug efflux that is mediated by their interaction with the mesenchymal niche, in both normal hematopoiesis and in AML blasts (Malfuson et al., 2014). N-cadherin, another cell adhesion molecule, has been shown to mediate the adhesion of CML cells to MSCs through an increase in the number of N-cadherin–β-catenin complexes and an increase in Wnt signaling, which has a crucial role in mediating resistance against tyrosine kinase inhibitors (Zhang et al., 2013). Furthermore, forced expression of galectin-3 in CML cell lines induced the expression of induced myeloid leukemia cell differentiation protein (Mcl-1), which promoted proliferation and drug resistance, whereas in the in vivo context, galectin-3 supported the localization of CML cells to the BMM (Yamamoto-Sugitani et al., 2011). More recently, galectin-3 was found to be elevated in the serum and ex vivo studies showed that galectin-3 also originates from stromal cells besides leukemic cells in B-ALL (Fei et al., 2013). In another study, soluble or stroma-cell-bound galectin-3 was taken up by ALL cells, where it modulates the NF-κB pathway and protects pre-B ALL cells from chemotherapeutic agents (Fei et al., 2015). Furthermore, the turnover of β1 integrins and CD44, both of which are associated with chemoresistance in leukemia cells, is dependent on galectin-3 (Lakshminarayan et al., 2014). In addition, some studies have described a role for the focal adhesion kinase (FAK) inhibitor VS-4718 in BCR–ABL1⁺ B-ALL with mutations in the DNA-binding protein Ikaros (IKZF1) in releasing the adhesion of leukemia cells from the stroma, thereby increasing their chemosensitivity (Churchman et al., 2016; Joshi et al., 2014).

As outlined above, the BMM provides a protective environment for LSCs and helps them to evade eradication, which is a prerequisite for a cure (Ishikawa et al., 2007; Krause and Scadden, 2015; Zhang et al., 2013). The majority of therapies against hematological malignancies and cancers, in general, are aimed at the cancer cells themselves. However, the ‘sheltering’ of leukemic stem cells by the BMM and their resulting persistence throughout therapy make it necessary for researchers and clinicians alike to aim at targeting their interactions with the BMM to augment the existing treatment strategies. Indeed, a number of past and recent studies provide hope that such innovative strategies are being attempted. These strategies can be divided into those that target (1) interactions of leukemia cells with the BMM, (2) leukemia cell intrinsic pathways and (3) direct BMM targeting strategies. An example for the first approach is granulocyte colony-stimulating factor (G-CSF), a well known HSC mobilization agent used in HSC transplantation; it induces HSC mobilization by modulation of niche components such as the CXCR4–CXCL12 axis and by inducing metalloprotease (Bendall and Bradstock, 2014). A randomized trial in CML has shown that G-CSF in combination with established chemotherapeutic treatment improved disease-free survival in adults (Löwenberg et al., 2003; Pabst et al., 2012). CXCR4, the receptor for the cytokine SDF-1α, is also frequently targeted directly (Abraham et al., 2017; Hoellenriegel et al., 2014; Sison et al., 2014, 2013). In addition, two independent studies have recently targeted CXCR4 with an IgG antibody (PF-06747143), which revealed a strong affinity to AML cells and inhibited their chemotaxis towards CXCL12. It also reduced the tumor burden in other hematological malignancies and in a xenotransplantation model for AML (Liu et al., 2017; Zhang et al., 2017).

E-selectin can be inhibited by the small-molecule antagonist GMI-1271 (Natoni et al., 2017).

The second approach of targeting leukemia cell intrinsic pathways (partly reviewed in Krause and Scadden, 2015) includes the inhibition of Axl and its interaction with Gas6 in AML (Ben-Batalla et al., 2013), for instance with the compound MYD1-72 (Kariolis et al., 2017), or the inhibition of CCL3 (Frisch et al., 2012; Schepers et al., 2013), placental growth factor (PlGF; Schmidt et al., 2011), IL-6 (Welner et al., 2015) and Jak2 in myeloproliferative neoplasia (MPN) (Meyer et al., 2015) and AML (Karjalainen et al., 2017). Other possible therapeutic strategies are inhibition of NF-κB, which lies downstream of the vascular cell adhesion molecule (VCAM)-1–integrin-β1 signaling axis, of various cytokines or their receptors (Jacamo et al., 2014) or targeting of CD44 (Erb et al., 2014; Singh et al., 2013). Other forms of treatment are the Bruton tyrosine kinase (BTK) inhibitor PCI-32765 in CLL (Byrd et al., 2013), or inhibition of the ‘moulding’ of the niche, for instance through the inhibition of exosome formation with carboxyamidotrizole-orotate (OTC) (Corrado et al., 2012) or that of nanotubes, resulting in a block in the delivery of the leukemia-promoting ‘cargo’ these tubes carry. Finally, with regard to the third approach that is directly aimed at the BMM, even the targeting of bone remodeling with parathyroid hormone (Krause et al., 2013) or treatment of hypoxic niches in the leukemic BMM (Benito et al., 2016) may be feasible (see poster and Box 2). In this respect, osteoblast-specific activation of the receptor for parathyroid hormone (PTH) and PTH-related peptide attenuated BCR–ABL1-driven CML owing to the suppressive effects of TGF-β1, which is derived from the remodeling bone, but had no effect on AML induced by the oncogene MLL–AF9 in murine models (Krause et al., 2013). Furthermore, in a xenotransplantation model of AML, the vascular damage that was induced by increased levels of ROS or nitric oxide (NO) in endothelial cells, which led to increased vascular leakiness and increased hypoxia, could be decreased by inhibitors of NO production, while improving normal hematopoiesis and treatment outcome (Passaro et al., 2017).

The normal hematopoietic and leukemic bone marrow microenvironments are highly complex entities, which interact in a distinct manner with either normal hematopoietic or leukemic stem cells. Owing to improved and emerging technologies in microscopy, as well as various molecular techniques, our understanding of these niches is growing steadily, but it is still far from being complete. In order to make further strides in our efforts to cure hematological malignancies, the exploitation of novel avenues of treatments, such as the targeting of the BMM – after it is thoroughly understood – seems like a logical and worthwhile cause.

The Georg-Speyer-Haus is funded jointly by the German Federal Ministry of Health (BMG) and the Ministry of Higher Education, Research and the Arts of the State of Hessen (HMWK). The LOEWE Center for Cell and Gene Therapy Frankfurt is funded by HMWK, reference number: III L 4-518/17.004 (2010).