Structural organization of the bone marrow and its role in hematopoiesis.

Abstract

Purpose of review

The bone marrow is the main site for hematopoiesis. It contains a unique microenvironment that provides niches that support self-renewal and differentiation of hematopoietic stem cells (HSC), multipotent progenitors (MPP), and lineage committed progenitors to produce the large number of blood cells required to sustain life. The bone marrow is notoriously difficult to image; because of this the anatomy of blood cell production- and how local signals spatially organize hematopoiesis-are not well defined. Here we review our current understanding of the spatial organization of the mouse bone marrow with a special focus in recent advances that are transforming our understanding of this tissue.

Recent Findings

Imaging studies of HSC and their interaction with candidate niches have relied on ex vivo imaging of fixed tissue. Two recent manuscripts demonstrating live imaging of subsets of HSC in unperturbed bone marrow have revealed unexpected HSC behavior and open the door to examine HSC regulation, in situ, over time. We also discuss recent findings showing that the bone marrow contains distinct microenvironments, spatially organized, that regulate unique aspects of hematopoiesis.

Summary

Defining the spatial architecture of hematopoiesis in the bone marrow is indispensable to understand how this tissue ensures stepwise, balanced, differentiation to meet organism demand; for deciphering alterations to hematopoiesis during disease; and for designing organ systems for blood cell production ex vivo.

INTRODUCTION

Hematopoiesis takes place in the bone marrow (BM) where hematopoietic stem cells and multipotent progenitors (HSPC) self-renew and progressively differentiate into lineage-specific, unipotent, progenitors responsible for production of each major blood lineage. The bone marrow has been studied in detail using multiple approaches including scRNAseq, and in vivo lineage tracing studies [1–9]. These and other studies have dramatically changed our understanding of how the different stem and progenitor populations differentiate, and how they are regulated by the BM microenvironment- the collection of hematopoietic and stromal cells and structures that supports differentiation- during normal and stress hematopoiesis. Our understanding of the spatial organization of hematopoiesis in the bone marrow is less comprehensive. Spatial analyses of differentiating progenitors, their offspring, and the supporting microenvironment are challenging due to several factors (reviewed in [10]); a) the bone marrow is fully enclosed by opaque bone which makes direct observation difficult and requires extensive preparation steps in order to generate high quality samples for imaging analyses; b) the hematology field has used increasingly complex combinations of cell surface markers -requiring simultaneous detection upwards of 15 antibodies- to define each hematopoietic progenitor and mature cell in the bone marrow. In contrast fluorescent analyses are generally limited to much fewer (4-7) parameters preventing simultaneous identification of multiple cell types. Further, many antibodies used to define cells by flow cytometry fail to detect the same cells in imaging analyses, either because the signals are too dim or because sample preparation destroyed the epitopes recognized by that antibody; c) scRNAseq analyses of stromal cells in the bone marrow have revealed extraordinary complexity [7–9**]. However, there are no validated antibodies to detect many of these stromal populations and the field relies in Cre/fluorescent reporter mice that identify some stromal components but fail to completely resolve the different populations [11]; d) the bone marrow contains large numbers of mature cells but stem cells and progenitors are exceedingly rare. This makes identification of sufficient numbers of HSPC for adequately powered statistical analyses very challenging and time consuming; e) different groups have used different statistical approaches and methods to define proximity of cells to structures and a global consensus on which approaches to use has yet to emerge. Despite numerous challenges the field has made tremendous progress in defining the architecture of the BM and deciphering how local cues from the microenvironment regulate stem and progenitor cells. Here we summarize our current understanding of the spatial organization of the bone marrow, its impact on hematopoiesis, and discuss recent discoveries that are transforming the field.

Overall organization of the bone marrow.

The main structures that spatially organize the bone marrow are the bone, the vasculature, and a network of reticular stromal cells. The bone completely encloses the bone marrow, defines its boundaries, and projects trabeculae that penetrate into the BM parenchyma (Fig. 1). The bone marrow vasculature is composed of rare arterioles that enter through the bone and transform into transitional vessels that give rise to an extremely dense network of fenestrated sinusoids that occupy most of the BM space (Fig. 1). The vasculature is tightly associated with a network of perivascular reticular cells that spreads through the BM. Hematopoiesis takes place in the spaces between vessels, bone, and reticular cells (Fig. 1). Many other types of stromal (non-hematopoietic) cells are present in the bone marrow including sympathetic nerves, Schwann cells, adipocytes, osteoblasts, osteocytes, osteoblastic precursors, and diverse types of fibroblasts. These are reviewed elsewhere [12–14]. These cells and structures –in association with different types of hematopoietic cells-cooperate to provide distinct microenvironments that regulate –and regionally organize-hematopoiesis in the bone marrow.

Figure 1.

Schematic representation of the spatial organization of the mouse bone marrow under homeostasis. The endosteum, the vasculature and a network of reticular stromal cells define the volumes available for hematopoiesis. vWF+ HSC reside in a sinusoidal/megakaryocytic/reticular niche far from arterioles and the endosteum while vWF- reside in an arteriolar niche enriched in Ng2+ cells [38]. Note that this arteriolar niche also contains sinusoids and reticular cells. HSC in the central BM constantly traffic between reticular cells [27**]. Subsets of HSC -reserve HSC [33*] and MFG-HSC [26] localize to endosteal regions where they proliferate in response to stress, likely in areas undergoing simultaneous bone deposition by osteoblasts and bone resorption by osteoclasts [26]. GMP are distributed through the BM but form clusters in respond to stress [50]. Lymphoid progenitors have been mapped to the endosteum [18] but also to different types of reticular cells [42,60]. Erythropoiesis takes place in erythroblastic islands presumably adjacent to the same sinusoids that support erythroid progenitors [61,62,64*].

Hematopoietic stem cell niches

The best studied microenvironments in the bone marrow are the hematopoietic stem cell niches, which are responsible for ensuring that HSC are maintained through the life of the organism. The discovery of a two color strategy (Lin⁻CD48⁻CD41⁻CD150⁺) to detect HSC using confocal microscopy [15] led to an explosion of studies that used imaging to identify candidate HSC niche components that were later validated using complementary approaches [15–22]. These analyses have been further refined by the development of mouse fluorescent reporter lines that identify populations highly enriched in HSC [23–27**]. These studies showed that –in the steady-state- HSC are always found as single cells and adjacent to perivascular cells and sinusoids. Most HSC exclusively localize to sinusoids but smaller fractions localize to areas that also contain arterioles and endosteal surfaces. Cells associated with each of these structures produce cytokines and growth factors that regulate HSC self-renewal and function (Fig. 1). The precise components of HSC niches and how they regulate HSC have been reviewed in detail elsewhere [13,28,29]. Here we will highlight recent insights from live imaging analyses of HSC.

Insights from native HSC imaging

Until recently live imaging of HSC in the bone marrow was restricted to experiments were HSC were prospectively isolated, transferred into recipient mice, and then imaged [30]. This has changed with the development of live imaging approaches of unperturbed HSC. Christodoulou et al., [26**] used Mds1^GFP+, and Mds1^GFP/+Flt3-Cre mice. In Mds1^GFP+ mice the Mds1 promoter drives GFP expression in HSC and multipotent progenitors. However, in the Mds1^GFP/+Flt3-Cre mice, Cre expression results in excision of the GFP cassette in all cells except a small (12%) subset of quiescent LT-HSC (MFG-HSC). Using live imaging of the calvarium they found that both the Mds1^GFP+ HSPC and MFG-HSC were adjacent (less than 10μm) to blood vessels. However, HSPC preferentially associated with transition zone vessels when compared to the MFG-HSC. In contrast the MFG-HSC were closer to the endosteum and sinusoids suggesting the existence of different microenvironments for HSC and downstream progenitors. Live imaging demonstrated that –in the steady-state- the MFG-HSC were largely non-motile (moving less than 10μm over a period of two hours) whereas the Mds1^GFP+ HSPC migrated more and further. Treatment with chemotherapy and G-CSF -which dramatically induces HSC proliferation and mobilization into the circulation- led to the formation of clonal MFG-HSC clusters in endosteal regions undergoing both bone deposition and remodeling. This study demonstrates live imaging of a subset of minimally motile LT-HSC and suggests that a unique endosteal microenvironment supports MFG-HSC expansion after chemotherapy injury. Note that multiple studies have shown that less than 10% of LT-HSC localize near the endosteum and that most are associated with sinusoids in the central BM [12,17,21,23]. These suggest that the MFG-HSC represents a subset of HSC that specifically associates with the endosteum (Fig. 1). A subset of macrophages also localizes near the endosteal surface (osteomacs). These macrophages promote HSC retention in the bone marrow, and are suppressed after mobilizing doses of G-CSF [31,32]. It would be of great interest to the field to examine whether these osteomacs localize near the MFG-HSC as this will further support the existence of a discrete niche for amplifying and mobilizing HSC in response to stress.

Upadhaya et al., [27**] used Pdzk1ip1-^CreER:tdTomato mice for live imaging of HSC. In these mice low dose tamoxifen expression results in TdTomato expression in 23% of LT-HSC. Live imaging of mouse calvarium or tibia showed that the labeled HSC are highly motile with ~90% of the labeled HSC moving more than 20μm whereas ~10% of the labeled HSC showed minimal movement. Combining Pdzk1ip1-^CreER:tdTomato with Fgd5Z^sGreen or KitLGFP/+ reporter mice allowed visualization of endothelial cells or stem cell factor (SCF)-producing perivascular cells. These confirmed the perivascular location of HSC but also revealed that –over short periods of time- HSC form multiple, close, transient contacts with various SCF-producing cells. Thus HSC might travel between different niches/microenvironments to receive different signals that regulate their behavior (Fig. 1). Surprisingly a drug treatment that inhibits the CXCR4 receptor and α4β1/α9β1 integrins -and mobilizes HSC to the circulation- also blocked HSC movement in the BM. This indicates that HSC movement requires CXCL12-CXCR4 and/or integrin signaling.

Although additional analyses are needed to resolve the observed discrepancies in motility between the HSC examined, the Christodoulou et al., and Upadhaya et al., studies open the door to deciphering HSC regulation by different signals, in situ, with single cell resolution.

Different bone marrow regions have unique functions in regulation of hematopoiesis

It is becoming increasingly clear that hematopoiesis in the bone marrow is spatially and regionally organized and that local cues produced by distinct microenvironments are responsible for regulating different HSPC. The best characterized example of this spatial heterogeneity is the data supporting the existence of distinct sinusoidal and arteriolar niches for HSC. As discussed in the previous sections most HSC localize reside exclusively in sinusoidal locations whereas smaller fractions also associate with arterioles and/or the endosteum [12,17,21,23,24,33*,34,35]. Note that the precise fractions of HSC associated which each structure –and whether these associations are specific-remains a source of controversy. Each group has used different methods to identify HSC and different criteria to define proximity to each type of structure. These further highlight a need for a common criteria in the field for defining cell proximity to niche components. Also note that –due to the abundance of sinusoids- almost all hematopoietic cells locate within 30μm of a sinusoid [36]. Therefore an arteriolar niche or endosteal niche is going to also contain sinusoids [12,26**,36]. Kunisaki et al., showed that 30% of Lin⁻CD48⁻CD150⁺ HSC localized near arterioles ensheathed by Ng2+ periarteriolar cells and that Ng2+ cell ablation caused loss of HSC quiescence and function [17]; another 30% of HSC specifically map within 5μm of megakaryocytes and loss of megakaryocytes or megakaryocyte-derived CXCL4 or TGFβ resulted in HSC proliferation in sinusoidal locations without affecting HSC in arteriolar locations [21,22]. Pinho et al., found that von Willebrand factor (vWF) positive HSC, which are biased towards megakaryocyte fates [37] selectively localized near megakaryocytes (60% of vWF⁺ HSC are within 5μm of a megakaryocyte) whereas vWF- HSC localized near arterioles. Megakaryocyte ablation specifically expanded vWF⁺ HSC [38]. Itkin et al., discovered that HSC could be fractionated based on intracellular ROS (reactive oxygen species) levels and that HSC with lower levels of ROS where enriched near arterioles whereas HSC with higher levels of ROS located near sinusoids [39]. Further data supporting a distinct arteriolar HSC niche was provided by Kusumbe et al., which showed that constitutive Notch signaling in the vasculature increased the number of arterioles in the BM followed by accumulation of HSC suggesting that arteriole number controls HSC abundance. Together these studies support the existence of a megakaryocyte/sinusoidal niche that maintains HSC biased towards megakaryocyte fates and an arteriolar niche that maintains more quiescent HSC (Figure 1).

There is also evidence supporting the existence of a distinct endosteal HSC niche. After adoptive transfer into recipient mice the donor HSC are selectively enriched near endosteal cells [30,40,41]. In agreement Zhao et al., discovered that CD48⁻CD49b⁻ HSC are resistant to chemotherapy and proposed that they represent a reserve HSC (rHSC) population. Sixteen percent of these rHSC localize -and amplify after chemotherapy – near the endosteum, adjacent to N-cadherin+ stromal cells that support them [33*]. Live imaging analyses also demonstrated that –after chemotherapy- a subset of HSC selectively proliferate in endosteal regions undergoing bone remodeling and deposition [26**]. Together these studies support the concept that the endosteum might provide a niche for regenerating HSC (Fig. 1).

The localization of progenitors downstream of HSC is less characterized. Multipotent progenitors are immediately below HSC in the hematopoietic hierarchy and are major contributors to blood cell production in the steady-state [1–3]. Live animal imaging of transplanted HSC -or a population enriched in MPP- into non-irradiated recipients showed that the MPP located further away from the endosteum [30]. In agreement live animal imaging of Mds1^GFP+ HSPC –also enriched in MPP- and MFG-HSC [26**] showed different spatial distributions for these two populations and increased HSPC localization near transitional vessels. These studies suggest that HSC and MPP might occupy different niches. In contrast, Cordeiro-Gomes et al., found similar spatial organization for HSC and MPP and –in rare occasions- observed colocalization of HSC and MPP suggesting that they occupy the same niche [42]. Note that each of these studies used different markers/reporter mice to define HSC/HSPC/MPP as well as different imaging approaches (live imaging of transplanted cells/live imaging of subsets of HSC and HSPC in the calvarium/fixed femur whole mounts) and additional studies are needed to resolve the question on whether HSC and MPP (of which there are multiple subsets [43]) occupy the same or distinct niches.

Several components of the bone marrow microenvironment including endothelial cells [44,45], perivascular cells [46–48], osteocytes [49], megakaryocytes [50], and even neutrophils [51] produce signals that support and regulate myeloid cell production in the steady-state and after stress (reviewed in [52]). However, the specific sites for myelopoiesis in the bone marrow, or whether myeloid progenitors and HSC share the same niche, remain unknown. This is mainly due to lack of approaches to image myeloid progenitors. Herault et al., were able to image a population of classically defined granulocyte monocyte progenitors (GMP) as Lin-Sca1-CD150-c-kit+FcγR+ cells [50]. Note that subsequent studies have shown that these phenotypically defined GMP are heterogeneous and contain bipotent and unipotent monocyte and granulocyte progenitors [4,53]. Herault et al., showed that these heterogeneous GMP were almost always found as single cells, distributed through the bone marrow. Insults that trigger emergency myeloid cell production induced formation of tightly packed GMP clusters. This cluster formation required signals provided from the microenvironment [50] suggesting that specific regions of the bone marrow support emergency myeloid progenitor expansion in response to stress.

Similarly, lymphopoiesis is dependent on signals produced by perivascular stromal cells [42], endothelial cells [42], and osteoblastic lineage cells, which include osteoblastic progenitors, osteoblasts, and osteocytes [18,20,41,54–59]. While these studies support the concept of an endosteal niche for lymphopoiesis the spatial localization of lymphoid progenitors is not clear. Ding et al., found that 30% of Lin⁻IL7Ra⁺ cells –which are enriched in lymphoid progenitor- were in contact with the endosteum [18]. In contrast, Tokoyoda et al., found that B220⁺flk2⁺ pre-pro-B cells and B220⁺c-kit⁺ pro-B cells were scattered through the central bone marrow. Additionally, 65% of Pre-pro-B cells and 0% pro-B cells were in contact with CXCL12 producing reticular cells whereas 11% of Pre-pro-B cells and 89% of pro-B cells contacted IL7-producing reticular cells. These suggest a perivascular location for lymphopoiesis and that CXCL12 and IL7 producing stromal cells provide niches for different stages of lymphocyte maturation [60]. Surprisingly, Cordeiro-Gomes et al., found that IL7-producing cells are a subset of CXCL12-producing reticular cells and that Ly6D⁺ common lymphoid progenitors localize to this subset. Since HSC also associate with IL7⁺CXCL12⁺ reticular stromal cells [42] this also suggested that common lymphoid progenitors and HSC occupy the same niche. Additional studies are necessary to reconcile these findings.

Erythropoiesis is also regionally organized; classical studies demonstrated that a subset of macrophages that localize near sinusoids provides a niche that supports islands of erythroblasts maturation [61,62] (for a recent review see [63]). Recently, Comazzetto et al., were able to image Lin-Sca1-c-kit+CD105+ erythroid progenitors. These selectively localize to reticular stromal cells in perisinusoidal locations. These stromal cells maintained these adjacent progenitors via SCF secretion [64**]. These indicate that sinusoids are the site of erythropoiesis.

Conclusions and open questions

The bone marrow is highly organized and contains specialized regions that provide distinct microenvironments that selectively regulate unique types of hematopoietic cells (Figure 1). The field has made tremendous progress in defining the spatial architecture of hematopoiesis. The development of approaches to image hematopoiesis in vivo will further transform the field by allowing visualization of cell decisions in real time. However, several challenges remain including: a) the lack of approaches to simultaneously image many types of hematopoietic progenitors and precursors. These prevent examination of stepwise differentiation in situ to determine how local signals impact progenitor function; b) scRNAseq analyses have identified several new types of stromal cells with unknown functions in hematopoiesis and shown that known populations –e.g endothelial cells and perivascular cells- are highly heterogeneous with different subsets producing unique combinations of cytokines and growth factors [7–9**]. Visualization of these novel populations and subsets will likely lead to the identification of unique niches for hematopoietic progenitors and precursors. The development of novel techniques allowing imaging of cytokines in the BM [65*] will be invaluable for these approaches; c) the bone marrow extracellular matrix is increasingly being recognized as a key regulator of hematopoiesis [66] that is spatially organized [67]. How differentiating hematopoietic cells interact with the extracellular matrix remains poorly understood; d) most studies in the field have focused in dissecting how each type of cell in the microenvironment interacts with- and regulates- one type of HSPC. However, multiple stromal cell types cooperate to regulate each type of HSPC and different stromal cells regulate different stages of HSPC maturation [58,60,68]; how the bone marrow ensures that each HSPC localizes to the right microenvironment as they mature remains unknown. Answers to these questions will define how the spatial architecture of the bone marrow regulates hematopoiesis during homeostasis and disease and allow the development of culture systems containing all the niche structures to necessary to produce large amounts of blood ex vivo.

Key points:

• The spatial organization of the bone marrow ensures that distinct microenvironments regulate different types of stem cells and progenitors.

• The best studies microenvironments are HSC niches and mounting evidence supports the existence of distinct sinusoidal and arteriolar HSC niches.

• It is becoming increasingly clear that the bone marrow contains distinct niches for progenitors downstream of HSC.

• Despite tremendous progress the spatial organization of hematopoiesis remains poorly understood and new approaches are needed.

• New live imaging studies of native HSC open the door to examine HSC regulation, in situ.