Dynamic niches in the origination and differentiation of haematopoietic stem cells

Introduction

In adult humans, haematopoietic stem cells (HSCs) are responsible for generating ~1×10⁹ red blood cells and ~1×10⁸ white blood cells every hour, including a full complement of platelets and other mature blood lineages. HSCs must also self-renew, and thus must constantly integrate and respond to myriad external inputs to maintain haematopoietic homeostasis. HSCs typically are depicted at the top of a hierarchical “lineage tree”, each branch point of which indicates a restriction in developmental potential (Fig. 1).

During mammalian development, haematopoiesis occurs in sequential stages: first primitive, and then definitive blood formation. These stages are temporally and anatomically distinct, invoking unique cellular and molecular regulators. The formation of primitive blood cells occurs early during fetal life, with coordinated progression from extraembryonic to intraembryonic sites of haematopoiesis. Within the embryo, definitive haematopoiesis undergoes developmentally stereotyped transitions; HSCs arising from the aorta-gonad-mesonephros (AGM) region migrate first to the placenta and fetal liver, and then to the spleen. Eventually, haematopoiesis shifts to the bone marrow, where homeostatic blood formation is maintained postnatally.

As proposed initially by Schofield¹ in his 1978 description of the regulation of blood formation in the marrow cavity, both primitive and definitive haematopoiesis require input from the cellular microenvironment, or ‘niche’. When first proposed, this concept held that a stem cell must associate “with other cells which determine its behaviour” in order to “prevent its maturation”¹; loss of this association was hypothesized to result in differentiation. This idea has evolved, and the concept of the niche now includes specific cell types, anatomical locations, soluble molecules, signalling cascades and gradients, as well as physical factors such as shear stress, oxygen tension, and temperature^2–7. Inputs from the niche can be permissive of, or conducive to, homeostatic HSC self-renewal and differentiation, but also may constrain normal haematopoiesis under pathological conditions such as myelodysplasia, aging, and haematologic malignancy. Yet, the definition of the niche as a microenvironment that provides spatially and temporally coordinated signals to support stem cell function has remained.

Niches were characterized initially in invertebrate model organisms, such as worms and flies^8–10, and subsequently identified in mammals using targeted genetic manipulations^11–15. Our conception of the haematopoietic stem cell niche also has been informed by studies in non-haematopoietic tissues, such as the hair follicle, auditory hair cell, and intestinal crypt. In these tissues, clear spatial constraints make identification of niche components more straightforward than in the haematopoietic system, where blood cells are widely disseminated and constantly in motion¹⁶. This spatially-defined concept of the niche, which has been derived from anatomically static tissues, has been translated across organisms and tissue types, providing a generalized model for stem cell regulation. However, in certain cases, direct evidence that this model is applicable still is lacking. Nonetheless, it is widely accepted that niches exist in most, if not all, tissues, and that they provide both basic cellular necessities, such as mechanical support, trophic factors, and hospitable physical and chemical conditions, as well as stem cell-specific self-renewal and differentiation cues (Fig. 2).

Several new and elegant techniques and model systems have been applied to HSC development, permitting an improved functional and anatomical dissection of HSC interactions with the niche. In particular, real time in vivo imaging has enabled the direct visualization of HSCs and their niches, providing key insights into the origins, dynamics, and physiological regulation of the anatomic compartments in which HSCs reside. It is clear now that signals from a diversity of non-haematopoietic cell types play a coordinated part in ensuring proper HSC function. Moreover, although certain features appear to be common among niches that serve HSCs and those that serve non-HSCs, our improved understanding of the HSC niche underscores important distinctions from the niches found in solid tissues. In particular, haematopoietic niches may not be anatomically restricted, and the signals provided to HSCs may be spatially as well as temporally compartmentalized. Thus, altering either the location or the timing of niche-dependent signalling may initiate divergent developmental outcomes.

In this Review, we discuss the HSC niche from an ontogenic perspective. We address current theories and open questions surrounding niche cells and their role in regulating fetal and adult definitive haematopoiesis. HSC biology is a rapidly expanding field, and recent advances have increased interest in the fundamental biology of these cells and in their potential as a source material for gene and cell therapy. Deeper understanding of the signals that stem cells receive from their niche as they grow and differentiate is essential for optimizing their therapeutic benefit.

Vertebrate haematopoiesis

Vertebrate haematopoiesis takes place in ontogenic waves. In both fish and mammals there are primitive phases, in which blood stem cells produce only a subset of specialized blood lineages, and definitive phases, in which true HSCs appear. In zebrafish, at least three distinct phases of haematopoietic development exist, and in mammals there are at least two. Primitive haematopoiesis is first detected extra-embryonically, in the yolk sac, as early as day e7.5 in mice and 30 days post-conception (pc) in humans^17–19 (Fig. 3). Although evidence exists that cells isolated from the yolk sac can reconstitute haematopoiesis fully^{20, 21}, it has recently become clear that multipotent HSCs also emerge directly from the AGM region (at day e10.5 in mice^{22, 23} and 4 weeks pc (wpc) in humans²⁴) to initiate definitive haematopoiesis. In all studied vertebrates, in fact, the first HSCs arise from the ventral wall of the dorsal aorta^25–30, although they subsequently seed additional tissues and organs in a species-specific manner. In mammals, haematopoiesis moves from the AGM to the placenta and fetal liver; for example, by day e11–12 in mice and by 5 wpc in humans. Haematopoiesis in mice shifts to the spleen by day e14 and begins to shift to the bone marrow by e18; in humans, it moves from the fetal liver to the bone marrow at 12 wpc. By birth in humans, the bone marrow supports the vast majority of haematopoiesis, and extramedullary multilineage haematopoiesis is negligible except in cases of pathology. By contrast, in mice, splenic haematopoiesis persists for several weeks postnatally³¹. Finally, in fish, the kidney serves as the predominant site of haematopoiesis in adult life.

The molecular underpinnings of the dynamic progression of blood cell formation during development are complex and still under investigation (see below). It is clear that many cytokines influence the differentiation and proliferation of haematopoietic stem and progenitor cells (HSPCs, a designation which includes both HSCs and multipotent and oligopotent progenitors). However, only a few have been demonstrated genetically to be necessary for HSC maintenance in vivo; these have been reviewed extensively elsewhere^32–34, and include angiopoietin³⁵, CXC chemokine ligand 12 (CXCL12; also known as SDF1)^{36, 37}, stem cell factor (SCF; also known as steel factor and KIT ligand)³⁸ and thrombopoietin^{39, 40}. In addition, developmental mediators, such as Notch (reviewed in ⁴¹) and Wnt (reviewed in ⁴²) are important for HSC maintenance. Other membrane-bound receptors, such as integrins⁴³ and cadherins⁴⁴, also may contribute. Many of these cytokines and developmental regulators are produced by niche cells and by haematopoietic cells.

HSCs require interactions with distinct niches at each stage of ontogeny. Thus, the sequential exposure of HSCs to different extracellular cues may have an important influence on their development and maturation; alternatively, perhaps the common features among these niches are exploited by HSCs to support their function in otherwise variant environments. Regardless, the anatomical relocalization of HSCs during development imparts unique challenges for the niches that support them, and suggests that haematopoietic niches may be one of the most dynamic of stem cell supportive niches.

The niche in fetal haematopoiesis

It has recently become possible to visualize individual HSCs arising from endothelial cells (ECs) of the ventral aspect of the dorsal aorta, using direct in vivo imaging techniques^{22, 45}. These imaging data provide strong support for the hemogenic endothelium hypothesis^{46, 47}, which posits that HSCs arise directly from committed ECs. The hemogenic endothelium model is related to, but distinct from, the “hemangioblast” hypothesis, which supposes the existence of a common, bipotent precursor that gives rise to both the vasculature and blood cells during development. Aside from the implications of this work with respect to the origin of HSCs, these imaging studies also argue that the niche for the very earliest HSCs is provided by the very same cells that give rise to those HSCs (Fig. 4).

Newly-formed HSCs emerging from the ventral aspect of the dorsal aorta in the AGM either bud off into the aortic circulation or remain embedded within the endothelium^{22, 28, 45}. The ECs that do not convert to HSCs are capable of supporting HSCs in vitro⁴⁸. One group has undertaken gene expression profiling of the haematopoietic AGM region and performed in situ hybridization to corroborate the results. Although it is difficult to identify cells definitively by in situ analysis of embryo sections, these experiments seem to suggest that p57 (also known as KIP2 and CDKN1C) and insulin-like growth factor 2 (IGF2) are upregulated in ECs underlying emerging HSCs⁴⁹. It remains unclear, however, precisely how these niche factors influence HSC function. p57 is a cyclin-dependent kinase inhibitor expressed in adult HSCs⁵⁰ and it is difficult to envision a cell-extrinsic role for this protein in HSC function. By contrast, IGF2 is a secreted growth factor, and culture of HSCs with IGF2 increases the frequency of mixed and erythroid colonies and decreases the frequency of macrophage and granulocyte-macrophage colonies in methylcellulose culture. These data led the authors to conclude that IGF2 signalling promotes the maintenance of primitive multipotent cells and early erythroid precursors at the expense of more mature myeloid precursors⁴⁹. Whatever mechanisms underlie the haematopoietic regulatory function of hemogenic endothelium, these ultimately vanish when distinct somitic endothelial precursors migrate to the ventral aspect of the dorsal aorta and replace the hemogenic endothelium, coincident with the disappearance of haematopoietic potential from the AGM⁵¹.

Within half a day of the emergence of HSCs from the ventral aortic endothelium, they can be found in the yolk sac, placenta, and fetal liver. Quantitative analysis suggests that HSCs are independently generated in the yolk sac, placenta and AGM, and cells generated in all of these anatomic sites may colonize the fetal liver^52–54. The mechanistic basis for HSC emergence in the yolk sac and placenta has not been defined, although by analogy with the AGM, HSCs here might also arise locally from developmentally regulated hemogenic endothelium. Likewise, the factors responsible for the subsequent engraftment of HSCs in the fetal liver are still being elucidated. The chemokine CXCL12, which is expressed by stromal cells, and its G-protein coupled receptor CXC chemokine receptor 4 (CXCR4; also known as fusin and CD184), which is expressed on haematopoietic cells and some cancer cells, are clearly important for proper HSC function and migration, including HSC engraftment into fetal liver^{36, 55, 56}. Likewise, the cytokine SCF, which is produced by stromal cells, and its receptor KIT (also known as CD117 and SCFR), which is expressed on pluripotent haematopoietic cells, have long been recognized as critical for HSC function^57–59. SCF–KIT signalling also appears to amplify the migratory response of fetal HSCs to CXCL12⁵⁷. Additional factors suggested to be important for fetal HSC migration and engraftment include α4 integrin⁴³, neural cadherin (N-cadherin) and osteopontin⁶⁰, which are membrane-bound adhesion molecules that can activate WNT signalling. Other factors, like the transcription factor pituitary homeobox 2 (PITX2), must be expressed in stromal cells in order to support normal fetal HSC function and maintenance, although the mechanism underlying this effect remains unclear⁶¹. More recently, BRG1-associated factor 250A (BAF250A; also known as ARID1A), a SWI/SNF (switch/sucrose nonfermentable) family member, has been shown to restrict HSC expansion in the fetal liver in a cell-non-autonomous fashion⁶². BAF250a might accomplish this by altering HSC–stromal cell interactions to reduce HSC expression of self-renewal and survival factors and enhance expression of genes associated with senescence and apoptosis⁶².

At about day e17.5 in the mouse, fetal haematopoiesis moves to the bone marrow⁵⁷. This relocalization is thought to result from the development, at around day e14.5, of osteoblast and chondrocyte precursor cells that are capable of forming an HSC niche⁶³. The signals that are responsible for transferring the bulk of haematopoiesis from the liver to the bone marrow are not entirely understood. However, CXCR4–CXCL12, SCF-KIT, TIE2–angiopoietin, integrin and CD44–epithelial cadherin (E-cadherin) adhesion and signalling pathways appear to be important for this process⁴⁴ and for retaining HSCs within the marrow space⁶⁴. In addition, transformation of the liver into a synthetic and digestive organ, which is marked by the differentiation of hepatoblasts into hepatocytes, beginning at about day e15.5 in the mouse⁶⁵, probably results in the replacement of haematopoietic niches with hepatic functional units. This change might drive the relocation of fetal liver HSCs to the bone marrow, where HSC–niche cell relationships enter their adult phase (discussed further below).

Taken together, these recent and historical findings support a co-developmental model of HSC–niche cell interactions throughout ontogeny, wherein haematopoiesis in a new niche occurs concurrently with the maturation of that niche’s ability to support haematopoiesis. This model also implies that the waning of haematopoiesis in a particular ontogenic niche may result from the execution of a developmental programme that makes the prior niche inhospitable (Fig. 4), or from the emergence of a new, more ‘attractive’ niche.

Acellular factors in the adult niche

In postnatal life, the bone marrow represents the primary reservoir for HSCs and haematopoietic progenitors in both mice and humans. The activity of adult HSCs is carefully regulated by a number of different signals, including many of the cytokines discussed above. Many non-proteinaceous factors, including calcium ions, oxygen tension, and reactive oxygen species, also are critical for proper HSC regulation, although such elements have only recently been integrated into the concept of the bone marrow niche. In addition, physical and mechanical inputs, such as shear force and rigidity, are becoming better defined as contributors to niche haematopoiesis. As the role of cytokines in the adult HSC niche has been extensively reviewed by others^{33, 34, 44, 59, 66, 67}, we focus here on some of the more-recently identified non-cytokine niche components that are implicated in adult haematopoiesis.

Cellular factors in the adult niche

The main function of the bone marrow is to provide new blood cells, as needed, throughout life. The bone marrow is highly complex, and conventional representations of bone marrow haematopoiesis often reference an “endosteal niche,” which is populated by osteoblasts, and a “vascular niche,” which is an anatomically distinct perivascular space. However, this conception is almost certainly an over-simplification, however. Recent advances have refined our understanding of how HSCs interact with niche cells, of which cell types have niche functions and of how acellular aspects of the microenvironment affect haematopoiesis. Further evidence indicates that a single HSC can receive input simultaneously from multiple, distinct types of niche cells and that HSCs may move among these niches quite readily, data that potentially collapses models based on anatomically separate niches (see the discussion below on osteoblasts and other niche cells). In aggregate, these studies suggest a new model of marrow haematopoiesis, in which one may visualize the entire bone marrow compartment as a niche, with discrete areas and cellular elements serving distinct functions at different stages of haematopoietic development and differentiation. In this section, we review recent advances in understanding the specific cell types that comprise the bone marrow niche, and how these cells interact with HSCs.

The niche in dysregulated haematopoiesis

Although most haematopoietic malignancies are likely to arise from mutations that inappropriately activate haematopoietic cell proliferation and survival pathways, recent data demonstrate that haematopoiesis can also be dysregulated by alterations in the niche, with defects in HSCs themselves arising secondarily. The most compelling recent example comes from work in which miRNA processing was disrupted in osteoblast precursors and mice developed myelodysplasia, which in rare cases progressed to myeloid leukaemia by 3 weeks of age⁹⁶. Notably, HSCs transplanted from these mice into wild-type recipients did not transfer myelodysplasia, indicating that the HSCs were not intrinsically competent to produce pathologic changes. Similarly, APC^min (adenomatous polyposis coli, multiple intestinal neoplasia) mice, which are heterozygous for a truncated, nonfunctional allele of the tumour suppressor and negative regulator of canonical WNT signalling APC, develop myelodysplasia as they age, and this phenotype is not rescued by transplanting wild-type CD45.1 bone marrow¹²⁵. Of note, these transplantation experiments used sublethally-irradiated recipients, leaving open the possibility that the continued formation of blood cells from residual APC^min HSPCs contributed to the development of myelodysplasia in these animals. Nonetheless, these studies add to a growing body of evidence indicating that malignancy and premalignancy can arise from microenvironmental defects^{126, 127}, or from cooperation between haematopoietic and niche cells. This appears to be the case for the myeloproliferation that develops in mice deficient in the retinoblastoma (RB) protein; haematopoietic dysfunction in this model requires deletion of Rb in both niche cells and myeloid cells¹²⁷.

Other suggestions of microenvironmentally-driven haematopoietic dysfunction come from the literature on ageing. Ageing is accompanied by well-described haematopoietic changes, including an accumulation of phenotypic HSCs with diminished repopulating capacity, as well as a propensity towards myelopoiesis instead of lymphopoiesis^128–133. Although HSC-intrinsic changes undoubtedly contribute to these age-associated phenotypes¹³¹, changes also occur in the microenvironment over time¹³³. For instance, mammalian target of rapamycin (mTOR) signalling is increased in aged HSCs, presumably as a result of microenvironmental cues¹³⁴. Heightened mTOR signalling in aged HSCs suppresses their engraftment function and appears to contributed to age-dependent deficits in immune function¹³⁴. Additionally, recent proteomic analysis of bone marrow stromal cells in culture demonstrates that older stromal cells produce more hydrogen peroxide than younger cells¹³⁵, which could lead to increased production of oxygen free radicals. This potentially more-toxic environment may have deleterious effects on HSC function.

Toxic insults that cause DNA damage, such as ionizing radiation, might also adversely affect haematopoiesis, in part owing to the preferential use of particular DNA damage response mechanisms in HSCs. Quiescent HSCs preferentially employ non-homologous end joining in response to radiation-induced DNA damage, and this repair pathway is prone to introducing base deletions and additions^{136, 137}. Irradiation also has profound effects on niche cells themselves, including osteoblasts and adipocytes. Although the marrow space becomes infiltrated by adipocytes soon after irradiation^{119, 138}, the first response to a radiation insult is a transient and reversible proliferation of osteoblasts, which peaks at 48h post-irradiation and disappears by 10 days post-transplant¹³⁹. Interestingly, emerging data indicates that irradiation is more disruptive of bone marrow architecture than chemotherapy (J.R. Harris, J.P. Chute and T. Reya, personal communication). The administration of cyclophosphamide to mice at conventional doses resulted in severe myelotoxicity but significantly less damage to the bone marrow architecture in comparison to radiation, which caused extensive destruction of both haematopoietic cells and bone marrow stromal cells, as well as delayed haematologic recovery. These findings suggest that systemic toxins are likely to have very distinct deleterious effects on HSCs and stromal cells.

Finally, obesity in humans has been correlated both with changes in the bone marrow microenvironment¹²⁰ and with an increased incidence of haematologic malignancy¹⁴⁰. Although the mechanisms underlying these findings remain to be elucidated, it is probable that they arise from a combination of systemic and cell-mediated interactions. Interestingly, obese diabetic BBZDR/Wor rats exhibit increased numbers of marrow adipocytes and impaired innervation of bone marrow stromal elements relative to controls, suggesting a link between obesity and disruption of the multicomponent HSC niche described above¹⁴¹. Whether this disruption is a function of specific adipocyte-mediated local signals (for example, NRP1), systemic inflammatory signals (for example, transforming growth factor-β), or of a niche interaction that is not yet described, is a topic of great interest for future study.

A look ahead: the HSC niche in therapy

Inadequate, impaired, or dysregulated haematopoiesis accounts for a substantial fraction of morbidity and mortality in today’s society and, until recently, most therapies have focused somewhat selectively on fixing blood cells and their progenitors directly. The new findings discussed in this Review indicate that haematopoiesis is likely to be driven as much by microenvironmental cues from the niche as by cell-intrinsic ones. There remain many open questions concerning the regulation of haematopoiesis by niche factors. Chief among them is if and how inputs from different niche components are spatially and temporally segregated from the perspective of the HSC. How HSCs influence many of the niche components discussed above is also unknown. Additionally, it is still unclear whether specific niche inputs vary in different bone marrow locations (for example, calvarial niches as compared with femoral diaphyseal niches and femoral metaphyseal niches), and how niche inputs in a given location change over time.

Despite the many unanswered questions, our new understanding of the importance of the microenvironment in haematopoiesis allows the development of novel therapeutic strategies that target the niche in order to correct haematopoietic dysfunction. These studies further highlight the limitations of therapies that are aimed only at HSCs. Already, researchers are developing innovative methods of therapeutically targeting the niche: by creating ectopic niches using biomimetic structures^{142, 143} or mesenchymal stem cells¹⁴⁴; by creating artificial niche-like environments ex vivo^{145, 146}; and by making use of agents known to modulate haematopoiesis in vivo via the niche^{147, 148}. Indeed, there are already many open clinical trials that seek to manipulate the niche in vivo or to reconstitute a synthetic niche in vitro, in the hope of effecting specific therapeutic changes within HSCs. In 2006, a multicenter Phase II clinical trial of parathyroid hormone (PTH; which stimulates bone formation and has been shown to increase HSC numbers endogenously and in transplant models¹⁴) administration to patients who received double-unit umbilical cord blood (UCB) transplantation was undertaken (ClinicalTrials.gov identifier: NCT00579722), although that trial was unfortunately halted because it reached an early stopping rule probably unrelated to the PTH therapy. Other trials using prostagalndin E2 (ClinicalTrials.gov identifier: NCT00890500) and Notch ligands (ClinicalTrials.gov identifier: NCT01175785) administered ex vivo to expand UCB stem cells before transplant are actively recruiting, as is a study in which UCB cells are co-cultured on MSCs (ClinicalTrials.gov identifier: NCT00498316). A trial using StemRegenin 1⁷¹ to expand UCB grafts will open soon, and medications like AMD3100 (ClinicalTrials.gov identifier: NCT00075335) and hepatocytes growth factor¹⁴⁹ are poised to join G-CSF and GM-CSF in the small but potent arsenal of HSPC mobilizing agents. As these therapies move into clinical practice, however, it will be increasingly important to understand not only how they affect haematopoiesis but also how they affect the haematopoietic niche. Understanding the composition of the niche and the complex interactions between niche elements and HSCs has never been more critical or timely. Only through a deeper understanding of this intricate support system can we truly learn how to engineer haematopoiesis to combat haematopoietic disease.