Induction of developmental hematopoiesis mediated by transcription factors and the hematopoietic microenvironment

Abstract

The induction of hematopoiesis in various cell types via transcription factor (TF) reprogramming has been demonstrated by several strategies. The eventual goal of these approaches is to generate a product for unmet needs in hematopoietic cell transplantation therapies. The most successful strategies hew closely to clues provided from developmental hematopoiesis in terms of factor expression and environmental cues. In this review, we aim to synthesize the TFs that play important roles in developmental hematopoiesis primarily and to also touch on adult hematopoiesis. Several aspects of cellular and molecular biology coalesce in this process, with TFs and surrounding cellular signals playing a major role in the overall development of the hematopoietic lineage. We attempt to put these elements into the context of reprogramming and highlight their roles.

Graphical abstract:

In this review, we aim to synthesize the TFs that play important roles in developmental hematopoiesis primarily and to also touch on adult hematopoiesis. Several aspects of cellular and molecular biology coalesce in this process, with TFs and surrounding cellular signals playing a major role in the overall development of the hematopoietic lineage. We attempt to put these elements into the context of reprogramming and highlight their roles.

Introduction

HSCs exist at the top of the hematopoietic hierarchy and give rise to myeloid and lymphoid progenitors. The progeny of these progenitors become increasingly lineage restricted, ultimately generating 100 billion blood cells that turnover per day in humans¹. In the setting of various hematologic malignancies, such as leukemia and lymphoma, current treatment modalities rely on allogeneic donor stem cells for the reconstitution of diseased bone marrow. This therapy is limited, however, due to the size and lack of diversity in the hematopoietic stem cell (HSC) donor pool. Further, even matched transplant recipients are at risk of developing graft-versus-host disease and graft failure due to imperfect human leukocyte antigen matching². Alternative sources of transplant material are very limited, making efforts towards generating patient-specific or haploidentical products both warranted and timely. Towards this, we have turned to known aspects of developmental hematopoiesis, with the hope of harnessing this information and applying it towards production of functional hematopoietic cells (HCs). Our initial approach was centered on the introduction of three TFs, GATA binding protein 2 (GATA2), growth factor independent 1B (GFI1B), and Fos proto-oncogene, ap-1 transcription factor subunit (FOS), in human and mouse fibroblasts^{3, 4}. Expression of these factors drives a hemogenic program that produces HSC-like cells over time. The reprogramming process appears to mimic features of the endothelial to hematopoietic transition observed during developmental hematopoiesis (recently reviewed in Ottersbach)⁵, with fibroblasts first converting to hemogenic endothelium/endothelial−like (HE) cells that then give rise to hematopoietic-like cells with prolonged culture. Our current studies show that another TF, growth factor independent 1 transcriptional receptor (GFI1), further enhances this process (Daniel et al., unpublished data). This observation is in line with the known cooperation between the TFs GATA2, GFI1, and GFI1B⁶ in hematopoiesis and the role of GFI1 and GFI1B proteins together during an endothelial to hematopoietic transition (EHT)^{7, 8}.

Together with TFs, the hematopoietic microenvironment plays a major role in inducing and maintaining HSCs as they emerge from the aorta-gonad-mesonephros (AGM). As HSCs mature throughout development, they travel to the fetal liver (FL) and finally to the bone marrow (BM)^{9, 10}. Microenvironmental cells impart multiple key signals required for the regulation of HSCs as they emerge and mature. This is reiterated and supported by in vitro studies demonstrating the efficacy of stromal cells in maintaining hematopoiesis^{11, 12}, and includes the mouse FL cell line AFT024^13–17. This cell line sustains HSCs ex vivo, demonstrating a capacity to be a surrogate HSC niche^{13, 14, 16, 18}. As such, the inclusion of a maturation step on AFT024 along with the inclusion of GFI1 to our reprogramming TFs further enhanced the biological activity of the reprogrammed cells.

Two recent studies have used cues from developmental hematopoiesis both in terms of inductive environments and TFs to generate repopulating HSCs. The Daley laboratory guided pluripotent stem cells along the HE pathway of development with morphogens and then added TFs to achieve HSCs¹⁹. In the second study, the Rafii group reprogramed endothelial cells (ECs) with TFs and then used an inductive endothelial support layer to generate transplantable HSCs²⁰. These studies garner hope and support for the use of reprogrammed HSCs in cell-based therapies. We refer the readers to reviews of this topic published prior to publication of these papers for more background on the topic^21–23. Herein we take a journey through developmental hematopoiesis and highlight the TFs involved at different stages, in particular, those we use in our reprogramming (GATA2, GFI1B, FOS, and GFI1) and the environmental cues that mediate the process. There are multiple mysteries that still exist, especially in human developmental hematopoiesis. Altogether, we strongly suggest that elucidation of the molecular mechanisms and the complex cellular interplay during developmental hematopoiesis is key in order to generate HSCs de novo.

TFs in Hematopoiesis

Mesodermal specification towards the hematopoietic fate

Hematopoiesis begins after mesoderm specifies towards this fate by first forming the hemangioblast as a precursor to HE. The mechanism behind this decision rests in the function of several TFs that both promote hematopoietic specification and restrict other lineage choices (Fig. 1). SCL/TAL1 acts as a major player in this decision, with a primary role in restricting the cardiac fate while promoting hemogenic competence²⁴. SCL continues to play a role across ontogeny. The formation of HE from mesoderm requires transcription factor 12 (TCF12). TCF12 deficiency in human embryonic stem cells (hESCs) results in an impairment of mesoderm formation with independent downstream effects on HE formation as well as T cell development²⁵. GATA factors, including GATA2 acts as another crucial TF in the earliest stages of hematopoiesis while necessary for primitive yolk sac hematopoiesis they are also necessary to initiate definitive blood formation within the embryo²⁶. On a more mechanistic level, GATA2 acts in a regulatory loop with SCL and FLI1, allowing for the specification and maintenance of hematopoietic fate on hemangioblasts in Xenopus. These early signals, instigated by vascular endothelial growth factor A (VEGFA) signaling as well as the TFs ETV2 and ETV6, allow for the acquisition of endothelial identity. They then migrate from the dorsal lateral plate mesoderm to form the dorsal aorta²⁷.

Figure 1.

TFs in developmental hematopoiesis. Illustration of developmental hematopoiesis from mesodermal specification through EHT to definitive hematopoiesis. Throughout these stages, various TFs work together to initiate different responses depending on the cellular and microenvironmental context. Select TFs are shown beneath each relevant stage.

Interestingly, the importance of the AP1 complex, composed of JUND and FOS proteins, in hematopoiesis has been shown. Forced coexpression of JUND and FOS in the Xenopus embryo led to an upregulation of key hematopoietic genes GATA1, SCL (TAL1), and LMO2, along with α-globin. Additionally, overexpression of JUND and FOS leads to blood cell formation in ectodermal cells and enhanced activity of SCL²⁸. A role for GFI1 in mesodermal specification towards HE has been suggested by studies in ESCs showing that a specific enhancer element is activated in Flk1⁺ mesoderm specified to the hematopoietic fate²⁹.

Developmental hematopoiesis occurs in two different forms, primitive and definitive, during vertebrate embryogenesis³⁰. Primitive hematopoiesis occurs in the extraembryonic yolk sac (YS), giving rise to transient blood populations, while definitive hematopoiesis initiates in the AGM and leads to the development of hematopoietic stem and progenitor cells (HSPCs) that eventually seed the BM for lifelong blood production. Although primitive hematopoiesis is important we will confine our discussion to definitive hematopoiesis, as this is the process HSC reprogramming studies are focused on.

Endothelial to hematopoietic transition

In 2010, three seminal papers emerged that put to rest the prolonged discussion about the existence of a specific type of endothelium that undergoes EHT, two in zebrafish^{31, 32} and one in mouse³³. All three demonstrated that hemogenic cells emerged from the endothelium in the aortic floor (Fig. 1). In vivo time-lapse imaging using an scl-isoform−specific transgenic zebrafish revealed that immediately prior to EHT, HE cells selectively express scl-β, whereas scl-α is expressed later in nascent HSCs as they egress from vascular aortic ECs³⁴. After specification to HE, these cells require additional signals to undergo EHT. Runt-related transcription factor 1 (RUNX1) plays a critical role in HE specification but becomes dispensable afterward³⁵. Upon removal of RUNX1, HE cells do not lose their endothelial identity but fail to obtain a hematopoietic fate. Additionally, these cells remain adherent, and the low numbers of cells that do separate from HE undergo apoptosis^{31, 36}. Genome-wide expression studies throughout EHT identified the transcriptional repressors GFI1 and GFI1B as direct targets of RUNX1 during this process⁷. This finding prompted rescue experiments in RUNX1-null HE cells, where overexpression of either GFI1 or GFI1B resulted in a restoration of several features of EHT. The emerging cells in these experiments expressing either of the GFI1 proteins adopted a round, nonadherent morphology and expressed hematopoietic genes while losing endothelial identity. Interestingly, these cells could not form hematopoietic colonies, indicating that GFI1 or GFI1B overexpression did not completely rescue EHT in the context of RUNX1 loss. In vivo, cells within intra-aortic hematopoietic clusters (IAHCs) and ECs in the AGM express GFI1, whereas fully formed IAHCs expressed GFI1B. This confirms the association of GFI1 proteins with active emerging hematopoiesis. Additionally, transplantation of GFI1⁺ or GFI1B⁺ cells from E11.5 AGM results in long-term reconstitution indicating that GFI1 proteins distinguish HE from other endothelial subsets⁸.

The effects of RUNX1 and GFI1 proteins coalesce into a role for Notch homolog, translocation-associated (Drosophila) (NOTCH). The AGM region requires the expression of NOTCH1 for initial HSC generation in the switch from endothelial to hematopoietic fate during EHT. NOTCH function requires RBPJ-binding sites found in a variety of genes bound by RUNX1 and the GFI1 proteins, demonstrating a complex interplay between NOTCH and these TFs for EHT⁸. Studies support the EHT model of RUNX1 expression occurring first in HE, which then induces the expression of GFI1 and GFI1B. These two proteins then go on to bind genes involved in the maintenance of endothelial identity and epigenetically silence them, subsequently permitting the generation of round, nonadherent HCs. The function of NOTCH specifically in this process, however, remains unknown. After EHT, RUNX1 becomes dispensable, necessitating other TFs to maintain this identity³⁵. To further explain the impact of NOTCH levels in HSC emergence, the role of FOS comes into play. It was found that another TF, nuclear receptor corepressor 2 (NCOR2), participates in HSC development through negative regulation of FOS. Knockdown of NCOR2 abrogates its ability to cooperate with histone deacetylase 3 (HDAC3) leading to an upregulation of FOS, which then induces the expression of VEGFRD to repress HSC emergence through increased NOTCH signaling³⁷. This highlights the importance of variable levels of NOTCH in hematopoiesis.

GATA2 also plays a clear role at this stage; GATA2-null mouse embryos possess very few HSPCs due to a defect in the AGM^{38, 39}. Deletion of GATA2 in mice leads to death at E10 due to a broad collapse of hematopoiesis, demonstrating the requirement of GATA2 for HSPC generation and function^{40, 41}. In GATA2-null hESCs, differentiated cells display disrupted EHT through a significant reduction in the production of CD34⁺CD43⁺ hematopoietic progenitor cells. Interestingly, these hESCs could still produce erythroblasts and macrophages upon differentiation, but could never generate granulocytes⁴². Some conflicting evidence regarding the role of GATA2 in specification of HE also exists. A GATA2-null hESC line carrying a DOX-inducible GATA2 transgene demonstrated a crucial role of GATA2 in primitive and definitive hematopoiesis. They questioned, however, the role of GATA2 in specifying HE, playing a role in EHT, or acting in both programs. Using this hESC system, they found that early activation of GATA2 during differentiation (therefore, during HE specification) suppressed HE formation, whereas inhibition of GATA2 after HE specification resulted in a significant reduction in blood production. These results indicate that in hESCs, GATA2 may not act during HE specification, but plays a major role during EHT⁴³.

Definitive hematopoiesis

After leaving the AGM, definitive HSCs are found in the placenta, expand in the fetal liver and then seed the BM, wherein a variety of TFs act to maintain HSC identity and function (Fig. 1). Key TFs from EHT—GFI1 and GFI1B—reappear in adult hematopoiesis. HSCs highly express both GFI1 and GFI1B⁴⁴. GFI1 also appears in granulocyte-macrophage progenitors, B cells, granulocytes, and immature T lymphocytes whereas GFI1B is expressed in erythroid and megakaryocytic cells^{45, 46}. GATA2 presents itself as a major player in adult HSCs. Single-cell expression analysis of 597 primary HSPCs from mouse BM allowed for the description of a gene regulatory network of 18 key hematopoietic TFs. A regulatory triad is imbedded in this network composed of GATA2, GFI1, and GFI1B. Within this triad, GATA2 functions to regulate cross inhibition between GFI1 and GFI1B during entry into the myeloid and lymphoid lineages⁶.

Other work confirms the importance of GATA2 in regulating these and other positive effectors of hematopoiesis. The same group that identified the regulatory triad of SCL, GATA2, and FLI1 in developmental hematopoiesis later described a “heptad” of TFs consisting of SCL, GATA2, FLI1, RUNX1, LYL1, ERG, and LMO2. These factors function together in HSPCs as well as in intermediates to regulate differentiation into erythroblasts and megakaryocytes^47–49. Taken together our reprogramming factors GATA2, GFI1, and GFI1B have major roles in adult hematopoiesis that highlights their cooperative roles in determining cell fate.

The involvement of FOS in adult hematopoiesis remains less clear compared to the previously discussed TFs. A study evaluating the transcriptional profiles of HSCs during development found that the definitive HSC population proliferating in the FL downregulates FOS and FOSB, which correlates with prior studies suggesting that FOS proteins act to regulate entry of HSCs into the cell cycle^{50, 51}. FOS was also identified in an HSC gain-of-function screen using repopulation capacity as a primary output that identified a panel of 18 nuclear factors involved in governing HSC fate further reinforcing its role in HSCs⁵².

Altogether, key TFs have been identified at various stages of hematopoietic development. They may play different roles at different stages but remain important throughout. Among these factors are those we are particularly interested in to further our reprogramming studies, these being GATA2, GFI1, GFI1B, and FOS.

The hematopoietic microenvironment throughout development

HSCs notoriously die or differentiate in culture, severely limiting the ability to expand or manipulate these cells in vitro for clinical or research purposes. This problem likely stems from an incomplete understanding of the in vivo microenvironment that supports and matures these cells. As proposed by Schofield in 1978, HSCs reside in sites of active hematopoiesis and remain in contact with other cells in a “stem cell niche.”⁵³ One year prior, a seminal paper was published demonstrating that BM stromal cells promoted the ex vivo proliferation and differentiation of HCs in what they termed long-term cultures (LTCs)⁵⁴. These supporting cells, and other identified niche cells, play a significant role in determining the behavior of the stem cell and whether or not it decides to self-renew or differentiate. This interaction exists through cell-cell, cell-extracellular matrix, and receptor–ligand interactions between the HSC and the variety of cells and other factors in the niche^55–57. It follows that the variety of cells identified and cell lines generated from different areas of development that support HSCs highlight the complexity of this system (Fig. 2).

Figure 2.

Schematic of definitive hematopoietic niches from embryo to adult. The different sites of mouse and human hematopoiesis throughout development are displayed. Hematopoietic progenitors are known to emerge from the AGM and seed different regions of the developing embryo. These include the placenta and FL where they undergo self-renewal expansion primarily in the FL. Cellular elements from these spaces have been isolated primarily as stromal cell lines and characterized for their ability to support HSCs. Maturing HSCs eventually seed the BM where they reside throughout adult hematopoiesis. Throughout their journey HSCs are exposed to a multitude of signals that promote their expansion, maintenance, or retention. The BM hematopoietic niche has undergone the most extensive investigation both in terms of characterization of cellular elements and the signaling that these elements elaborate and mediate.

The BM acts as the major site of hematopoiesis in most adult vertebrates, but several other sites throughout development support HSCs as they mature. During embryonic development, HSCs travel from the AGM and placenta to the FL where they undergo their greatest period of self-renewal expansion to eventually reside in the BM⁵⁸. Given the heterogeneity of cells types in these sites, it has been a long-standing challenge to recapitulate this system in vitro.

The AGM

The AGM develops from paraaortic splanchnopleura (P-Sp) and mesoderm. Definitive hematopoiesis is initiated there at E10 in the mouse⁵⁹. Although de novo HSCs emerge in the AGM they remain there for only a short time and are largely gone by E12. Additionally, the AGM generates a limited number of HSCs in this region but they can be produced and expanded using an organ culture approach⁶⁰. Delta-like 1 (DLK1), a known positive regulator of HSCs in the FL^61–64 appears to function as a negative regulator in the smooth muscle of the dorsal aorta. In a direct cell contact manner, DLK1 limits HSC expansion in the AGM⁶⁵.

After discovery of the anatomical birthplace of definitive HSCs attempts to recapitulate this environment in vitro to maintain and expand HSCs have been undertaken. Early work derived two immortalized endothelial lines from day 11 murine CD34⁺ cells from the dorsal aorta. Interestingly, one of the lines supported the expansion of LSK CD34⁺ FL cells while the other induced the differentiation of these cells into erythroid, myeloid, and B lymphoid cells. Additionally, the LSK CD34⁺ cells required direct contact with the derived ECs for this induction⁶⁶. Additional studies of 100 stromal cell lines generated from cells isolated from parts of the AGM revealed that the dorsal aortic mesenchyme as well as the urogenital ridge would act as potent microenvironments for HSC growth in vitro⁶⁷. Further evidence of the utility of AGM cells as a supportive HSC niche comes from studies of YS-derived HCs. Most studies agree that the YS does not generate definitive HSCs but the question remained as to whether or not cells from this area possess the potential to obtain definitive HSC function. Stromal cells derived from E10.5 AGM were seeded HCs from E8.0 YS or intraembryonic P-Sp. After 4 days, the seeded cells could reconstitute definitive hematopoiesis in transplanted mice⁶⁸. Although these results are striking they have yet to be confirmed in other studies. These cells also permitted the expansion of human CD34⁺CD38⁻ progenitors 20-fold when supplemented with the cytokines stem cell factor (SCF) and thrombopoietin (TPO),⁶⁹ further reinforcing the inductive power of the AGM microenvironment.

In attempts to elucidate the molecular signals provided by AGM-derived stromal cells that support hematopoietic maturation in vitro, Wnt5a was identified as a key factor responsible for preserving HSCs⁷⁰. Supplementation of Wnt5a to nonsupportive cultures restored maintenance of HSCs in vitro, and antibodies against Wnt5a compromised HSC maintenance. Additional work supported this finding (albeit in adult BM) by demonstrating that Wnt signaling in general promotes quiescence and self-renewal in vivo^{71, 72}. AGM-derived primary ECs with forced AKT expression promote HSC induction and self-renewal ex vivo. These cells also express various NOTCH ligands, including JAG1, JAG2, DLL1, and DLL4, and can induce HSC generation from VE-cadherin⁺ precursors derived from E9.0–E10.0 HE⁷³. These endothelial cell co-cultures also amplify HSCs from E11 AGM-derived Ve-Cadherin⁺CD45⁺ HCs. These cells reproduce the vascular niche for developing HSCs, a known niche proven to support HSCs in vitro^{20, 74}.

The placenta

The small numbers of definitive HSCs formed in the AGM begged the question of whether some other site existed that also generated definitive HSCs de novo. The avian allantois, a known mesodermal appendage in avian embryos serves as a site of de novo hematopoiesis in birds^{75, 76}. In the mouse, the allantois forms both the umbilical cord and the mesodermal components of the fetal placenta after fusion with the chorion⁷⁷. This fusion generates the placental labyrinth, which consists of endothelial cell−lined fetal capillaries and trophoblast-lined maternal blood sinuses⁷⁸. In 2003, the Dieterlen-Lievre laboratory found via in vitro clonogenic assays that the placenta served as a rich source of multipotent hematopoietic progenitors suggesting that the placenta should be considered a hematopoietic organ⁷⁹.

Using long-term transplantation assays to assess a spatial and temporal analysis of HSC activity during mouse embryonic development, it was determined that the placenta contributes to definitive hematopoiesis as early as E10.5, similar to the AGM^{77, 80}. At this time point, however, the emergence of functional HSCs begins at low levels but becomes significantly more robust at E11.5. Since circulation is established prior to E10.5 it is not possible to unequivocally state that definitive hematopoiesis arises de novo in the placenta. Nevertheless, the placenta is now recognized as a major hematopoietic organ during development in both mouse and human⁸¹. Interestingly, GATA2 and FOS along with RUNX1 were found to be particularly important in the mouse placenta in both endothelial and HCs⁸⁰. The lack of hemogenic precursor−specific markers hinders the ability to fully study the earliest stages of emerging hematopoiesis. To this end, our group identified a Prom1⁺Sca1⁺CD34⁺CD45⁻ (PS34) surface phenotype in developing reprogrammed cells that possess both endothelial and hematopoietic gene profiles³. Using this phenotype, systematic dissection of the hematopoietic organs of the developing mouse embryo revealed these cells both in the AGM and placenta⁸². In the placenta, PS34 cells originate from the fetus, localize to the vascular labyrinth and interestingly remain restricted to the maternal–fetal interface, agreeing with previous studies regarding HSC isolation from placentas^{77, 80, 83}. These cells also possess endothelial and hematopoietic gene profiles, and require a NOTCH signal before generating transplantable HCs. Interestingly, the NOTCH signal in this study comes from OP9-DL1 cells (BM stromal cells derived from newborn B6C3F1-op/op mouse calvaria that genetically lack macrophage colony−stimulating factor (M-CSF) that were stably transduced with a delta-like canonical NOTCH ligand 1 (DLL1) expression construct⁸⁴), highlighting the interplay between multiple cell types across hematopoietic organs while still supporting HSC function⁸⁵.

To molecularly dissect the placental niche, endothelial and mesenchymal cells were isolated to specifically investigate hematopoietic cytokine gene expression. It was confirmed that placental hematopoietic cell clusters originate from the allantois, rely on regulation from endothelial niche cells, and require SCF signaling in this region for proper maturation and function⁸⁶. Another group further confirmed the presence of CD34⁺ HCs with multilineage potential from either fresh or cryopreserved tissue⁸⁷. Staining for HSC markers Sca1, CD34, and CD31 marked ECs within the placental labyrinth, further suggesting that these cells act as at least one source of HSCs⁸⁰ and provide a niche where HSCs expand primarily through SCF stimulation^{86, 88, 89}.

With the identification of HSCs in the placenta, others sought to identify placenta-derived stroma to culture HSCs and subsequent blood products in vitro. The role of stromal cells derived from mesenchymal cells in human placenta was investigated^{90, 91}. These cells lack the expression of hematopoietic and endothelial cell markers. Similar to BM-derived mesenchymal stromal cells (MSCs), placenta-derived cells suppress T cell proliferation, highlighting their similarities to the commonly used BM cells⁹². CD34⁺ enriched cord blood (CB) cells coinjected with placenta-derived adherent stromal cells engrafted at much higher efficiencies than without these stroma, demonstrating a positive effect of placenta-derived products for transplantation⁹³. The same group that established the presence of HSCs in the human placenta also demonstrated that human placenta-derived cell lines support human hematopoietic progenitors in vitro⁸¹. Interestingly, these cell lines formed tubules in culture indicative of endothelial potential, demonstrating the mesenchymal potential of these cells similar to previously described AGM hematopoietic-supportive stromal lines⁹⁴.

The fetal liver

Interestingly, the FL does not appear to generate HSCs de novo, but rather becomes seeded by circulating HCs^{95, 96}. Following the generation of definitive HSCs in the AGM, the FL becomes the unique niche for HSC expansion without loss of self-renewal or differentiation capacity. This fact suggests that the FL forms a conducive microenvironment to support HSCs, as these cells do not lose their stem cell qualities even though they rapidly cycle at this stage⁹⁷. In the mouse, HSCs begin to migrate to the FL on E11.5. Between E12.5 and E16.5, they undergo the aforementioned massive expansion but also differentiate to form a large pool of hematopoietic progenitors⁵⁸. In these 4 days, the competitive repopulating units (CRUs) magnify 38-fold and subsequently migrate to the BM where they become quiescent⁹⁸. This expanding HSC pool can outcompete HSCs derived from the BM when transplanted into irradiated recipients^99–101. It was hypothesized that rare stromal elements present in the FL would mediate this expansion and if found would support HSC in vitro.

AFT024 came out of an extensive study that established over 200 immortalized stromal cell lines from day 14.5 murine FL. This line supported enriched HSC in LTC for 4−7 weeks that were capable of repopulating transplanted mice at efficiencies equal to freshly purified cells¹³. This clearly indicates that AFT024 provides a milieu necessary for stimulating HSPC proliferation without losing repopulating activity, a hallmark feature of the FL niche. Upon identification of AFT024, multiple follow-up studies demonstrated the marked utility of these cells. AFT024 promoted significantly greater cell expansion of human CD34⁺CD38⁻ progenitors compared with primary human stroma, and maintained the phenotype and function of these human hematopoietic progenitors¹⁴. The line could also support human progenitors for at least 2−3 weeks ex vivo, and permit superior levels of long-term multilineage engraftment compared with cultures using postnatal human BM stromal cells¹⁶. Given the success of this cell line to support and maintain both mouse and human HSCs for several weeks in vitro, attempts to identify what signals AFT024 provides to generate such an effect were undertaken. A noncontact culture system demonstrated the importance of direct cell–cell contact of purified progenitors on AFT024. Few LTC-initiating cell (LTC-IC) colonies emerged from noncontact cultures unless human cytokines were supplemented to the cultures. Additional studies showed that AFT024 expressed cytokines, growth factors, and glycosaminoglycans (GAGs) that are needed to support hematopoiesis¹⁰².

Subtractive hybridization of cDNA clones between AFT024 and a nonsupportive FL cell line identified DLK1 as a critical molecule preferentially expressed by AFT024 cells that serves as a positive stem cell regulator^{18, 61}. Overexpression of DLK1 in a nonsupportive stromal cell line allowed significant enhancement of HSC reminiscent cobblestone-like colony formation. This molecule appears to require contact cultures, as soluble DLK1 in semisolid media failed to affect the colony-forming potential of purified progenitors⁶¹. These findings are in contrast to studies that found DLK1 to be a negative regulator of HSPCs in the AGM⁶⁵. This phenomenon could simply be due to the nature of the tissue and the complexity and requirements of the HSC niches at different times of ontogeny. DLK1⁺ fetal hepatic progenitors supported a small increase in HSC numbers and a large expansion of hematopoietic progenitors^{62, 63}. In this system, only cultures in direct cell–cell contact demonstrated long-term expansion. These cocultures, which utilized serum-free, low-cytokine conditions, expanded HSCs sevenfold. Another study found that Nestin⁺ stromal cells in the FL reside in close proximity to portal vessels, and express mesenchymal lineage markers as well as DLK1⁶⁴. These cells highly express HSC niche and expansion factors, further implicating DLK1⁺ cells as putative HSC niche cells in the FL.

The bone marrow

The BM acts as the major site of hematopoiesis throughout adult life. The bone cavity can be subdivided into endosteal, subendosteal, vascular/perivascular, and central marrow regions¹⁰³. These regions are composed of a myriad of cells that form complex and heterogeneous structures, so-called BM niches, that support and balance self-renewal and cell fate decisions of HSCs and their progeny. Early studies indicated osteoblasts and other bone lining cells in the endosteal zone of trabecular bone as important components of the HSC niche. Initially, osteoblasts were shown to expand HSCs in vitro^{104, 105}. Subsequent studies, using a pharmacological or genetic expansion of osteoblastic cells, demonstrated an increase in the number of HSCs in the BM^{106, 107}. In contrast, ablation of osteoblasts resulted in a significant reduction in the number of a highly quiescent HSC subpopulation, despite only a modest increase in overall HSC and progenitor cell numbers, which can be attributed to cell cycle entry for proliferation and differentiation of these highly quiescent HSCs¹⁰⁸. Other bone lining cells, known as osteolineage cells, have been shown to be important in the regulation of HSC quiescence through the expression of angiogenin and the adhesion molecule embigin^{109, 110}.

Vascular/perivascular components of the BM niche encompass ECs, mesenchymal-derived perivascular stromal cells, Nestin⁺ MSCs, NG2⁺ pericytes, sinusoid-associated macrophages, and nonmyelinating Schwann cells. ECs are major sources of cytokines, such as C-X-C motif chemokine 12 (CXCL12) and SCF, that control HSC function^111–114. ECs also express the angiocrine factor Jagged-1 through which they control homeostatic and regenerative hematopoiesis in a Notch-dependent manner^{115, 116}. The importance of ECs in HSC regulation is further supported by data demonstrating that ECs boost the formation of hematopoietic multipotent progenitors from pluripotent stem cells¹¹⁷, promote the reprogramming of human and mouse ECs to HCs^{20, 74}, and support the ex vivo expansion of BM-repopulating HSPCs^{88, 118, 119}. Perivascular stromal cells, also known as CXCL12-abundant reticular (CAR) cells, are also an active source of the known HSC-regulating cytokines SCF and CXCL12^{111, 112}. NG2⁺ pericytes ensheathe arterioles that are associated with quiescent HSCs^120–122. Ablation of NG2⁺ pericytes as well as NG2⁺ pericyte–specific deletion of CXCL12 resulted in relocation of HSCs away from arterioles and loss of HSC quiescence^{120, 122}. Imaging analysis revealed that about one-third of HSCs are located adjacent to sinusoid-associated macrophages in the BM^{123, 124}. In vivo ablation of megakaryocytes resulted in a substantial reduction in the number of HSCs caused by loss of quiescence and cell cycle entry ultimately leading to HSC exhaustion^{123, 124}. Megakaryocytes secrete CXLC4/PF4 and the biologically active form of transforming growth factor-beta (TGF-β), and express the cell membrane protein CLEC2^123–125. Deletion of CXLC4 in the BM resulted in an increased number of HSCs and increased HSC proliferation¹²³. Ablation of megakaryocytes led to a reduction in biologically active TGF-β1 in BM and nuclear-localized phosphorylated SMAD2/3 in HSCs, and megakaryocyte-specific deletion of TGF-β1 resulted in increased HSC activation and proliferation¹²⁴. Megakaryocyte-specific deletion of CLEC2 led to a reduction in BM HSC quiescence and repopulation potential, along with extramedullary hematopoiesis¹²⁵. This decline in HSC potential may be a consequence of reduced THPO and/or CXLC4 levels.

Very recently, the Frenette laboratory has shown that Nestin⁺ MSCs rapidly lose an HSC supportive capacity in vitro due to loss of expression of niche factors, such as SCF and CXCL12, can that this activity can be restored by the expression of five different TFs (KLF7, Ostf1, Xbp1, Irf3, and Irf7)¹²⁶. These factors were isolated by an elegant screen of 28 factors highly expressed in freshly isolated Nestin⁺ stroma. They went on to show by transplantation assays that reprogrammed MSCs would expand repopulating HSCs by sevenfold. These studies suggest that reprogramming can work both ways in the hematopoietic system and aid in regenerative clinical applications¹²⁶.

The sympathetic nervous system is implicated in the maintenance of HSCs in the BM through CXCL12 and TGF-β.^127–129 Sympathetic nerves sit atop arteries and arterioles and are ensheathed by nonmyelinating Schwann cells, which are a major source of the active form of TGF-β¹²⁷. Surgical denervation of sympathetic nerves reduced the number of active TGF-β–producing cells and led to a rapid reduction in HSC numbers in the BM, supporting the notion that nonmyelinating Schwann cells are an important component of the HSC BM niche¹²⁷.

Additionally, nonclassical niche cells that regulate HSC behavior include macrophages, neutrophils, and other myeloid cells, and adipocytes. Macrophages are found throughout the BM. Trophic endosteal macrophages, known as osteomacs, are a specific subset of macrophages that localize to the endosteal surface and support osteoblast function and act on perivascular stromal cells. Ablation of macrophages resulted in the loss of endosteal osteoblasts, a strong decrease in HSC-trophic cytokine levels, such as CXLC12 and ANGPT1, at the endosteum, and HSC mobilization from the BM^130–132. Additional support for the role of macrophages in HSC regulation comes from a report, which showed that macrophage-specific interferon-gamma (IFN-γ) signaling during infection is responsible for loss of HSC quiescence resulting in a reduced HSC pool¹³³. Moreover, a recent paper implicated that aged BM macrophages with decreased phagocytic capacity have an instructive role together with the inflammatory cytokine interleukin (IL)-1β in age-associated lineage skewing of HSC towards platelets¹³⁴. The clearance of neutrophils, which egress from and circulate back to the BM by macrophages via phagocytosis, has been shown to rhythmically modulate the HSC niche and thereby control HSPC egress in steady-state hematopoiesis¹³⁵. Chen and colleagues recently demonstrated that histamine production is restricted to myeloid cells and some myeloid-biased HSCs¹³⁶. These histamine-producing cells were found immediately adjacent to CXCL12-producing perivascular cells in the BM. Mice incapable of producing histamine showed increased HSC proliferation and myeloid colony expansion. HSCs from these mice or from mice deficient in the histamine receptor HRH2 showed a deficiency in the reconstitution of all lineages, particularly in long-term myeloid reconstitution upon competitive transplantation¹³⁶. Adipocytes have been implicated to play a role in the regulation of HSC behavior, albeit clarity is lacking. Within adipocyte-rich BM, a reduction in number, frequency, and cycling capacity of HSCs during homeostasis was reported¹³⁷. In addition, “fatless” mice, which were genetically or pharmacologically rendered incapable of adipocyte formation, demonstrated enhanced BM engraftment after irradiation and transplantation¹³⁷. A subsequent study, however, found no effect of increased BM adipocyte numbers on HSC maintenance under homeostatic conditions¹³⁸. Two more recent studies do implicate a role for adipocytes in regulating HSC behavior in post-ionizing radiation or post-chemotherapy conditions via Adiponectin and/or SCF production but not in the steady state^{139, 140}.

Altogether, it is obvious how complicated the BM space within which HSCs live is structured (Fig. 2). In the last two decades, it has been systematically and intricately studied. A myriad of cell types have been identified and the signaling systems they employ are being elucidated. This bounty of knowledge will inevitably lead to a more complete understanding of how to make, expand, and maintain an HSC.

Concluding remarks

The plethora of knowledge that has been gained over the last 20 years has contributed to an enhanced understanding of the cellular and molecular nature of hematopoiesis. Key strides have been made towards revealing the processes of developmental hematopoiesis and the hematopoietic microenvironment. We and others have begun to capitalize on them in attempts to develop HSCs from somatic and pluripotent cells. The goal is to generate HSCs for stem cell transplantation. But there are multiple other applications; the ability to generate different blood products, to provide platforms for drug screening, to create stocks of rare haplotypes for transplantation, and to develop platforms for genetic correction of disease, to name a few. As with all good science, it is a back-and-forth process, one technology informs another. Altogether, this will lead us to the successful application of reprogramming technologies to the treatment of hematopoietic diseases.