Main

Defining the embryonic origins of specific cell lineages is important for understanding how tissues of the adult organism develop. The signaling events that induce the molecular programs governing lineage-specific fate 'decisions' in embryonic cells provide insight into the complexity of lineage relationships, cell diversity and, ultimately, tissue function in the adult. The process of blood cell development in the mammalian conceptus is particularly complex, as it occurs in many sites that are separated both temporally and spatially. Furthermore, unlike stationary tissues, cells of the hematopoietic system circulate and thus their ancestry and the distinct characteristics associated with their site of origin are confounded by the natural mobility of the system. Studies have begun to show the lineage relationships between and molecular programs controlling hematopoietic cell emergence in the conceptus and the legacy of the cells emerging from distinct anatomic sites. This review focuses on the embryonic origins of the hematopoietic system as well as the environments and molecules affecting the development of adult mammalian hematopoietic stem cells.

Sites and cells: where does it start?

The conceptus consists of embryonic tissues that will ultimately become part of the fetus, as well as extra-embryonic tissues that support fetal development. It has been long recognized that the first blood cells in the vertebrate conceptus appear in the extra-embryonic yolk sac concomitant with the developing vasculature. The yolk sac of early chick embryos has been shown by histological studies to contain the first visible hematopoietic cells, primitive erythrocytes1. The close physical association of primitive erythrocytes and their synchronous appearance with endothelial cells led to the postulation of a common mesodermal precursor for these two lineages called the 'hemangioblast'2. Studies using the in vitro differentiation of totipotent mouse embryonic stem cells produced the first functional evidence of mammalian hemangioblasts3,4, and subsequent analyses of early stage mouse conceptuses showed presumptive hemangioblasts expressing both the mesodermal marker Brachyury and fetal liver kinase 1 in the posterior region of the primitive streak5. These hemangioblasts migrate to the yolk sac, at which point they become committed endothelial and hematopoietic progenitors (Fig. 1a, left), several of which contribute to the formation of each blood island6,7. Thus, during mammalian embryonic development, the earliest cohort of mesodermal cells emigrating from the primitive streak take on endothelial and hematopoietic fate before blood island formation and give rise to primitive red blood cells and some of the yolk sac vasculature. The remainder of the yolk sac vasculature is derived from angioblasts that also emerge from the posterior primitive streak and do not contribute to blood8.

Figure 1: Vertebrate hematopoietic development.
figure 1

(a) Mesodermal migration during the early-streak stage (left) and mid- to late-streak stage (right) in the mouse conceptus. In the early-streak stage, mesoderm emerging from the primitive streak forms the extra-embryonic yolk sac and, slightly later, the allantois. At the mid- to late-streak stage, mesoderm emerging from the anterior primitive streak forms the axial, paraxial and lateral mesoderm in the rostral region of the embryo. The mesoderm from the posterior primitive streak forms the paraxial and lateral mesoderm of the remaining trunk region of the embryo. Red arrows indicate the emigration of mesoderm after egression from the primitive streak. (b) Interspecies and intraspecies grafting in early precirculation avian embryos (such as a quail embryo body grafted onto a chick yolk sac) identifies the origins of adult blood as the embryo body and not the yolk sac. (c) After genetic marking of amphibian embryos at the 32-cell stage (left), the progeny are traced to larval stages (right). The C3 blastomere gives rise to the DLP mesoderm and more specifically to the dorsal aorta and the hematopoietic clusters in the lumen. The D4 blastomere gives rise to the posterior VBI (pVBI) and the C1 and D1 blastomeres give rise to the anterior VBI (aVBI). Drawings are adapted from refs. 11,15 and the Edinburgh mouse atlas project website.

Full size image

It was suggested in the 1970s that cells of the yolk sac are the source of the hematopoietic system in the adult mammal and that yolk sac cells emigrate to the fetal liver and thereafter to the bone marrow, where they reside throughout adulthood9,10. However, tissue-grafting approaches have conclusively demonstrated that the yolk sac is not the source of adult blood in nonmammalian vertebrates11,12 (Fig. 1b, c). Interspecies and intraspecies grafting of avian embryo body and yolk sac (as well as the allantois), before the emergence of blood cells and the onset of circulation, has shown that the adult hematopoietic system originates from cells in the body of the embryo and from the extra-embryonic allantois, and not from the yolk sac11,13,14 (Fig. 1b). Likewise, orthotopic grafting of the ventral blood island (VBI; yolk sac equivalent) and dorsal lateral plate (DLP; intrabody region) in amphibian embryos has shown that the DLP is the source of adult hematopoietic cells12. Molecular marking experiments in Xenopus laevis embryos at the early cleavage stages has shown that the separation of presumptive VBI and DLP mesoderm occurs very early15 (Fig. 1c). The progeny of the C3 blastomere of regularly cleaving embryos contribute to intraembryonic (DLP) hematopoiesis, whereas completely distinct blastomeres contribute to the anterior (C1 and D1) and posterior (D4) VBI. Separation at this early stage indicates that the progeny of these cells experience spatially and temporally distinct signaling events during gastrulation and embryonic patterning. Indeed, the three blood compartments in X. laevis (anterior VBI, posterior VBI and DLP) are specified from mesoderm that encounters different amounts of bone morphogenic protein (BMP)16. In addition, the morphogen fibroblast growth factor (FGF) differentially affects the timing of expression of TAL1 and RUNX1, which encode pivotal transcription factors (SCL and Runx1, respectively) that specify hematopoietic fate in the anterior versus posterior VBI17. Notably, by grafting VBI or DLP cells to the reciprocal site in the amphibian embryo, the prospective hematopoietic cells can be 'reprogrammed' to the alternative adult or primitive hematopoietic fate. However, this is possible only in an early developmental window of time18, after which primitive and adult states are thought to be epigenetically fixed.

Evidence for the separation of presumptive mesoderm in the mouse epiblast has been provided by single-cell marking experiments. In experiments labeling prospective mesodermal cells at early primitive streak stages (at embryonic day 6.5 (E6.5) to E7), cells from the posterior primitive streak were found to contribute to extra-embryonic hematopoietic tissue, that is, yolk sac and allantois19 (Fig. 1a, left), and not to the body of the embryo. Those results have been confirmed with an orthotopic grafting method8 that further showed that at mid-streak stages, the mesoderm derivatives in the rostral embryo arise from epiblast cells that ingress through the anterior primitive streak before the cells that give rise to axial, paraxial and lateral (blood-forming) mesoderm of the anterior trunk (Fig. 1a, right). Finally, the mesoderm emigrating from more caudal regions of the streak forms the paraxial and lateral mesoderm of the remaining trunk regions8 (Fig. 1a, right). Notably, the entire epiblast of embryos of the early- and mid-streak stage contains hemogenic potential, but that potential thereafter becomes restricted to the trunk and posterior region of the embryo20. Thus, in notable similarity to the mesodermal fate map of the chick8, presumptive extra-embryonic blood-forming mesoderm in the mouse conceptus is the first to emerge from the primitive streak during gastrulation, followed by the intraembryonic mesoderm in a rostral-to-caudal sequence.

From primitive erythrocytes to adult hematopoietic stem cells

In vivo transplantation assays in mammals have provided great insight into the specific cells that can generate a complete functional adult hematopoietic system. Hematopoietic stem cells (HSCs) are at the foundation of the adult blood differentiation hierarchy and provide continuous hematopoietic cell production throughout life. Transplantation of cells from various regions of the E8–E12 mouse conceptus has shown that HSCs conferring complete, long-term, multilineage, substantial hematopoietic repopulation of irradiated adult recipient mice appear beginning only at E10.5 in the aorta-gonad-mesonephros (AGM) region of the embryo body and in the vitelline and umbilical arteries21,22,23 (Fig. 2, right). The appearance of HSCs 3 days after primitive erythrocytes are generated in the yolk sac (Table 1 and Fig. 3) makes it unlikely that they share a recent common ancestry. These adult repopulating HSCs (which are as potent as adult bone marrow HSCs) are autonomously generated in the AGM, as shown by explant cultures21, and are located in the ventral aspect of the dorsal aorta24,25,26. Slightly thereafter, HSCs are found in other tissues, such as the placenta, yolk sac and liver21,22,27,28. The liver does not generate hematopoietic cells de novo but is instead colonized beginning at late E9 by hematopoietic cells made in other tissues29,30 (Fig. 3). The identification of the yolk sac and placenta as de novo generators of HSCs is precluded by the circulation (which is established at the four- to six-somite-pair stage, approximately E8.25–E8.5)31, which easily distributes HSCs throughout the conceptus. However, quantitative spatial and temporal analyses of HSCs suggest that the yolk sac32 and placenta27,28 do contribute to the HSC pool in the liver either through the population expansion of pre-existing HSCs or by de novo generation of HSCs. It has been shown that there are more HSCs in the fetal liver than can be accounted for by HSC generation in the AGM alone32. The additive production of HSCs by AGM, yolk sac and placenta27 is most likely responsible for the large numbers of fetal liver HSCs, although the liver itself may expand these cell populations33. Thus, within a span of 3 days, the mouse conceptus generates at least two very distinct and unrelated classes of functional hematopoietic cells: primitive erythrocytes at E7.5 and definitive adult repopulating HSCs at E10.5.

Figure 2: The generation of hematopoietic cells in the mouse conceptus.
figure 2

At least five classes of hematopoietic cells (arrows), as defined by function, are progressively generated in the mouse conceptus at E7.5, E8.25, E9.0 and E10.5 (left to right). The primitive class is derived from hemangioblasts, and the pro-, meso-, meta- and adult-definitive classes are thought to be derived from specialized vascular cells called 'hemogenic endothelium'. Some of these cells are derived from distinct mesodermal lineages emigrating from the primitive streak (Fig. 1). The E7.5 and E8.25 conceptuses show the outgrowing allantois that will fuse with the chorion to form the placenta. The circulation is established at E8.25–8.5. The E9 embryo has turned and is enveloped in the yolk sac. Colonization of the liver by hematopoietic progenitors begins at late E9. The E10.5 conceptus contains hematopoietic clusters in the dorsal aorta in the AGM region, the vitelline (V) and umbilical (U) arteries, and the first adult HSCs are found in these vessels. CFU-S, colony-forming units in the spleen. Drawings adapted from the Edinburgh mouse atlas project website.

Full size image
Table 1 Spatial and temporal appearance of hematopoietic potential and cells in the mouse conceptus
Full size table
Figure 3: Timeline of hematopoietic events in the mouse conceptus.
figure 3

Arrows above indicate the onset of specific hematopoietic cell generation and/or appearance; arrows below indicate the earliest time of colonization of the secondary hematopoietic territories.

Full size image

Other classes of hematopoietic cells are also generated and/or detected in the mouse conceptus between the time primitive erythrocytes appear and when HSCs appear, including myeloid progenitors, lymphoid-myeloid (multipotent) progenitors, progenitors capable of forming colonies on the spleens of irradiated mice (colony-forming units–spleen) and neonatal repopulating HSCs (Table 1). At E8.25, after the first wave of primitive erythropoiesis and before the circulation is established, myeloid progenitors are detected in the yolk sac and are thus generated in that site34,35. After the circulation is established, myeloid progenitors are also found in the trunk region36. However, before circulation, both the yolk sac and para-aortic splanchnopleura (pSp; prospective AGM region) contain cells with the potential to become myeloid progenitors, as shown in explant cultures37. Similar cultures of precirculation allantoises38,39 have also demonstrated cells with myeloid potential. By E9, the placenta contains an abundance of myeloid progenitors40. More potent colony-forming units–spleen are present in both the yolk sac and the AGM beginning at E9 (ref. 41). Quantitative analyses have shown that there are more colony-forming units–spleen in the AGM than in the yolk sac, which suggests that the AGM generates these cells21. Studies of two types of mutant mice, deficient in the endothelial marker VE-cadherin (Cdh5−/− mice) or deficient in solute carrier family 8, member 1 protein (Slc8a1−/− mice) have provided in vivo evidence of the de novo production of definitive myeloid progenitors in the yolk sac. Conceptuses deficient in these genes do not establish circulation between the yolk sac and embryo body; in Cdh5−/− conceptuses, there is no vascular connection, whereas in Slc8a1−/− conceptuses, the vitelline vessels are intact but there is no heartbeat to promote the circulation. E9.5 Cdh5−/− conceptuses and their respective wild-type counterparts have similar numbers of myeloid progenitors in the yolk sac, although Cdh5−/− conceptuses have fewer macrophage and mixed colony-forming progenitors42. Slc8a1−/− conceptuses have as many myeloid progenitors of all types in the yolk sac as the cumulative number of progenitors in all anatomic sites of Slc8a1+/+ conceptuses43. No progenitors are found in the Slc8a1−/− pSp, which suggests that the yolk sac normally generates all of these progenitors and distributes them to the pSp and liver. Alternatively, it is possible that Slc8a1−/− conceptuses, which lack hemodynamic stress, do not produce the proper signals to induce myeloid progenitor formation in the pSp44. Notwithstanding those potential caveats, it is apparent that several types of definitive myeloid progenitors are generated de novo in the yolk sac. It would be useful to examine the allantois in such circulation-defective embryos for the de novo generation of myeloid progenitors.

A progenitor with more complex lymphoid-myeloid potential is found before the onset of circulation in the E8 pSp-AGM of the embryo after explant culture37 (Table 1). When transplanted into irradiated immunodeficient recipients, this lymphoid-myeloid progenitor contributes to low, long-term, multilineage hematopoietic repopulation45. Yolk sac explants, in contrast, do not contain such cells until the circulation is established, which suggests that cells with lymphoid-myeloid potential are generated de novo in the pSp-AGM and could theoretically emigrate to the yolk sac. Similarly, in explant cultures with human yolk sac and pSp-AGM tissues isolated before the onset of circulation, only the pSp-AGM tissue contains cells with lymphoid-myeloid potential46.

A potent neonatal engrafting HSC has been identified in the E9 mouse yolk sac and AGM. Both tissues contain c-Kit+CD34+ cells that when injected directly in the liver of neonatal recipient mice can yield considerable multilineage engraftment47. The yolk sac contains more of these cells than does the AGM. In contrast to E10.5 AGM HSCs (also c-Kit+CD34+), which fully engraft adult recipients, the E9 c-Kit+CD34+ neonatal repopulating cells are incapable of engraftment when injected directly into adult mice47. Hence, there are at least five broad classes of hematopoietic cells in the mammalian conceptus as defined by activity in in vitro clonogenic or transplantation assays: primitive, pro-definitive (myeloid progenitors), meso-definitive (lymphoid-myeloid progenitors), meta-definitive (neonatal repopulating HSCs) and adult-definitive (adult repopulating HSCs; Fig. 2). Some of these classes of cells are generated independently of each other and in distinct anatomical sites.

Direct precursors of definitive hematopoietic cells

The 'definitive' classes of hematopoietic progenitor–stem cells shown here (Fig. 2) are thought to arise through a slightly different process than do the primitive erythroid progenitors. Discrete subsets of vascular endothelial cells seem to show hemogenic potential during development1. In many species, histological staining and immunostaining studies have shown hematopoietic clusters tightly adherent to the ventral endothelium of the dorsal aorta, as well as the vitelline and umbilical arteries48. These hematopoietic clusters seem to be emerging from the endothelium during the time the first definitive HSCs are detected. Metabolic lineage tracing (with acetylated low-density lipoprotein labeled with the fluorescent red dye DiI (AcLDL-DiI)) or retroviral labeling of endothelial cells before hematopoietic cell appearance in chick embryos has confirmed the endothelial-hematopoietic lineage relationship of aortic hematopoietic clusters49. At 1 day after labeling of endothelial cells, AcLDL-DiI+CD45+ hematopoietic cells are found in the lumen adhering to the ventral aspect of the aorta. Slightly later, the underlying aortic mesenchyme also contains AcLDL-DiI+CD45+ cells, which suggests that labeled endothelial cells ingress into the tissue. Clonal retroviral marking of the avian embryonic vasculature has provided additional data confirming the fate commitment of an endothelial cell to the hematopoietic lineage50. Similar attempts have been made in mouse embryos cultured ex utero to establish lineage relationships between endothelial cells and hematopoietic cells. At 12 hours after intracardiac injection of AcLDL-DiI into E10 mouse embryos, AcLDL-DiI–marked definitive erythroid cells were found in the circulation51. These data collectively indicate an endothelial origin for definitive hematopoietic cells.

In the midgestation mouse embryo, the phenotypic profile of adult HSCs and the spatial localization of these cells in the AGM have been useful in providing information about their direct precursors. Indeed, all HSCs in the AGM are CD45+ (ref. 25), Ly6A (Sca-1)–GFP+ (ref. 26), c-Kit+CD34+ (ref. 52), Runx1+ (ref. 25), SCL+ (ref. 53) and GATA-2+ (refs. 54, 55). These markers (with the exception of CD45) are also expressed by some or all endothelial cells in the ventral aspect of the dorsal aorta at E10–E11. Most or all AGM HSCs express cell surface VE-cadherin25,56, which is typically thought of as an endothelial marker. Whether the 'hemogenic' endothelium provides full endothelial function in the midgestation mouse aorta or if instead it is composed of cells from the underlying mesenchyme that temporarily assume endothelial characteristics on their way to forming hematopoietic clusters is a matter of conjecture and/or semantics.

Some studies have suggested that HSCs are generated from mesenchyme located directly under endothelial cells in the ventral aspect of the dorsal aorta25 or in discrete patches ventral-lateral to the dorsal aorta (subaortic patches)57. Runx1 is expressed in mesenchymal cells underlying the ventral aspect of the dorsal aorta, and Runx1+ cells isolated from Runx1-haploinsufficient embryos on the basis of a mesenchymal phenotype (CD45−CD31−VE-cadherin−) have HSC activity25. However, CD45−CD31−VE-cadherin− cells similarly isolated from wild-type embryos do not include HSCs25. Transplantation data show that cells from the subaortic patches (CD45−c-Kit+AA4.1+) have some repopulating activity in immunodeficient adult recipients (0.4–1.9% engraftment)57; however, they are not as potent as the Runx1+, Ly6a-GFP+, VE-cadherin+ or CD45+ HSCs, which are located mainly in the hematopoietic clusters and aortic endothelium, that provide up to 100% engraftment of irradiated adult recipients25,26. The hematopoietic cells located in the subaortic patches may be precursors of the fully potent HSCs found in the aortic endothelial hematopoietic clusters or may represent differentiated progeny of hemogenic endothelium that has ingressed (as in the chick embryo) into this site, or they may be an unrelated population of hematopoietic cells. These mouse data collectively indicate that the direct precursors of HSCs are mainly 'hemogenic endothelial' cells. Vascular endothelium of the human embryo has also been shown to have blood-forming potential58.

It seems that cells in the aortic hematopoietic clusters are not homogeneous. For example, only some cells in chick aortic clusters are CD41+ (ref. 59). Also, the Ly6A-GFP transgene marker associated with mouse aortic HSCs is expressed by only some cells in the aortic clusters and endothelium28,60. This could indicate that some emergent Ly6A-GFP+ HSCs undergo differentiation (becoming GFP−) while residing in the cluster or that adjacent Ly6A-GFP− hemogenic endothelium is contributing to Ly6A-GFP− hematopoietic cells in the clusters. In contrast to the strict ventral localization of hematopoietic clusters in the chick aorta, studies of mice have identified hematopoietic clusters on both the ventral and dorsal aspects of the dorsal aorta24. Functional studies indicate that definitive hematopoietic progenitors reside on both aspects of the aorta, but only the ventral aspect contains fully potent HSCs24. Notably, the chick aortic endothelium has dual origins: the dorsal aspect is derived from paraxial (somitic) mesoderm, and the ventral aspect is derived from splanchnic mesoderm61. Data have shown that chick somitic endothelial cells replace the splanchnopleural hemogenic endothelium in the floor of the aorta once the hematopoietic cluster phase has ended62. In the mouse, a small contribution of somite-derived endothelial cells is also found in the aorta, but in the lateral aspect63. To facilitate studies of the genetic programs that regulate hematopoietic specification, it will be useful to establish the specific mesodermal origins of the hemogenic versus nonhemogenic endothelium in the mouse conceptus.

Extrinsic factors and 'master regulators'

Prospective hematopoietic cells are specified by sets of intrinsic 'master regulators' (transcription factors) and are influenced by morphogens and factors emanating from the surrounding cellular environment; that is, developing adjacent germ cell layers and tissues. The de novo generation of mouse hematopoietic cells in the bilaminar yolk sac (endoderm and mesoderm), the chorio-allantoic placenta (mesoderm and trophectoderm) and the more complex AGM region (dorsal ectoderm, mesoderm and ventral endoderm) suggests distinct interactions and/or genetic programs are operative in each of these sites. Yet the genetic program leading to hematopoietic specification should overlap to some degree among the distinct anatomical territories.

Interactions between endoderm and prospective hematopoietic mesoderm are necessary for hemogenic induction in the chick embryo. Blood island generation occurs only when the mesothelial and endoderm germ layers are cultured together; when they are cultured separately, no primitive erythroblasts form64,65,66. Similarly, somitic mesoderm, which normally contributes only to endothelium in the dorsal aspect of the dorsal aorta and not to the ventral endothelium or hematopoietic clusters, could be 'reprogrammed' to assume the latter fates after transient exposure to endoderm before grafting67. Several signaling molecules, including VEGF, bFGF and TGF-β1, could substitute for this endodermal signal67.

Studies of mouse conceptuses have shown that contact with visceral endoderm is necessary for primitive hematopoiesis in yolk sac explants and that exposure to endoderm can respecify prospective neurectoderm to assume a hematopoietic fate68,69. This endoderm signal can be replaced in vitro by heparin-acrylic beads soaked in Indian hedgehog, which is normally produced by the visceral endoderm. Its expression pattern, together with the explant data, suggests that hedgehog signaling is essential for primitive erythropoiesis69. However, deletion of Indian hedgehog or its receptor (Smoothened) in mice does not eliminate primitive erythropoiesis in the yolk sac, although it does profoundly affect yolk sac vascularization70.

Hedgehog signaling is essential for hematopoiesis in the zebrafish equivalent of the AGM. Hedgehog is situated at the beginning of a signaling cascade that includes the 'downstream' effectors VEGF, Notch, GATA-2 and Runx1 and culminates in the formation of blood cells in the dorsal aorta71. VEGF, together with factors identified in the chick (bFGF, TGF-β and BMP4) are generally thought of as ventralizing factors, whereas dorsalizing factors that antagonize hematopoietic induction include EGF and TGF-α67. Most of the ventralizing factors also seem to be involved in hematopoiesis in the mouse. Embryonic stem cell differentiation cultures and gene-targeting studies have shown involvement of the FGF, TGF and VEGF–Flk-1 signaling axes in vasculogenesis and hematopoiesis3,72,73,74. VEGF is expressed by the yolk sac endoderm, whereas Flk-1 is expressed by the mesoderm, and both are expressed intraembryonically as well75,76. BMP signaling is also important for initiating the blood program in the mouse conceptus. Bmp4−/− embryos generally die around the gastrulation stage, and those that do survive have much less yolk sac mesoderm and erythropoiesis74. The addition of BMP4 to embryonic stem cell–differentiation cultures77 and presumptive anterior head-fold explants induces hematopoietic cell formation20. BMP4 also increases the number of HSCs in AGM explants60. Notably, BMP4 is located in the mesenchyme underlying aortic clusters in the mouse60 and human78 embryo. Ventralizing factors control the expression of pivotal hematopoietic transcription factors such as SCL and GATA-1 that are important in hematopoiesis79.

Notch1 signaling is selectively important for AGM but not yolk sac hematopoiesis. Mutations that affect Notch signaling in zebrafish eliminate Runx1 expression and hematopoietic cluster formation in the AGM71,80. Notch1-deficient mouse conceptuses die at E10 and have almost normal numbers of yolk sac primitive erythroid and erythroid-myeloid progenitors but have no AGM hematopoiesis or HSCs81. Notch1 and Notch4 and their ligands Delta-like 4, Jagged 1 and Jagged 2 are expressed in endothelial cells lining the dorsal aorta82. Overexpression of Runx1 in Notch signaling mutants in both zebrafish and mice restores AGM hematopoiesis, which indicates that Runx1 is at least in part genetically downstream of Notch80,83.

Mice lacking the transcription factor GATA-2 suffer from severely impaired primitive erythropoiesis and a complete lack of other committed progenitors and HSCs and die at E10.5 (ref. 84). GATA-2 is expressed in the aortic endothelium55 and is thought to affect the expansion of the hemogenic population emerging from these cells54. Notably, Gata2 haploinsufficiency profoundly decreases the number of AGM HSCs, but yolk sac HSCs are affected only slightly. Runx1, another pivotal transcription factor for definitive hematopoiesis, is expressed ventrally in the mesenchyme, endothelium and hematopoietic clusters of the dorsal aorta25,85. Runx1-deficient conceptuses have essentially normal primitive erythropoiesis but have no myeloid or lymphoid-myeloid progenitors of any sort and have no AGM HSCs86,87,88. The Ets-family transcription factor PU.1, which is required for definitive hematopoiesis, is a critical 'downstream' target of Runx1 (refs. 89,90,91,92). Runx1 haploinsufficiency leads to an increase in AGM HSCs when these are directly isolated from the embryo and transplanted into irradiated adult mice25. However, when hematopoietic tissues of Runx1+/− conceptuses are first cultured as explants and then transplanted, they have notably different responses to Runx1 haploinsufficiency. AGM explants have many fewer HSCs, but both yolk sac and placenta have many more HSCs, which suggests that different regulatory networks, 'downstream' targets, interacting molecules and/or developmental timing are operative in these tissues93. Because transcription factors work in complexes and act at several points in hematopoietic development, the interplay between specific transcription factors is an important aspect of hematopoietic specification. The finding that HSC-specific enhancers of Tal1 and Runx1 can bind multiprotein complexes containing GATA and Ets factors94 and GATA, Ets and SCL factors95, respectively, suggests a higher order of regulatory complexity. In addition, prostaglandin E2 and interleukin 1, which are normally associated with the regulation of inflammatory molecules, also affect hematopoiesis in the zebrafish and mouse AGM96 (C. Orelio, personal communication). Thus, an understanding of how the 'master regulators' are controlled and are 'fine tuned' in terms of their amounts in different hematopoietic subpopulations and sites will provide insight into the genetic network that governs hematopoietic emergence in the conceptus. By analogy to the embryonic stem cell program97, it is likely that a small set of factors (Runx1, GATA, Ets and SCL) establishes hematopoietic cell identity in the conceptus.

Tracing cells to secondary hematopoietic territories

Once the various hematopoietic progenitors and HSCs emerge from their anatomically distinct sites, they are thought to enter the circulation and colonize the fetal liver (Fig. 3). In zebrafish, CD41+ hematopoietic cells enter the circulation by intravasation through the posterior cardinal veins98. The mode of entry of mammalian progenitors and HSCs into the circulation is presumed to occur through the release of cells from hematopoietic clusters into the artery or yolk sac vasculature, but this has not been directly demonstrated. In the mouse, the liver rudiment is colonized from late E9 onward, and later the spleen and thymus are seeded either directly or from the fetal liver99,100. Although it seems peculiar that the conceptus first produces hematopoietic cells without the full potency of an adult HSC, these early classes of cells that are limited in potency and lifespan may provide maturation signals to the rudiments of the secondary hematopoietic territories. Indeed, in the chick, several waves of early lymphocyte–like cells enter the thymus at receptive times and promote thymic growth101. Similarly, mouse genetic models have demonstrated interactive 'dialogs' between early lymphocytes entering the developing thymus and thymic epithelium102. Thus, in addition to the well known function of primitive erythrocytes in gaseous exchange, definitive myeloid and lymphoid-myeloid progenitors may function to stimulate the growth of secondary hemato-lymphoid territories.

Attempts have been made to trace the lineage of cells in the conceptus that give rise to the permanent hematopoietic system in the adult mouse. In place of the ex utero studies that can be done with nonmammalian vertebrates, lineage tracing in the mouse has relied on Cre-loxP recombination-marking technology together with tissue-specific and inducible transgenes103. Several transcriptional control elements have been used to direct expression of Cre recombinase to hematopoietic cells in the conceptus; these elements include CD41, which is expressed in a subset of definitive hematopoietic progenitors34,104; SCL, which is expressed in endothelial cells and definitive HSCs53; and Runx1, which is expressed in all definitive hematopoietic cells, hemogenic endothelium, and some mesenchymal cells25. Genes encoding all three of those elements are also expressed in the mesodermal precursors of the primitive erythrocytes. Rosa26 reporter strains are used to detect the cells in which Cre is active and the progeny of such cells.

CD41-Cre marks a high percentage of lymphoid and myeloid cells in the fetus but only 5% of adult bone marrow cells105. Therefore, although CD41 marks most progenitors in the yolk sac34,104, most adult hematopoietic cells do not transit through a CD41-expressing precursor. Scl- and Runx1-marking experiments have used tamoxifen-inducible Cre (Scl-CreERT and Runx1-CreERT, respectively) to control the temporal window of active recombination106,107. Activation of Scl-CreERT pregnant females at E10 and E11 of gestation marks approximately 10% of the cells in the adult bone marrow, which indicates that the progeny of SCL-expressing cells at midgestation contribute to adult hematopoiesis. In experiments with Runx1-CreERT, after pregnant females are injected with tamoxifen between E7.5 and E10.5 (ref. 107), varying numbers of marked cells are found in the adult bone marrow depending on the day of injection. The authors of those last studies106,107 conclude that the adult hematopoietic cells are derived from yolk sac cells, as Runx1 is highly expressed in the yolk sac at the time of tamoxifen injection. However, others have found Runx1-expressing cells at the base of the allantois and in the pSp within 0.5 days to 1 day after the onset of Runx1 expression in the yolk sac39,95. The inability to precisely stage embryos in utero, together with the uncertain kinetics of CreERT activity and recombination, makes it difficult to discern from what anatomical site the marked cells are derived. Additionally, the allele encoding Runx1-CreERT creates a Runx1 haploinsufficiency known to affect both the temporal and spatial appearance of HSCs in the conceptus86. Hence, although the data confirm that Runx1-expressing cells from the early conceptus contribute to hematopoiesis in the adult, improvements in directing Cre recombination to specific sites of hematopoietic cell emergence will be necessary to unambiguously identify the anatomic origins of adult HSCs.

Summary

Hematopoietic development in the mammalian conceptus occurs in several mesodermal lineages, as defined by cells emigrating from the primitive streak to three distinct hemogenic tissues: yolk sac, pSp-AGM and chorio-allantoic placenta. At least five distinct classes of hematopoietic activities have been described with an increasingly progressive generation of more complex hematopoietic activities that culminate in the de novo generation of adult repopulating HSCs. The direct precursors of hematopoietic cells in the conceptus are hemangioblasts and the hemogenic endothelium of the main embryonic vasculature. The genetic program directing hematopoietic fate determination involves known developmental signaling pathways that converge on the directed expression of a small set of pivotal hematopoietic transcription factors. It is the balance of the amounts and timing of the expression of these factors in the different embryonic tissues that drives the fate determination and emergence of hematopoietic progenitors and HSCs. And although the specific lineages of cells contributing to the mammalian adult hematopoietic system are in question, the legacy of Runx1- and SCL-expressing embryonic cells in the adult is certain. The future challenge is to determine the precise balance of factors necessary to direct HSC fate ex vivo in highly expandable and accessible populations of cells, such as embryonic (stem) cells or other somatic cells. This knowledge and insight gained from the study of HSC emergence in vertebrate embryos will ultimately be directed toward the de novo production of fully potent transplantable HSCs for cellular and molecular clinical therapies of human blood-related genetic diseases and leukemias.