Assembling the layers of the hematopoietic system: A window of opportunity for thymopoiesis in the embryo

During embryonic development, several independent generations of hematopoietic cells were identified. They occur in the yolk sac and the intra‐embryonic major arteries, in a narrow window of development. They arise sequentially, starting with primitive erythrocytes in the yolk sac blood islands, progressing to less differentiated erythromyeloid progenitors still in the yolk sac, and culminating with multipotent progenitors, some of which will generate the adult hematopoietic stem cell compartment. All these cells contribute to the formation of a layered hematopoietic system that reflects adaptative strategies to the fetal environment and the embryo's needs. It is mostly composed, at these stages, of erythrocytes and tissue‐resident macrophages both of yolk sac origin, the latter persisting throughout life. We propose that subsets of lymphocytes of embryonic origin derive from a different intra‐embryonic generation of multipotent cells occurring before the emergence of hematopoietic stem cell progenitors. These multipotent cells have a limited lifespan and generate cells that provide basic protection against pathogens before the adaptive immune system is functional, contribute to tissue development and homeostasis, and shape the establishment of a functional thymus. Understanding the properties of these cells will impact the understanding of childhood leukemia and of adult autoimmune pathology and thymic involution.


| FORE WORD
More than 30 years ago, Herzenberg and Herzenberg proposed that the mammalian hematopoietic system is composed of cells originated from different waves of hematopoietic stem cells (HSC) with specific differentiation potential, emerging sequentially during embryonic development. 1 This notion integrated findings suggesting the embryonic origin of some adult lymphocyte subsets that were not easily generated from adult bone marrow (BM) progenitors and was published in a paper entitled "Toward a Layered Immune System".
Although the tools to analyze hematopoiesis and the knowledge on progenitor subsets evolved, the notion that the hematopoietic compartment is composed of cells that originate from developmentally restricted progenitors and that persist for life is of actuality.

| Extra-embryonic generation: the yolk sac
Colony-forming cells in the spleen (CFU-S) is an in vivo assay that measures the presence of highly proliferative hematopoietic progenitors considered, for a long time, to comprise HSC. 2 The first multilineage colony-forming cells were identified in the yolk sac (YS) at embryonic day (E)8 as CFU-S analyzed 8 days after transplantation (CFU-S 8 ). 3 In vitro colony formation in semi-solid medium was also first detected at this stage, peaked in numbers at E10 and decreased thereafter. 3 These results were consistent with a YS origin of HSC. However, pioneer experiments in the avian model challenged this idea. Chicken chimeras derived from chick embryos that developed in a major histocompatibility complex (MHC) allogeneic egg showed that adult hematopoietic cells derive from the embryo proper and not from the YS. 4 Primitive erythrocytes were initially identified as the first embryonic hematopoietic cells of YS origin. 5 Circulating primitive erythrocytes are nucleated and express specific hemoglobin chains that have increased capacity to transport oxygen compared to adult hemoglobin. 6 They can be found around E7 of mouse gestation in YS blood islands, peak by the time circulation is established (E8-8.5), and decrease after E12.5, although a few were detected after birth. 5 They circulate but are also found in the fetal liver (FL), the first main site of embryonic hematopoietic differentiation. Limited numbers of primitive macrophages and megakaryocytes have also been reported in the YS blood islands. 5 Around E8-8.5, a second layer of hematopoietic progenitors emerges in the vasculature of the YS. It is composed of erythromyeloid progenitors (EMP) that generate definitive erythrocytes and multiple myeloid lineages. 7,8 However, they fail to generate lymphocytes in vitro or in vivo. The YS has long been considered a provider of cells that transiently ensure embryo survival. It came therefore as a surprise that YS EMP-derived macrophages persist for life as tissue-resident macrophages. 9 The best example are microglia, the macrophages residing in the central nervous system (CNS) that, under physiologic conditions, stay throughout adult life with the same original cell composition and are not replaced by BM-derived cells (discussed below). Subsequently, YS mast cells were also found to have a long lifespan. 10 This observation indicates that mature blood cells emerging in the embryo in an HSC-independent manner persist and are functional for long periods of time, thus challenging the paradigm that all adult hematopoietic cells are derived from HSC.

| Intra-embryonic generation: major arteries
Experiments in several models including the chicken, 4,11 the mouse, [12][13][14] and the zebrafish 15,16 indicated an intra-embryonic origin of lymphoid progenitors that paralleled that of HSC. This generation was restricted to the major arteries (aorta, omphalo-mesenteric, and vitelline arteries) 17 and occurred in close vicinity to endothelial cells lining the vessel. The emergence of hematopoietic progenitors was directly observed in the zebrafish larvae. 15,16 These observations indicated that hematopoietic progenitors emerge from endothelial cells through a process, designated endothelial-to-hematopoietic transition (EHT), that involves progressive morphological alterations, loss of endothelial cell markers, and acquisition of hematopoietic specific transcripts. It was subsequently proposed that EMP in the YS emerge by a similar mechanism. 18 For the purpose of this review, we will use the term "embryo-derived" for all hematopoietic progenitors that are generated within the embryo proper.
The above experiments indicated that a third layer of hematopoietic progenitors emerges in the intra-embryonic dorsal aorta also called aorta-gonads-mesonephros (AGM) and in the omphalomesenteric and vitelline arteries that connect circulation between the embryo proper and the extra-embryonic tissues (YS and placenta). 17 These progenitors can generate erythroid, megakaryocyte, myeloid, and lymphoid cells at the clonal level. They migrate to the FL where they establish the adult HSC compartment. The emergence of the third layer is limited to the period between E9 and E11.5 and partially overlaps in time, in the mouse, with the previous EMP generation. 19 The first emerging multipotent progenitors were assessed for their capacity to reconstitute the hematopoietic compartment of myeloablated recipient mice for long periods of time. This ability designated long-term reconstitution (LTR) activity is taken as the golden standard to evaluate HSC. These experiments revealed a poor LTR capacity of E9-E10 progenitors. 20 Although CFU-S were detected in the intra-embryonic compartment earlier than in YS, 13,21 very low numbers of cells with LTR activity were found before E12.
It was subsequently shown that emerging multipotent progenitors needed to undergo maturation in an organ culture of AGM or FL in order to acquire LTR activity and that these newly generated multipotent progenitors were designated pre-HSC 22 or immature-HSC. 23 One reason for their inefficient reconstitution capacity is their low expression of MHC-class I on the surface that activates host NK cell activity, 14 but other factors are probably also involved in their fast elimination from hosts by innate immune mechanisms. These experiments dissociated CFU-S 11 and LTR activity, indicating that the CFU-S assay does not detect HSC. 24 They also indicated that testing the in vivo activity of the emerging multipotent progenitors should not be done in the same assays used to detect adult HSC.

| The fetal liver, the major embryonic hematopoietic organ
The FL is the main hematopoietic organ in the mouse embryo. It develops from the hepatic endodermal bud (that gives rise to the liver parenchyma), that proliferates in cords that invade the mesodermal septum transversum (giving liver fibroblasts or stellate cells, mesothelial and endothelial cells), and it provides the environmental signals that support hematopoietic differentiation and the establishment of the HSC compartment. 25,26 It was early recognized that hematopoietic cells in FL are originated elsewhere. 25 Later studies, using organ cultures and, again, in vivo reconstitution assays, confirmed that the FL is initially colonized at E10.5 (30 somite [S] stage), that is, shortly after being produced in the AGM, by immature HSC, which will fully develop to HSC in situ. 23 Ema and Nakauchi further showed that, in FL, the number of cells displaying LTR activity increased more than 30-fold between E12.5 and E16.5, and decreased thereafter. These observations were interpreted as reflecting a single wave of HSC expansion in FL and a progressive exit to colonize the BM or the spleen. Interestingly, E12.5 is also the time when the first B cell progenitors were detected in FL. 29,30 Although these conclusions remain unchallenged, the extent to which HSC expand in the FL is a matter of debate (see below).
Little is known about the properties of the FL stromal cells that sustain hematopoiesis. Endothelial cells, 31

| A layered immune system
In the late 80 s it was reported that a small subset of B cells (designated B1 cells to differentiate them from B2 conventional B cells), reside in the peritoneal and pleural cavities, expressed the T cell marker CD5 and were endowed with some surprising properties. 35 45 In the absence of a consensual view, we will consider here that FL progenitors contribute more efficiently to the B1 compartment than their BM counterparts.
Around the time B1 cells were described it became also evident that subsets of mouse γδ T cells expressing invariant T cell receptors (TCR) 46 also had specific properties not shared with other T cells. only in the FL is RoRγt expressed in lymphoid progenitors before tissue residency. 58 Thus, the immune system, like embryonic erythroid and myeloid cells, is composed of cells that were generated in a developmentally restricted manner. The origin and function of these different hematopoietic progenitors will be discussed below.

| Yolk Sac Hematopoiesis -longer than expected
As previously mentioned, the hematopoietic system is established by multiple waves of progenitors. This includes all multipotent progenitors that may or may not undergo maturation into HSC and their progenies ( Figure 1).  9 While about half of those cells were slowly replaced by HSC-derived progeny, microglia cells were only marginally replaced during a year period. 9,60 In fact, the contribution of BM-derived precursors and/or circulating monocytes to the microglia pool has only been observed in situations where the integrity of the blood-brain barrier (BBB) has been disrupted, e.g., using radiation, or in the context of disease. [61][62][63] In the developing CNS of rodents, primates and humans, microglia cells were progressively found at key sites of neurogenesis such as the ventricular and subventricular zone, [64][65][66] where they regulate the size of the neural precursor pool by phagocytosis. 67 YS-derived hematopoiesis has proven to also have a higher relevance than initially thought regarding other blood populations.
Indeed, lineage-tracing analysis of YS versus HSC-derived hematopoiesis showed that there is a minimal contribution of HSC to various cellular subsets in the embryo. It is the case for mast cells, which are predominantly YS-derived in the embryo and at perinatal stages in most connective tissues. 10,70 Furthermore, recent evidence showed that, until birth, red blood cell production is maintained by YS EMP with minimal contribution from HSC. 71 Remarkably, YS-derived erythrocyte progenitors require 10-fold lower concentrations of erythropoietin (Epo) than their HSC-derived counterparts and have, therefore, a selective advantage in an Epo-poor environment such as the embryo. This can explain why patients with mutations altering the receptor-binding kinetics of Epo only display anemia postnatally. 72 These studies have challenged the dogma by which HSC assume blood production from the moment they are generated and are the source of all hematopoietic lineages in the adult, thus providing evidence for the existence of a developmentally regulated "layered" hematopoietic system. The obvious selective pressure for such phenomenon lies in the differences in environments that exist in FL and BM for hematopoiesis. The FL exhibits lower cytokine concentrations and, therefore, progenitors that respond to lower concentrations of chemokines and cytokines, like YS progenitors will have a selective advantage over HSC and their progeny. 44,71,73 One can speculate that, by taking over the production of myeloid

| Embryo-derived progenitors -multipotency versus stemness
The first embryo-derived hematopoietic progenitors appear in clusters in the dorsal aorta and other major arteries through a process of EHT performed by specialized endothelial cells named hemogenic endothelium (HE). 15,16,74 These progenitors are multipotent, giving rise to all blood lineages, and the only ones with lymphoid potential. 12,13 The EHT process starts with the upregulation of the transcription factor Runx1, essential for HSC generation, that activates the hematopoietic transcriptional program. 75 Whether all multipotent progenitors generated in the major arteries will give rise to HSC has recently been questioned. 83  Another study points to heterogeneity in the HE in terms of HSC potential as Mecom, a critical regulator of HSC, is preferentially expressed in progenitors that will mature into phenotypic HSC. 83 Accumulating evidence points, therefore, to the existence of multipotent progenitors independent of HSC that will be further described below.

| The establishment of the HSC compartment
At this stage, we will discuss the attempts to evaluate the degree of expansion of the different hematopoietic compartments, as well as how the existence of multiple layers can be integrated in these views.
The organization of the hematopoietic system is classically viewed as a hierarchical tree, with HSC continuously sustaining all mature blood lineages throughout life. However, this concept has only been based on a "retrospective" assessment of the two hallmarks of HSC In contrast to the adult system, it has been widely accepted that HSC are highly proliferative during embryonic development, expanding significantly in the FL, 28   Pre. ---1700 Pre.

| Thymic epithelial regeneration
The thymus is a highly dynamic organ that increases in size for the Thus, the first wave of TSP would fulfill the essential function of inducing TEC maturation before TCRαβ T cells develop, thereby shaping the thymic architecture required for thymocyte selection ( Figure 2B). It is also conceivable that the TEC compartment has a window of opportunity for adequate maturation and that the equilibrium between mature TEC and TECP determines, at this stage, the proper response to disturbances in the adult TEC compartment. If that were the case, a late or inadequate crosstalk between thymocytes and TEC would result in a life-long defective compartment with potential consequences for development of autoimmunity and premature thymic involution.
After birth, invariant Vγ5 + T cells migrate to the skin where they expand and contribute to the integrity of the epithelial barrier, and to the tissue repair upon injury. 48,138 Thymic LTi are also gradually lost although their fate is presently unknown. 132 To date, the role of invariant Vγ6 + T cells in thymic organogenesis has not been addressed.
These cells are also of exclusive embryonic origin, and their development is coupled with their capacity to produce high levels of IL-17.
While some stay resident in the thymus, the vast majority leave to the periphery in the perinatal period and colonize the mucosa of the genitourinary tract as well as the airway epithelia where they play an important role in acute inflammation, and tissue repair. 139 Although the first TSP also produce αβ T cells, the absence of terminal deoxynucleotidyltransferase (TdT) and, therefore, of N sequence additions at that stage limits the diversity of their TCR. 135 They are thought to be rapidly replaced by the highly diverse T cells produced from the second wave TSP that colonize the thymus after E15.5. These progenitors are endowed with high proliferative capacity and generate large numbers of TCR-diverse, αβ-and γδ-T cells (Figure 2A). Second wave TSP share essentially all key properties with adult TSP. 134 Similarly, again exclusively in the referred time window, the FL also harbors a subset of lymphomyeloid-primed progenitors (LMPP) that are PIR + , whose transcription signature is already biased toward restricted T/ILC potential. 73  Whether the first wave of TSP originates from YS or intraembryonic multilineage progenitors that emerge from the dorsal aorta has been a matter of debate due to the fact that, in mice, the time window of hematopoiesis at both sites largely overlaps. It has been proposed that YS-derived cells have the potential to generate lymphoid cells. 140 These results were based on the observation that lymphoid-associated transcripts (namely Il7Ra, Rag1, and Rag2)

| The origin of fetal TSP
were detected in YS cells, prior to the emergence of HSC. 140 Along the same line, it was also suggested that YS progenitors were the source of first wave TSP, as assessed by the emergence of invariant Vγ5 + T cells, in experiments using a temporally-controlled lineage tracing model. 141 However, as the authors themselves acknowledged, the results were as well compatible with an intra-embryonic, HSC-independent, origin of these cells.
Recently, Elsaid et al 135  In humans, evidence also exists supporting the concept of a  conditions that label YS-EMP, indicating that they originate from different progenitors also found by Elsaid et al. 135 However, also conditions that label a majority of HSC yield low frequency of YFP labeled embryonic lymphoid cells, suggesting that they do not originate from HSC or their immediate progenitors.

A transient population of multipotent progenitors appears bi-
ased to the production of B1 cells and invariant γδ T cells in vivo, and these cells emerge in the mouse embryo before HSC are detected. 142 3. Human fetal hematopoietic progenitors were shown to generate T cells with different properties than their adult BM counterparts. 150 4. CXCR4 expression in the dorsal aorta HE distinguishes an early and a late hematopoietic generation that differ in their capacity to integrate the HSC compartment. 84 Another compelling argument is the analysis of the incidence of childhood leukemias: 1. The higher incidence of myeloid leukemias in aged individuals contrasts with the higher incidence of acute lymphoid leukemias (ALL) in infants. 151 It was recently suggested that the target of childhood ALL is a progenitor cell that is lymphoid biased and can only be found during embryonic development.
2. Another interesting observation is the high susceptibility of children with Down syndrome (DS) to develop leukemias, not found in adult DS patients. 152 It is, therefore, tempting to speculate that the target cell for the infant DS leukemia is no longer present in the adult BM thus accounting for the discrepancy between young and adult DS patients in developing the disease. 153 3. The common ALL translocation, MLL-AF9, induces different diseases depending on whether it is induced in neonatal human progenitors or in adult CD34 + cells. It is therefore tempting to speculate that, in humans, lymphoid-biased transient multipotent cells, equivalent to the ones we propose here, are targets of the translocations that lead to ALL. 154 The first wave of TSP induces the maturation of the thymic architecture, essential to the selection of conventional T cell. 53,135 Genetic or environmentally induced variations in the maturation of the thymic epithelial compartment might impact immunity for life, increasing risks of autoimmune disorders or premature thymic involution. The invariant γδ T cells also have roles in the periphery by improving wound healing, in the skin, 155 and epithelial regeneration, in the lungs. 139 Likewise, it could be speculated that altering the dynamic of secondary lymphoid organ formation initiated by LTi could modulate memory immune responses and even cancer development. The natural antibody repertoire of B1 cells appears to be selected by early exposure to bacterial products and might be essential as first line of defense specially in newborns where the adaptive immune system is immature. 38 All these properties would have conferred a selective pressure to maintain this new, yet transient, generation of progenitors that do not contribute to the HSC compartment and possibly also not to adult hematopoiesis.
Further studies are required to assess how long these hematopoietic progenitors persist, whether they can reach the BM, how to distinguish them from other multipotent cells and what are the molecular basis for their different behavior.

ACK N O WLE D G E M ENTS
We thank all members from the Lymphocyte and Immunity Unit, Elisa Gomez-Perdiguero and the members from her laboratory for excellent discussions.

CO N FLI C T O F I NTE R E S T
All authors declare no have a conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available on request from the authors.