Multiple waves of fetal-derived immune cells constitute adult immune system

Summary

It has been over three decades since Drs. Herzenberg and Herzenberg proposed the layered immune system hypothesis, suggesting that different types of stem cells with distinct hematopoietic potential produce specific immune cells. This layering of immune system development is now supported by recent studies showing the presence of fetal-derived immune cells that function in adults. It has been shown that various immune cells arise at different embryonic ages via multiple waves of hematopoiesis from special endothelial cells (ECs), referred to as hemogenic ECs. However, it remains unknown whether these fetal-derived immune cells are produced by hematopoietic stem cells (HSCs) during the fetal to neonatal period. To address this question, many advanced tools have been used, including lineage-tracing mouse models, cellular barcoding techniques, clonal assays, and transplantation assays at the single-cell level. In this review, we will review the history of the search for the origins of HSCs, B-1a progenitors, and mast cells in the mouse embryo. HSCs can produce both B-1a and mast cells within a very limited time window, and this ability declines after embryonic day (E) 14.5. Furthermore, the latest data have revealed that HSC-independent adaptive immune cells exist in adult mice, which implies more complicated developmental pathways of immune cells. We propose revised road maps of immune cell development.

The long history supporting the layered immune system theory

The stem cell hypothesis is a central dogma in hematology and immunology that proposes that all blood cells are produced by HSCs. An HSC is defined as a cell that has self-renewal and multilineage differentiation capacity, which can be proven only by transplantation assays. Numerous efforts to isolate HSCs from adult bone marrow (BM) at the single-cell level have been made^1,2, and one HSC has been shown to repopulate blood cells of all five lineages (erythroid, megakaryocyte, myeloid, T-cell and B-cell lineages) following transplantation into lethally irradiated congenic recipient mice³. Precise observation of blood lineage engraftment patterns by single HSCs has revealed that the outputs of HSCs are different between HSCs of different ages and even after secondary transplantation, highlighting the heterogeneity of HSCs^3,4. However, these studies did not always include donor-derived tissue-resident immune cells in their analysis, leaving the origins of some tissue-resident immune cells unclarified.

It is now recognized that some tissue-resident immune cells are fetal-derived and are not the progeny of adult HSCs. For example, Vγ3-T cells in the skin are produced only during the fetal period from extra-embryonic yolk sac (YS) precursors^5,6 or fetal stem/progenitor cells^7,8. Similarly, brain microglia are derived from early YS progenitors and are not replaced by BM HSCs^9–12. Mast cells are easily generated from adult BM cells in vitro, but it has been challenging to repopulate mast cells in vivo following HSC transplantation¹³. Peritoneal B-1 cells, specifically CD5⁺ B-1a cells, are not fully replaced by BM progenitors and are produced more efficiently from fetal liver (FL) progenitors, while conventional B cells (B-2 cells) are fully repopulated by both adult BM and FL HSCs^14–17. These observations led Drs. Herzenberg and Herzenberg to propose the layered immune system theory in 1989,¹⁸ in which several types of stem cells arising at different times during development sequentially produce different types of immune cells. Since then, the layered immune system theory has been supported by many pieces of data, and multiple waves of hematopoiesis, in which many immune cells arise sequentially in an HSC-independent manner followed by HSC emergence as a final hematopoietic wave during embryonic development, have been recognized. One important difference from the original layered immune theory is that it is ECs and not HSCs that produce different types of immune cells within the limited time window of embryonic development, as discussed in detail in a later section.

This review will highlight recent findings from various laboratories, with an emphasis on our work, that has traced the fetal emergence of HSCs, the first lymphoid potential, and mast cells. Based on this information, we present an updated model of immune system layering.

CD5⁺ B-1 cells: historical view of the research in finding their origin.

There are two main B-cell subsets: conventional B-2 cells and innate-like B-1 cells. B-2 cells include follicular (FO) B cells in the spleen and lymph nodes (LNs), while B-1 cells reside primarily in the body cavities and to a limited degree in the spleen and BM. B-1 cells express unique surface markers such as IgM^hiIgD^loCD23^lo and CD11b (in the case of peritoneal B-1 cells), while B-2 cells express IgM^loIgD^hiCD23^hi.¹⁹ B-1 cells are further categorized based on CD5 expression into CD5⁺ B-1a cells and CD5⁻ B-1b cells. B-1 cells secrete natural IgM antibodies in a T-cell-independent manner, and these IgM antibodies are known to be biased toward the germline IgH repertoire, frequently using VH11 or VH12, with no N-additions to the CDH3 region (CDH3-R).¹⁷ These features are linked to their fetal origin.

B-2 cells ultimately develop from HSCs in the BM through well-defined progenitor stages that are phenotypically resolved based on Hardy’s classification²⁰. On the other hand, B-1 cells are mainly derived from FL progenitors, as shown by transplantation assays^15,16. Further detailed transplantation studies have indicated that while FL pro-B cells can repopulate both B-1 and B-2 cells, adult BM pro-B cells efficiently repopulate B-1b and B-2 cells but inefficiently repopulate peritoneal B-1a cells¹⁶. These results indicate that peritoneal B-1a cells primarily have a fetal origin, as we discuss extensively in subsequent sections, and that adult BM progenitors have a limited capacity to produce B-1a cells in the peritoneal cavity.

So, what is the difference between BM pro-B cells and FL pro-B cells? Yuan et al. examined the difference in miRNA expression profiles between pro-B cells in the FL and BM and found that expression of the let-7 family was largely absent in FL pro-B cells²¹. They also found that expression of Lin28b, which specifically regulates the let-7 family, was abundant in FL B progenitors, common lymphoid progenitors (CLPs), and lineage (lin)⁻Sca-1⁺c-kit⁺ (LSK) hematopoietic stem and progenitor cells (HSPCs). When Lin28b was ectopically overexpressed in adult BM LSK cells, these stem/progenitor cells repopulated B-1a cells in the peritoneal cavity and γδ T cells in the thymus following transplantation into Rag1^−/− recipient mice (although the percentage of engrafting Lin28b⁺GFP⁺ cells was low, and there was no Vγ3-T-cell engraftment). These data indicate that Lin28b plays a significant role in the fate determination of fetal-derived lymphocytes and that its overexpression in adult BM progenitors results in an adult to fetal program shift.

Zhou et al. extended these results and demonstrated that ectopic expression of Lin28b in adult BM pro-B cells induced a B-2 to B-1a cell differentiation switch²². The researchers also determined that Arid3b, a target of let-7, regulates B-1a cell specification. Overexpressing Arid 3b in BM pro-B cells also induced differentiation toward B-1a cells, while knockdown of Arid3b in fetal pro-B cells blocked B-1a cell development. This elegant study indicated that Lin28b regulates B-1a cell differentiation at the pro-B-cell stage. However, the BCR repertoire of B-1a cells produced from adult BM pro-B cells that express Lin28b (converted B-1a cells) differed from that of B-1a cells in the peritoneal cavity; that is, the converted B-1a cells had CDH3 N-additions similar to those of FO B-2 cells. These results raise a question of whether the phenotypic change induced by Lin28b overexpression in B-2 pro-B cells truly changes them into B-1a cells. If it does, it would be of interest to examine whether the converted B-1a cells secrete natural IgM antibodies and if these natural IgM antibodies recognize phosphorylcholine or phosphatidylcholine as fetal-derived B-1a cell-derived IgM. Also, it would be interesting to examine if these converted B-1a cells can class switch to produce IgG as B-2 cells do. In addition, it has been reported that adult BM progenitors can produce B-1a cells that have a distinct repertoire compared with fetal-derived B-1a cells, with more N-additions to CDH3-R^{23 24}, similar to converted B-1a cells. Therefore, it is plausible that a portion of adult BM progenitors may use the same Lin28b–let-7–Arid3a pathway to produce B-1a cells in steady state conditions, which raises the question of whether the converted B-1a cells are similar to adult BM-derived B-1a cells that display N-additions.^24,25

B-1 cell origin: do HSCs produce B-1a cells?

As discussed above, peritoneal B-1a cells are repopulated more efficiently by FL than adult BM progenitors, while conventional B-2 cells are repopulated by both BM and FL progenitors upon transplantation¹⁶. It has also been reported that highly purified long-term (LT)-HSCs (CD150⁺CD48⁻ LSK cells) in adult BM fail to repopulate peritoneal B-1a cells²⁶, further supporting the fetal origin of B-1a cells. However, what cells are the ultimate progenitors for B-1a cells? Montecino-Rodriguez et al. identified B-1 cell-specific progenitors with a lin-AA4.1⁺CD19⁺B220^lo phenotype in the FL and fetal and neonatal BM²⁷. These cells had undergone DJ but not VDJ recombination and were classified as the B-1 cell equivalent of B-2 pro-B cells in adult BM²⁰. These cells were also shown to be derived from B-1-restricted CLPs²⁸. FL LSK cells, including HSCs and multipotent progenitors (MPPs), which are upstream of CLPs, efficiently repopulate B-1a cells in the recipient peritoneal cavity; therefore, it had been assumed that the ultimate origin of B-1a cells is HSCs in the FL and that HSCs lose B-1a cell production ability with age because it is well known that HSCs change their characteristics during the transition from fetal-neonatal periods to adulthood^29,30, including surface marker expression³¹, gene expression³², cell cycle characteristics³³, and differentiation capacities^4,34,35.

Beaudin et al. found “developmental transient HSCs” in the FL that produce innate-like B-1a cells upon transplantation³⁶. In the study, the researchers utilized Flk switch mice in which all cells initially express tdTomato and permanently switch to green fluorescent protein (GFP) expression upon Flk2 expression³⁷. Adult LT-HSCs do not express Flk2, and Flk2 expression is considered to identify downstream MPPs³⁸. However, surprisingly, Beaudin et al. found GFP⁺ HSCs in the FL that repopulated the recipient blood system long-term, and more importantly, the GFP⁺ HSCs efficiently repopulated peritoneal B-1a cells and splenic marginal zone (MZ) B cells, while the GFP⁻ HSCs primarily repopulated splenic FO B cells. Because Flk2⁺ (GFP⁺) cells are not HSCs in adult BM, the FL Flk2Cre⁺ LSK cells that display long-term repopulation ability, including the ability to repopulate B-1a cells, are referred to as “developmental transient HSCs”, and are considered to diminish during development.

This study raised an intriguing question of whether such unappreciated stem/progenitor cells that are not detectable in a regular experimental setting exist. However, Flk2⁺ LSK cells, and not purified CD150⁺CD48⁻ LSK cells, have been used for transplantation assays; therefore, the B-1a cell repopulation ability might have been overestimated because our data show that the CD150⁻CD48⁺ MPPs within the LSK population have more B-1a cell repopulation ability, which will be discussed later.

Additional evidence that FL HSCs possess B-1a cell repopulation ability was provided by Kristiansen et al., who used a cellular barcoding system³⁹. In the study, LSK cells from embryonic day (E) 14.5 FL were transfected with a barcoded lentivirus in vitro and transplanted into lethally irradiated recipient mice, and the barcodes among the peritoneal B-1a cells, spleen FO B cells (B-2 cells), and splenic granulocytes were examined. The researchers found shared barcodes among these three populations, leading to the conclusion that the FL HSCs produced these three types of cells. Barber et al. also showed that CD150⁺ LT-HSCs from neonatal BM repopulate B-1a cells²⁸. These results indicate that FL and neonatal HSCs can produce peritoneal B-1a cells.

However, we and others found that FL LT-HSCs did not efficiently repopulate peritoneal B-1a cells^40,41. In our study, LT-HSCs, short-term (ST)-HSCs, and MPPs were sorted from E15.5 FL based on their expression of CD150⁺ and CD48⁺ among LSK cells and transplanted into lethally irradiated congenic mice or sublethally irradiated NOD/SCID/IL2Rγc⁻/⁻ NSG neonates⁴¹ (Fig. 1A). In both studies, LT-HSCs did not repopulate B-1a cells efficiently; instead, MPPs in the FL were the main providers of B-1a cells (Fig. 1B, C). Of note, when the numbers of donor-derived B-1a cells were compared, ST-HSCs and MPPs showed similar B-1a repopulation abilities (Fig. 1E). Importantly, the median fluorescence intensity (MFI) of CD5 was higher in MPP-derived B-1a cells than in LT-HSCs or ST-HSCs (Fig. 1F). As such, in transplantation settings, all three populations in FL LSK cells have B-1a cell repopulating ability. However, if LT-HSCs are the ultimate origin of B-1a cells, LT-HSCs and MPPs should show similar B-1a cell repopulating abilities. The discrepancy in the B-1a repopulation percentages between FL LT-HSCs and MPPs raises a question of whether FL MPPs are derived from LT-HSCs.

Figure 1. FL HSCs fail to repopulate B-1a cells.

A. Gating strategy for LT-HSCs, ST-HSCs, and MPPs based on CD150 and CD48 expression among LSK cells from the E15.5 FL. B, C. Recipient peritoneal cavity repopulated by LT-HSCs (B) and MPPs (C). D. Donor-derived repopulation of peritoneal B-cell subsets. E. Numbers of donor-derived B-1a cells in the recipient peritoneal cavity repopulated by LT-HSCs, ST-HSCs, and MPPs. F. Intensity of CD5 expression among B-1a cell populations repopulated by different donor cells. (Modified from Kobayashi et al.⁴¹)

Embryonic hematopoiesis and development of the first HSCs in mouse embryos

The first blood cells observed in embryos on E7.5 are not HSCs but erythrocytes found in “blood islands” or “blood bands,” in which erythrocytes aggregate and surround the extraembryonic YS^42,43. These erythrocytes have large nuclei and express embryonic hemoglobin in contrast to the small enucleated erythrocytes with adult-type hemoglobin circulating in adult peripheral blood (PB).^44,45 Thus, these embryonic-type erythrocytes are referred to as “primitive erythrocytes” and differ from adult-type erythrocytes. This classification was based on cell morphology; it was later found that primitive nucleated erythrocytes are actually erythroblasts that correspond to erythroid progenitors in adult BM, and these primitive erythroblasts mature in the embryonic circulation into enucleated red blood cells.⁴⁶ At the early embryonic time point, there are also “primitive megakaryocytes” and “primitive macrophages” in the YS.^9,47 These macrophages have been found to migrate into the brain before the blood‒brain barrier is established and form microglia^9–12.

The production of these cells is considered to constitute the first wave of hematopoiesis and mainly occurs in the YS within a very short time period between E7.5 and E8.5. A second wave of embryonic hematopoiesis results in the production of erythro-myeloid progenitors (EMPs) in the YS from approximately E8.5 to E9.5, and these cells have colony-forming ability in vitro and express CD41 and c-kit.^48,49 These EMPs seed the FL and differentiate into mature erythrocytes, which are referred to as “definitive erythrocytes” and are essential for blood homeostasis in the embryo. As such, embryonic hematopoietic processes, especially erythropoiesis, are actively seen in the YS and then move to the FL. Therefore, it was previously assumed that the HSCs that produced these erythrocytes arose in the YS and migrated into the FL⁵⁰. However, research in other species suggested that HSCs arise in the paraaortic region of the embryo^51,52, which prompted active investigation of and debate about where HSCs originate during embryonic development^53,54. Do HSCs originate in the YS or in the paraaortic region?

The first site of HSC emergence and the endothelial origin of HSCs

Based on transplantation assays, the first adult repopulating cells are detected in the aorta-gonad-mesonephros (AGM) region at E10.5⁵⁵. In this study, only 3 out of 96 mice transplanted with AGM cells exhibited engraftment, while no engraftment occurred in the 74 mice transplanted with YS cells⁵⁵. Subsequent studies have shown that only a few HSCs exist in the AGM region of the whole embryo at E10.5 and that HSCs are also found in the YS and the placenta at E11.5^{56 57}. HSC activity has also been found in the embryonic head⁵⁸. Rybtsov et al. revised the first day of adult-repopulating HSC detection to E11.5⁵⁹ because there are almost no adult HSCs detectable by transplantation assays at E10.5, and ex vivo cultures are required for the detection of HSC activity; therefore, the cells that develop into HSCs via ex vivo culture have been termed “HSC precursors (pre-HSCs).”^59,60

Then, what cells are the precursors of the first HSCs? Round blood-like cells were observed budding from the ECs of the dorsal aorta of chick embryos⁶¹, suggesting that ECs in the dorsal aorta produce blood cells. Injections of acetyl-LDL-DIL, which labeled the ECs of the chick embryos, demonstrated that labeled hematopoietic cells were budding from the ECs, a direct indication of blood cell production by ECs. This research was extended to determine where HSCs emerged in the intraembryonic site of the mouse embryo.

Intra-aortic hematopoietic clusters (IAHCs) on the ventral side of the dorsal aorta were observed at E10.5^62,63. Cdh5 encodes VE-cadherin, an EC-specific marker. Experiments in the EC-specific lineage-tracing mouse model (Cdh5Cre: ROSA26-lacZ mice) revealed that all blood cells in the FL and adult BM were EC-derived⁶⁴. Ly6a encodes Sca-1, a marker of ECs and adult HSCs. Experiments using Ly6a-GFP transgenic embryos showed that GFP was expressed in ECs and hematopoietic cells in the dorsal aorta and that only Ly6a-GFP⁺ cells contained functional HSCs after 3 days of coculture⁶⁵. Additionally, the moment when GFP⁺ HSCs divide from aortic ECs in the embryo was observed using live imaging⁶⁶.

IAHCs express Runx1, an essential transcription factor for hematopoiesis, and only Runx1⁺ cells of the AGM region display transplantation ability⁶³. Runx1 deletion induces a lack of all definitive blood cells, including a lack of IAHCs, and embryos die at E13.5⁶⁷, suggesting that IAHCs contain HSCs. Importantly, when Runx1 is deleted specifically in ECs (as in Cdh5Cre:Runx1-flox/flox mice), all definitive blood cells and transplantable HSCs disappear, and the embryos show the same phonotype as global Runx1 knockout (KO) embryos⁶⁸. In such KO embryos, there is also a lack of IAHCs in the dorsal aorta. Interestingly, if Runx1 is deleted in hematopoietic cells (as in vav-1Cre:Runx1-flox/flox mice), HSCs are present in the embryo. The study indicated that Runx1 is indispensable for the endothelial to hematopoietic transition but is not required thereafter⁶⁸. Then, is Runx1 important for ECs to transition to only HSCs? The answer is no. Runx1 KO mice lack all erythro-myeloid progenitors in the YS⁶⁹. We also confirmed that no lymphoid cell production occurs in the Runx1 KO YS and para-aortic splanchnopleura (P-Sp) cells in OP9 cultures (unpublished data). Thus, Runx1 is essential for all blood cell production (except for primitive hematopoiesis) from ECs in the YS and P-Sp/AGM regions.

HSCs of YS origin

While these studies have provided strong evidence that HSCs arise from the ECs of the dorsal aorta in the AGM region, the presence of de novo HSCs in the YS has also been investigated because the YS is the most productive hematopoietic site before active hematopoiesis moves to the FL, and an in utero transplantation study showed the presence of HSCs in the early YS⁷⁰. Because embryonic blood cells have immature characteristics, such as a lack of MHC expression, in the embryonic environment, adult recipient mice may not be the best model to detect the HSC potential of embryonic tissues. For this reason, Yoder et al. used neonatal mice in which the embryonic environment in hematopoietic tissues (e.g., the liver) is preserved⁷¹. The neonatal recipient mice were preconditioned in utero at E17 and E18 by busulfan injection into the pregnant mother, and E10.5 YS cells were injected into the liver of neonatal mice one or two days after birth. This experiment showed successful engraftment of YS cells, and a subsequent study demonstrated that CD34⁺c-kit⁺ cells from E9.5 YS engrafted in neonatal mice but not in adults⁷². Thus, these YS-derived HSCs can engraft only in neonatal mice and are referred to as neonatal HSCs.

A follow-up study from Samokhvalov et al. also reported the presence of HSCs in the YS⁷³. They generated an interesting mouse model in which Runx1⁺ cells are temporally marked at the time of tamoxifen (TAM) injection (Mer-Cre-Mer-Runx1: ROSA-YFP or LacZ mice). They marked Runx1-expressing cells at E7.5 when Runx1 expression is limited to the YS blood band and traced Runx1⁺ progeny at later stages. Runx1-derived cells were found in the lining of dorsal aorta ECs at E10.5 together with β-gal⁺ hematopoietic cells. They also confirmed Runx1⁺ progeny in all blood lineages, including in HSCs in the FL and adult BM. Further, the study demonstrated that Runx1⁺ cells in the early YS (extraembryonic mesoderm) migrated to the embryo, not through circulation⁷⁴, and contributed to the hemogenic ECs of the dorsal aorta that become HSCs at a later stage. These studies propose an intriguing hypothesis that some HSC precursors migrate from the early extraembryonic mesoderm to the P-Sp region.

Another report supporting the idea of a YS origin of HSCs used Lyve-1 Cre reporter mice.⁷⁵ Lyve-1 is a marker of lymphatic ECs in adult mice, but it is expressed in the YS ECs from E9.5 to E11.5 and in a very low percentage of ECs in the AGM region and placenta⁷⁶. Lyve-1⁺ progeny were found in up to 50% of HSCs and HPCs of the FL⁷⁵. Although the possibility cannot be denied that only a small percentage of the Lyve-1⁺ cells in the AGM region contributed to the HSC population, this study suggests that YS ECs produce HSCs. However, Ganuza et al. demonstrated that organ cultures of Lyve-1 Cre embryos at E8.5 and E9.5 could give rise to transplantable HSCs⁷⁷, while YS organ cultures failed to repopulate the recipient mice, and they claimed that the ultimate origin of HSCs was an intraembryonic site, not the YS. This study suggested that Lyve1⁺ cells in the P-Sp region at a pre-circulation stage could become HSCs. However, YS organ cultures have often failed in other reports, suggesting the possibility that organ culture is not an optimal condition for YS HECs to mature into HSCs; thus, the possibility of YS-derived HSCs cannot be denied.

Xu et al. established a stromal cell line, named AGMS-3, derived from E10.5 AGM cells because they thought the stromal cells in the AGM region should be able to support HSC production^78,79. When they cocultured E8.5 YS and P-Sp cells with AGM-3 stromal cells and transplanted these cells into lethally irradiated mice, both YS- and P-Sp-derived cells fully repopulated the multilineage blood cells⁷⁹, showing the importance of microenvironments in HSC development. Using these specific stromal cells, HSC potential was shown in both the YS- and P-Sp.

Despite the many negative results indicating that HSCs do not develop in the YS, there are also reports that conclude that they do. Thus, the question of whether HSCs emerge de novo in the YS remains controversial. This raises another question: if HSCs also originate in the YS, do YS-derived HSCs produce different types of immune cells, as proposed in the layered immune system theory? These possibilities need to be investigated.

Seeking the lymphoid potential in the embryo

All lymphoid cells were traditionally believed to be the progeny of HSCs; therefore, the lymphoid potential found in the early YS and embryos was considered to be associated with the HSC origin. This view led to extensive examinations of lymphoid potential during embryonic development in search of the HSC origin using organ cultures or stromal cell cocultures. For example, it has been reported that cells from the YS and P-Sp regions from 10-somite pair (sp) stage (E8.5) embryos have B-cell potential following ex vivo culture on the S17 stromal cell line⁸⁰. This study showed a much higher frequency of B progenitor potential in the P-Sp than in the YS. The expanded clones from the P-Sp cells at the 20–25 sp stage (E9–9.5) showed potential for cells of multiple lineages, including B, T, and myeloid cells. The B cells produced in this culture system included both CD5⁺ IgM⁺ B-1a-like cells and IgM^lowB220^hi B-2-like cells. CD4⁺ αβT cells were also produced in fetal thymic organ cultures of P-Sp-derived cells. Other studies showed the B-cell and T-cell potential of Tie2⁺ or VE-cadherin⁺ ECs from the E9.5 P-Sp region and/or YS^81,82. While lymphoid potential in the P-Sp region was consistently detected, the lymphoid potential in the YS differed depending on the laboratory that conducted the study and the stromal cells used, raising controversies regarding whether the YS has de novo lymphoid potential.

The YS is connected to the embryo via the vitelline artery and vein, and the systemic circulation mixes the cells that are produced in the YS and P-Sp regions. The embryonic heartbeat begins at E8.25 (4–6 sp stage). Therefore, to examine the de novo lymphoid potential of each tissue, Cumano et al. separated YS and P-Sp cells at the precirculation stage and performed organ cultures using these cells⁸³. They found that only P-Sp cells produced B, CD4⁺CD8⁺, αβT, γδT, and myeloid cells on S17 stromal cells. Their subsequent work demonstrated that only P-Sp cells from the precirculation stage had multilineage repopulation ability in Rag1: IL2Rγc^−/− (Rag1γc^−/−) mice after organ culture.⁸⁴ Rag1γc^−/− mice lack natural killer (NK) cells, so these mice accept CD45⁺ cells from the AGM with 20x lower expression of MHC class I. These data strongly support that HSCs originate from the P-Sp/AGM region and not the YS.

The majority of publications report dominant lymphoid potential and HSC emergence in the P-Sp/AGM region. However, the question remains: Is there no lymphoid potential in the YS at all? The YS contains neonatal repopulating HSCs and is also transplantable when injected in utero or cocultured with AGM-S3 stromal cells. To determine if lymphocyte progenitors arise de novo in the YS, we used an Ncx1 KO mouse model, in which Ncx1 (that encodes a Na+-Ca2+ channel) is deleted. As a result, there is no heartbeat or blood circulation in the KO embryo⁸⁵. Ncx1 KO embryos develop normally until E9.5 but die around E11.5, at the time of HSC emergence. This mouse model enabled us to segregate YS-derived and P-Sp-derived hematopoiesis because there is no mixture of blood progenitors arising from the YS and P-Sp due to the absence of a heartbeat. We cocultured YS and P-Sp cells from wild-type (WT) and Ncx1 KO embryos with OP9 stromal cells, which support B-cell development⁸⁶, and found that AA4.1⁺CD19⁺B220⁺ B progenitors were produced from both YS and P-Sp areas of WT and Ncx1 KO embryos as early as E8.25. We also confirmed that this B-cell potential was detected only in VE-cad⁺CD41⁻ ECs and not in CD41⁺ hematopoietic cells until E9.75. After E9.75, we were able to produce B cells from CD41⁺ HPCs. These results indicate that ECs produce or transit to CD41⁺ B-progenitors around E9.75.

We further asked which B cell subsets would be repopulated by B progenitors derived from ex vivo culture of YS and P-Sp cells. YS- and P-Sp-derived AA4.1⁺CD19⁺B220⁺ B progenitors were transplanted into sublethally irradiated NSG neonates and repopulated only peritoneal B-1a, B-1b, and splenic MZ B cells but not FO B-2 cells. More importantly, when freshly isolated YS and P-Sp cells from E9.5 embryos were directly injected into sublethally irradiated NSG neonates, both YS and P-Sp cells repopulated peritoneal B-1 cells and splenic MZ B cells. These data indicate that the B-cell potential arising in the YS and P-Sp as early as E8.25 (and up to E9.5) is biased toward the B-1 lineage. Because HSCs are not detected in E8.5 embryos, these B cells from the YS/P-Sp region are considered to be produced independently of HSCs and arise directly from ECs.

To further validate the presence of HSC-independent B-1 cells, we utilized a unique HSC-deficient mouse model⁸⁷. Cbfβ is a co-binding factor of Runx1 and increases the affinity of Runx1 DNA binding. Cbfβ^−/− embryos show the same phenotype as Runx1^−/− embryos, including embryonic lethality at E13.5 and a lack of definitive hematopoiesis and HSCs. When Cbfβ-GFP fusion protein was overexpressed in Tek (Tie2, an EC marker)-expressing ECs, hematopoietic production by ECs was partially rescued, and the embryos survived one day after birth without HSCs⁸⁸. In the FL of these HSC-deficient embryos, B-cell progenitors were present and matured into B-1 cells upon transplantation. This study showed the presence of HSC-independent B-1 progenitors in vivo.

The HSC precursor population contains various hematopoietic progenitors, including MPPs and B-1 precursors

To assess the HSC potential of the tested cells in the early embryo, organ cultures and stromal cell cocultures were utilized to determine the transplantability of the tested cells. Because HSCs arise from HECs, HSC precursors are basically ECs and need to transit and mature into blood cells that possess functional transplantation ability. The stromal cells that support HSC development include S17, OP-9, AGMS-3, and Akt-overexpressing AGM-ECs (Akt-AGM-ECs)^78,89–91. The S-17 and OP-9 cell lines were established from mouse BM, while AGMS-3 and Akt-AGM-EC were established from the AGM stroma or ECs of mouse embryos at E10–11, the time of HSC emergence based on the idea that the AGM EC/stromal niche fosters the development of HSC precursors⁹¹.

Rybtsov et al. identified these intermediate EC-derived immature cells destined to become HSCs at E10.5 and E11.5 in the AGM region and named them HSC precursors (pre-HSCs)⁵⁹. Pre-HSCs express the EC marker VE-cadherin (VC) but lack CD45 at E10.5 (type-I cells) and become CD45⁺ after maturation (type-II cells) (Fig. 2). Other studies further purified this population as VC⁺c-kit⁺CD41^dimEPCR⁺Dll4⁺CD61⁺ cells^41,92–94. Once pre-HSCs are cultured ex vivo (using OP9 aggregation cultures or AGM-EC or OP9-DL cultures), the cultured cells express conventional HSC markers (e.g., CD150, EPCR, Fgd5) and repopulate multilineage blood cells in preconditioned recipient mice.

Figure 2. Transition from hemogenic ECs to HSCs through pre-HSC stages.

Surface markers at each stage are depicted. Created with BioRender.com.

While pre-HSCs can be detected in the E10.5 AGM region via ex vivo culture, we found that E9.5 YS and P-Sp cells repopulated B-1 cells following injection into NSG neonates.⁸⁶ Yoder et al. and Arora et al. have also found neonatal repopulating HSCs in the E9.5–E10.5 YS and/or P-Sp regions upon transplantation without an ex vivo culture step^72,95. Is there any relationship between pre-HSCs, neonatal repopulating B-1 progenitors, and neonatal HSCs in the E10.5 YS and/or AGM region? To address this question, we tested whether pre-HSCs have B-1 repopulation ability.⁴¹ Because the term pre-HSCs indicates the cells that mature into functional HSCs, instead, we refer to phenotypic pre-HSCs as pre-HSPCs. Pre-HSPCs, FACS sorted as VC⁺c-kit⁺ (VK) cells from the E10.5 YS and AGM regions were transplanted into sublethally irradiated NSG neonates. There were three engraftment patterns in the recipient mice: multilineage repopulation, T-cell and B-1 cell repopulation, and B-1 cell only repopulation (Table 1). Among the 30 mice transplanted with E10.5 AGM VK cells, approximately 30% showed multilineage repopulation, including repopulation of myeloid, T, and B cells. B-2 cells were dominantly engrafted in the spleen and BM; however, there were also donor-derived B-1a and B-1b cells in the peritoneal cavity and MZ B cells in the spleen. Because one to ten embryo equivalent (e.e.) VK cells were injected, it was not clear whether this multilineage repopulation was derived from a single HSC or from multiple progenitors. Approximately 13% of the recipient mice displayed dominant T-cell repopulation in the PB and showed donor-derived B-1a and B-1b cells in the peritoneal cavity and MZ B cells in the spleen but not FO B cells. Additionally, approximately 23% of the recipient mice did not show donor-derived cells in the PB but were actually engrafted with peritoneal B-1a and B-1b cells and splenic MZ B cells.

Table 1.

Engraftment patterns by pre-HSPC population

	E10.5 AGM and YS VK			E11.5 AGM VK41E
	HSC	T & B-1	B-1 only	HSC	MPP	B-1
PB	+++	T only	ND	+++	+++	B, M
PerC	B-2>B-1, T, M	T, B-1	B-1a, b	B-2>B-1, T, M	B-1>B-2, T, M	B-a, b
Spleen	FO>MZ, T	T, MZ	B-1, MZ	FO>MZ, T	MZ>FO, T	MZ
BM	+++	T only	No	+++	+++	ND
BM HSPC	Yes	ND	ND	Yes	ND	ND

+++: T, B, and myeloid cell (M) repopulation, ND: not detectable

These results strongly indicate that VK cells include independent T and B-1 progenitors separate from HSCs. Importantly, of 23 mice transplanted with VK cells from the E10.5 YS, one displayed multilineage, three exhibited T-cell and B-1, and five had B-1 and MZ cell repopulation. Thus, the E10.5 YS contains pre-HSCs and independent B-1 and/or T-cell progenitors. When we further purified pre-HSPCs and transplanted only ten VC⁺c-kit⁺CD41^dimEPCR⁺ (VK41E) cells into NSG neonates, two out of five animals were engrafted but only with B-1a, B-1b, and MZ B cells. This result validated that B-1-specific progenitors were present among VK41E cells and that B-1 and MZ B cells are of the same lineage, while B-1b and MZ B cells can also be derived from adult BM HSCs.

In our study, B-2 cells were dominantly reconstituted by E10.5 VK cells only when multilineage repopulation was observed; therefore, we speculate that gaining HSC ability increases B-2 potential. The B progenitor colony-forming assays that were developed by Montecino-Rodriguez⁹⁶ were used to assess E10.5 EPCR⁺ VK cells. B-1 and B-2 common progenitor colonies were detected after cocultures with AGM-ECs, while only B-1 colonies were detected without coculture, which suggested that ECs produce progenitor populations that possess only B-1 cell potential and gain B-2 and multilineage repopulation ability during the maturation process. Accordingly, when a single Fgd5⁺c-kit⁺ cell from E10–11 embryos was cocultured with AGM-ECs and transplanted into recipient mice, the mice showed multilineage repopulation, including repopulation of B-1a and B-2 cells⁹³. Since these donor cells were originally from one cell, this result indicates that the first HSCs in the AGM region possess the ability to repopulate both B-1a and B-2 cells. This result suggests that there is a clear difference between the B-1a cell repopulation ability of pre-HSCs/the first HSCs at E10.5 and the E15.5 FL HSCs that have minimal B-1a cell repopulation ability, indicating a dynamic change in the hematopoietic capacity of HSCs during development.

We also tested the blood output capacity of VK41E cells taken from the AGM region at E11.5, when the first HSCs arise in the fetus. Ten to fifty E11.5 AGM VK41E cells were injected into sublethally irradiated NSG neonates⁹⁷. Again, there were three types of engraftment patterns, but these were slightly different from those observed after E10.5 VK cell transplantation (Table 1). Among the 25 transplanted mice, 60% showed donor cell engraftment in the PB: 1) 20% of the mice displayed donor-derived multilineage repopulation; 2) 24% of the mice showed multilineage repopulation but without HSC repopulation in BM; and 3) 16% of the mice showed only B-1 and MZ B-cell engraftment with a minimal percentage of B-2 cells. Similar to the E10.5 multilineage repopulation, the E11.5 multilineage repopulation included CD11b⁺ myeloid cells, CD4 and CD8 T cells, a large number of B-2 cells, and a significant percentage of B-1a, B-1b, and MZ B cells. Importantly, donor-derived LSK cells (an enriched stem/progenitor population) were also repopulated in recipient BM, and these cells were successfully repopulated in secondary transplanted mice, which also showed donor-derived LSK cells in BM. These results suggested the presence of definitive HSCs among the E11.5 VK41E cells, and given that only 10 to 50 VK41E cells were injected, it is plausible that the B-1a and B-2 cells in the recipient mice were produced from a single HSC. The second repopulation pattern was multilineage repopulation without donor LSK cells in BM. The recipient mice showed limited percentages of donor cells in PB. Donor-derived cells included T, B-1a, and B-1b cells in the peritoneal cavity and MZ B cells in the spleen, with limited B-2 cell repopulation. In recipient BM, there were a few donor cells (<3%) but no donor-derived LSK cells. We referred to this repopulation pattern as MPP repopulation. The third repopulation pattern was repopulation of only B-1 and MZ B cells with rare B-2 cells in some cases. There were no donor-derived cells in BM. These results clearly show that E11.5 VK41E cells include independent MPPs and B-1 progenitors separate from HSCs.

In line with the results of that study, single-cell clonal culture of CXCR4⁺/⁻VK41E cells from the E11.5 AGM region showed that CXCR4⁺VK41E cells became HSCs, while CXCR4⁻ cells failed to produce HSCs but were still multipotent, indicating independent developmental processes of HSCs and MPPs from HECs⁹⁴.

In light of all the previous studies demonstrating B-1 cell potential in early YS and P-Sp cells and the ability of HSC-independent MPPs to repopulate B-1a, B-1b, and MZ B cells, we next asked whether HSCs, MPPs, and B-1 progenitors are produced independently in vivo.

HSCs do not contribute to FL hematopoiesis, including B-progenitor production

As mentioned in the earlier section, B-cell potential in the YS and P-Sp region before HSC emergence has been shown; however, this does not mean that committed lymphoid cells exist in the tissues because CD45⁺ cells do not produce lymphoid cells before E9.5.⁸⁶ Instead, ECs that can differentiate into lymphocytes exist. Then, when are the first lymphocytes detected in the embryo? The first detectable B-lymphocytes are Pax5⁺ B-lymphocytes, which have been reported in the E10.5 AGM region⁹⁸. The next earliest lymphoid progenitors have been reported as lin-CD45⁺c-kit⁺ interleukin-7 receptor (IL7R)⁺ cells in the FL at E12–13⁹⁹. While CD45⁺c-kit⁺IL7R⁻ cells have multilineage differentiation potential that has been shown by in vitro clonal culture assays, the IL7R⁺ cells produce only T or B cells⁹⁹. In line with this, lymphoid-restricted progenitors were also identified in E11.5 FL as lin⁻Flt3⁺c-kit⁺IL7R⁺ cells that express Rag1 and other lymphoid marker genes¹⁰⁰. Additionally, lin⁻AA4.1⁺CD19⁺B220⁻ B-1-specific progenitors have been detected in the FL and fetal BM²⁷. Do these cells come from the HSCs in the FL? Because only a few HSCs are detectable in whole embryos at E10–11¹⁰¹, it is assumed that the earliest lymphoid progenitors found in the early FL are produced independently from HSCs, presumably from ECs at the earlier stages.

To address the question of whether FL lymphoid progenitors are derived from HSCs or directly from HECs, Cdh5CreERT2:ROSA-tdTomato mice (iCdh5 mice) were utilized⁹⁷. In this mouse model, Cdh5⁺ ECs permanently become Tomato⁺ upon TAM injection, and Tomato⁺ EC-derived blood cells are traced as Tomato⁺ cells. After E7.5 TAM injection, 20–30% of MPPs, lymphoid progenitors, and CD19⁺ B-progenitors were Tomato⁺, while only 5% of HSCs were Tomato⁺ in the FL, indicating that these progenitors originated from E7.5 ECs, not from HSCs. When E9.5 ECs were labeled, HSCs, MPPs, and lymphoid progenitors in the E15.5 FL showed similar Tomato⁺ percentages, indicating that these cells were produced simultaneously from ECs. However, it is still possible that the first few HSCs that have just arisen from ECs undergo asymmetric cell division and produce one MPP and one daughter HSC and that all these progenitors in the FL originate from the first HSCs. A precise examination of the cell division history, such as that enabled by “DNA-recoding” studies¹⁰², would be required to address this question.

Another recent study by Yokomizo et al. showed that FL HSCs do not contribute to FL hematopoiesis using lineage-tracing mouse models¹⁰³. Hlf has recently been reported to be expressed in pre-HSCs in the AGM region as well as in human HSCs^104–107. They found that HlfCreERT2 specifically marked pre-HSCs and not EMPs in the YS and that HlfCre-derived cells contributed to most hematopoietic progenitors in the FL, suggesting that Hlf⁺ pre-HSC populations contain greater percentages of HSC-independent hematopoietic progenitors than HSCs. Furthermore, Evi1 is highly expressed in the HSCs, and Evi1CreERT2 was able to specifically label HSCs in the FL. These data clearly show that Evi1⁺ HSCs do not contribute to hematopoiesis at the FL stage and gradually produce cells of multiple lineages after birth. This means that the main FL hematopoiesis is supported by HSC-independent hematopoietic progenitors that have originated in the early embryo, in line with our data.

HSCs do not produce all of the B-2 cells

Another study by Patel et al. is also in line with these results¹⁰⁷. They used inducible Flt3Cre mice to trace FL MPP-derived cells (referred to as eMPPs) after birth because they previously found that steady-state hematopoiesis is driven by MPPs rather than HSCs.¹⁰⁸ Interestingly, when the eMPPs in the FL were labeled at E14.5 and the animals were examined postnatally, as many as 40% of T and B cells in the PB were eMPP-derived for up to 21 months after birth.

In line with these data, we also showed that E14.5 FL HSCs and P2 BM HSCs do not contribute to hematopoiesis within the first 4 weeks after birth⁹⁷ by using Fgd5CreERT2: ROSA-Tomato (iFdg5) mice, which specifically label HSCs¹⁰⁹. While the Tomato percentages of MPPs and splenic macrophages were promptly increased, splenic T cells and FO B cells gradually developed a Tomato⁺ cell population, and only up to 60% of the cells were Tomato⁺ (HSC-derived), even when the mice were 12 months old. On the other hand, a low percentage of peritoneal B-1a cells were Tomato⁺, clearly indicating that HSCs do not efficiently contribute to the B-1a cell population. Since our procedure of TAM administration by oral gavage achieved close to 100% efficiency in labeling BM HSCs, it is reasonable to conclude that B-1a cells are minimally produced by BM HSCs and that fetal-derived B-2 cells are slowly replaced by HSC-derived B-2 cells (Fig. 3).

Figure 3. EC-derived and HSC-derived lineage tracing interpretation and multiple waves.

Representative Tomato% kinetics in the EC-lineage tracing mice (A) and in the HSC-lineage tracing mice (B). Representative Tomato% of EC-derived spleen FO labeled at different embryonic stages are shown. Created with BioRender.com.

Because HSCs do not fully replace B-2 cells even one year after birth, fetal-derived B-2 cells still exist in aged mice. To confirm that these cells are of fetal EC origin, we labeled ECs at E7.5, E8.5, E9.5, E10.5, and E11.5 by injecting TAMs into pregnant iCdh5 mice and examining EC-derived hematopoietic lineages at various time points after birth. Fig. 3A shows what EC-labeling results mean. Because not all the ECs produce blood cells immediately after TAM injection and may produce HSCs at E10–11, EC-labeling includes EC-derived HSC-independent blood cells and a part of HSC-derived cells. HSC-lineage tracing study already showed its maximum output to each lineage (Fig. 3B); thus, the estimated frequency of HSC-independent B-2 cells can be calculated and showed the persistence in aged mice. While HSC labeling showed a gradual increase in Tomato % of B-2 and T cells (Fig. 3B), EC labeling showed maintained Tomato % of B-2 and T cells, indicating that EC-derived B-2 cells are gradually replaced by HSC-derived cells (Fig. 3A). Interestingly, ECs labeled at different embryonic time points showed different kinetics after birth (Fig. 3C). Thus, we obtained the following findings: 1) HSCs were labeled by all stages of ECs from E7.5 to E11.5; however, the highest Tomato⁺ cell percentages were shown with E9.5 and E10.5 EC labeling, and the same Tomato⁺ cell percentage was maintained at later time points (>6 months). 2) FO B and T cells were labeled by ECs from E7.5 to E11.5. E8.5 and E10.5 EC-derived FO B cells and T cells were maintained at later time points. 3) Peritoneal B-1a cells were marked by ECs from E7.5 to E10.5, but much less at E11.5, and the E7.5 EC-derived B-1a cell percentage declined by >6 months, while E8.5–E10.5 EC-derived B-1a cells were maintained. These results indicate that not only B-1a cells but also some adaptive immune B-2 and T cells are derived directly from HECs.

HSCs minimally produce peritoneal B-1a cells after the FL stage

Our HSC lineage-tracing mouse model revealed that postnatal HSCs labeled at P2 minimally produce peritoneal B-1a cells, up to 10% of the total cells at most⁹⁷. Although the efficiency of HSC labeling in the FL stage is not high, as many as 30% of E14.5 HSCs were labeled, and there were no Tomato⁺ HSC-derived lymphoid progenitors at birth⁹⁷. Therefore, the results of P2 HSC labeling reflect overall HSC-derived hematopoiesis, and at least, it can be said that B-1a cells are not produced by HSCs after E14.5 under steady-state conditions. In the EC lineage-tracing study, B-1a cells were well marked by E9.5–10.5 ECs. A transplantation study showed that the pre-HSC population (VK41E cells) from E10–11 contained B-1-specific precursors, MPPs and HSCs, both of which possess B-1a cell repopulation ability. Therefore, it seems that E9–11 ECs give rise to solo B-1a progenitors as well as MPPs and HSCs with B-1a production ability; however, the B-1a cell potential of HSCs seems to decline after E14.5.

HSCs have different characteristics between the fetal to neonatal stage and adult stage, including differences in gene expression profiles (Lin28b, Sox17, etc.), surface marker expression (CD34), cell cycle characteristics, and differentiation capacity²⁹. Similarly, E10–13 is a time frame when pre-HSCs develop and mature into functional HSCs; therefore, the characteristics of pre-HSCs/HSCs during this time (referred to as immature HSC, or imm-HSC) are different from those of HSCs after E14.5 FL. For example, 1) the repopulation ability of imm-HSCs is unstable, and 2) the surface markers of imm-HSCs have not yet been well defined; thus, the current markers include a heterogeneous population. 3) imm-HSCs are strongly lymphoid biased (although FL HSCs are also lymphoid biased), and 4) imm-HSCs are capable of repopulating tissue-resident immune cells that cannot be repopulated by adult HSCs, including peritoneal B-1a cells and mast cells (which will be discussed in a later section). It is important to understand the molecular mechanisms that enable imm-HSCs to mature into functional stable HSCs after E14.

Multiple waves of B lymphopoiesis in mouse embryos

The accumulated evidence indicates that there are several waves of B-lymphopoiesis. For B-1a cells, our pre-HSPC transplantation assays showed several progenitor populations: 1) solo B-1 progenitors, 2) B-1 and B-2 progenitors (minimal B-2 potential), 3) B-1 and B-2 progenitors (B-2 dominant), and 4) HSC-derived B-1a cells (minimal B-1a potential). The presence of restricted B lineage progenitors is supported by evidence that 1) YS and P-Sp ECs produce only B-1 cells⁸⁶, 2) AGM-EC cultures induce bipotent B-1 and B-2 progenitors at a single-cell level from E10–11 AGM pre-HSCs⁹³, 3) FL MPPs contain common B-1 and B-2 progenitors⁹⁷, 4) E11.5 multilineage-repopulating cells can repopulate both B-1a and B-2 cells⁹⁷, and 5) HSCs after E15.5 and postnatal HSCs do not produce B-1a cells^40,41,97. The B-1 only progenitors and B-1 and B-2 progenitors detected at E10–11 must have been produced directly by ECs. Our EC labeling at E7.5–10.5 marked both B-1a and B-2 cells after birth; therefore, it is still challenging to separate a B-1 only wave from a common B-1/B-2 wave using the EC lineage tracing system. However, a study by Montecino-Rodriguez et al. clearly separated these different types of B progenitors using PU.1 hypomorphic mouse embryos⁹⁶. In that study, PU.1 hypomorphic mice showed an 80% reduction in the expression of Spi-1 (encoding PU.1) by deleting an upstream regulatory element (URE) located 14 kb from the Spi-1 transcription start site (URE^Δ/Δ mice). URE^Δ/Δ embryos showed a lack of the earliest B-1 progenitors derived from the YS and a lack of B-2 cell development (but intact B-1 cell development) from MPPs and CLPs in neonatal BM, while MPPs and CLPs in the late FL could produce B-2 cells. Gene expression profiling revealed that fetal- and adult-derived B-1 and B-2 cell development was clearly segregated into three waves for B-1 cells and two waves for B-2 cells. Importantly, we assume that this first fetal B-2 wave corresponds to our common B-1/B-2 progenitor wave found in the E10–11 AGM region. The fetal Spi1-independent B-2 cell potential was detected after E15.5 in the FL and MPP and CLP populations, in which we found strong B-1a repopulation ability as well as B-2 cell repopulation ability upon transplantation⁴¹. It is interesting that peritoneal B-2 and splenic MZ B-cell numbers were not altered despite the severe reduction in splenic FO B cells in 6–8-week-old URE^Δ/Δ mice. Our E8.5 EC and P2 HSC lineage-tracing data showed that all peritoneal B-1a, B-1b, and B-2 cells were EC derived, while the fetal-derived B-1b and B-2 cells (but not B-1a cells) were gradually replaced by HSC-derived cells (Fig. 3). Considering that MPPs and CLPs in the E15.5 URE^Δ/Δ FL retain B-2 potential, and the majority of FL cells, MPPs and CLPs are HSC-independent progenitors, these PU.1-independent B-progenitors may be HSC-independent and contribute to peritoneal B-2 and splenic MZ B cells in early life.

Montecino-Rodriguez et al. also identified a lin-Sca-1⁺c-kit^hiIL7R⁺ (CD127⁺) population, which does not exist in adult BM, in the E12.5 YS and FL that they proposed to contain the precursors of B-1 cells. These cells are reduced in number or lacking in URE^Δ/Δ mice. We also confirmed that IL7R⁺c-kit⁺ progenitors from E12.5 FL repopulate peritoneal B-1a cells (unpublished data). This population also seems to correspond to the lin⁻Flt3⁺c-kit⁺IL7R⁺ cells that have been reported by Boiers et al¹⁰⁰ as immune-restricted progenitors. These data indicate the possibility that these cells are HSC-independent, PU.1-dependent B-1 progenitors derived from the YS.

It has been debated whether B-1a and B-2 cells are separated lineages and whether they share the same precursors and selected based on BCR signaling. Intriguingly, B-1a and B-2 common progenitors seem to exist in the AGM region. However, it remains unknown what stage of progenitors they are. Are they derived from the same ECs but separate before they commit to blood cells or even early lymphoid progenitors? If this is the case, B-1a and B-2 cells would not be selected by BCR signaling at this bifurcation point. Alternatively, can “fetal-specific CLP-like cells”, which have not yet been identified, produce both B-1a and B-2 cells? We found the presence of a common progenitor for B-1 and B-2 cells after AGM-EC coculture; however, the characteristics of this progenitor population that gives rise to B-1 and B-2 progenitor colonies remain unknown. Because there are multiple waves of B-1 and B-2 production, it is plausible that these inconsistencies might reflect the fact that populations of different origins were assessed. Additionally, the results of BCR overexpression experiments¹¹⁰ in which B-2 cells switch to B-1a cells do not contradict that B-1a and B-2 cells have different origins¹¹¹, as discussed in more detail in the review from Montecino-Rodriguez and Dorshkind in this volume.

T-cell development

In the search for the origin of HSCs, T-cell potential has also been examined in the embryo at early stages. The earliest thymus-seeding cells have been identified in thymus rudiment at E11.25¹¹². This period is almost the same as when the first HSCs emerge in the AGM region, raising the question of where these progenitors come from. To detect the T-cell potential of tissues in vitro, Notch signaling is indispensable¹¹³. Therefore, traditionally, thymic organ cultures or stromal cells expressing the delta-like proteins delta (DL)-1 and DL-4 have been used¹¹⁴. Ohmura et al. examined the multipotent progenitor potential of AGM cells using thymic organ cultures and showed that they had T-progenitor potential as well as multipotent progenitor potential¹¹⁵. As mentioned above, T-cell potential was proven in the P-Sp region but not in the YS at the precirculation stage,⁸³ while other groups have shown T-cell potential in the YS and/or P-Sp regions at E9.5^81,82.

We also utilized DL1-expressing OP9 (OP9-DL1) cells to test the T-cell potential of WT and Ncx1 KO (no circulation) YS and P-Sp cells⁶. We detected both γδT cells (including Vγ3 T cells) and aβ T cells as well as CD4⁺CD8⁺ double-positive (DP) T cells in the OP9-DL1 cell cultures from the WT and Ncx1 KO YS and P-Sp cells. These T cells were produced as early as E8.25, at a time before the heartbeat starts (precirculation stage); thus, the YS and P-Sp regions have independent T-cell potential. When the T cells that were produced from the YS and P-Sp regions were transplanted into NSG neonates, thymic repopulation was observed two weeks after transplantation. Because the YS/P-Sp culture did not include HSCs, thymus repopulation was transient, but donor-derived T cells were detected in the spleen and BM three months after transplantation. We also detected naive, memory, and FOXP3⁺ regulatory T cells in the spleen. Interestingly, the TCR repertoire of the YS-derived CD4⁺ and CD8⁺ T cells was skewed, while P-Sp-derived T cells showed broad TCR repertoires. These T cells were functional, showing proliferation and secreting IL-2 upon anti-CD3 and anti-CD28 antibody stimulation. When freshly isolated YS and P-Sp cells from E9.5 embryos were directly injected into sublethally irradiated NSG neonates, only those mice that received YS cells displayed T-cell repopulation in PB. These data provide direct evidence that solo T-cell progenitors exist in the E9.5 YS before HSC emergence and confirm that the YS is a site of de novo lymphocyte production. In addition, the T-cell potential observed before HSC emergence suggests the contribution of these cells to multiple T-cell waves in the fetal thymus and to innate-type T cells, including Vγ3 T cells in the skin and Vγ2 T cells that secrete Th17¹¹⁶.

It has been recognized that neonatal CD4⁺ and CD8⁺ T cells are distinct lineages from their adult counterparts,^117–122 as discussed in more detail by Tabilas et al. in this volume. For example, neonatal CD8⁺ T cells display more effector functions than experienced memory T cells, which are referred to as virtual memory T cells. These cells rapidly proliferate and secrete cytokines. Time stamp experiments showed that these virtual memory T cells are fetal-derived^117,118. Neonatal CD4⁺ cells are known to be biased toward Th2 cells^121,122. We further showed that the Th2 loci of CD4⁺ T cells in the fetal and neonatal thymus are highly hypomethylated and epigenetically biased toward Th2 cells. This hypomethylated status of Th2 loci is only observed during the fetal and neonatal periods.¹²⁰ These specific features of neonatal T cells that are lacking in their adult counterparts may be due to their origins (e.g., HSC-independent, YS, or P-Sp origin), and the relationship between neonatal T-cell origin and function has to be investigated.

How long do HSC-independent MPPs persist into adult life?

Recent studies, including ours, have indicated that FL MPPs are produced directly from ECs and independently of HSCs at similar embryonic time points^97,103,107. This idea is also supported by the evidence discussed above suggesting that HSCs in the FL do not differentiate into downstream progenitors¹⁰³ and that the AGM region contains HSC-independent MPPs, as shown by clonal culture and transplantation assays⁹⁷. Then, how long do HSC-independent MPPs persist into adult life and support hematopoiesis? It is more likely that hematopoiesis during the neonatal period is supported by fetal progenitors, not LT-HSCs, in the BM because our data indicate that LT-HSCs produce some MPPs and myeloid cells but not lymphoid cells. Recent reports combining clonal analysis and fate mapping studies that used inducible Flt3 labeling of E14.5 MPPs (referred to as embryonic MPPs, or eMPPs) have been suggested that eMPPs last up to two years and are responsible for 40% of the T and B cells in the adult PB¹⁰⁷. On the other hand, our data indicated that 50%, 75%, and 80–90% of BM MPPs were replaced by HSC-derived cells at 4 weeks, 3–4 months, and 1 year of age. Our data also indicated that 60% of spleen T and B cells were replaced by HSC-derived cells, which means that approximately 40% of these lymphocytes were fetal-derived. In this sense, the percentages of HSC-independent T and B cells showed agreement between these studies.

While our lineage-tracing study estimated the number of HSC-derived MPPs in adult BM, the in vivo barcoding study had limited barcoding efficiency, thus making it a challenge to estimate the actual percentage of the HSC-independent eMPP population. Furthermore, it is known that the Cre-recombinase efficiency is different between R26^LSL-tdTomato and R26^LSL-EYFP mice. When R26^LSL-EYFP mice were used for E14.5 Flt3⁺ cell fate mapping, the percentage of YFP⁺ LT-HSCs was close to zero, while the YFP⁺ MPP percentage was 7.7 and that of ST-HSCs was 5 in BM at 13 months of age. These numbers are in line with the percentage of HSC-independent MPPs in our results. Another important point is that Flt3⁺ fate mapping labels a high percentage of ST-HSCs and approximately 75% of MPPs. It is also known that the phenotypic ST-HSCs in the FL are actually not ST-HSCs and rather contain long-term multilineage-repopulating cells. Therefore, when the eMPP number declines, the labeled ST-HSCs can replenish MPPs, which will confound estimations of HSC-independent eMPPs. Another possibility is that our TAM injection into P2 Fdg5CreERT2 neonates may also mark ST-HSCs and MPPs as well as HSCs. Because FL LT-HSCs do not actively differentiate into ST-HSCs and/or MPPs at birth, it is plausible that our system might have labeled a portion of fetal-derived (presumably HSC-independent) ST-HSCs and MPPs and might have overestimated the percentage of HSC-derived progenitors.

Thus, it remains unclear to what extent lifelong HSC-independent MPPs exist. However, there is agreement that HSC-independent mature lymphocytes persist almost the entire lives of mice. It would be intriguing to investigate the different functional roles of HSC-independent and HSC-dependent lymphocytes.

Notch-independent hematopoiesis

HSC-independent hematopoiesis seems to last longer than previously believed. In the embryo, HSC-independent and HSC-dependent hematopoiesis can be segregated based on their dependence on Notch signaling.

Notch signaling plays essential roles in embryonic tissue development. During hematopoiesis in the mouse embryo, Notch signaling is critical for HSCs to be produced from ECs via endothelial to hematopoietic transition (EHT) in the dorsal aorta at E10–11^123-125. Rbpj is the target of all canonical Notch signaling pathways, and Rbpj KO embryos die at approximately E10.5¹²⁶, at the time of HSC emergence. In Rbpj KO embryos, colony-forming cells are diminished, and Runx1 and Gata2 expression is severely reduced in the ECs of the dorsal aorta¹²³. Notch1, but not Notch2, has also been shown to be essential for HSC production in the AGM region¹²⁴. Interestingly, although all hematopoietic capability was diminished in Notch1^−/− P-Sp cells, the colony-forming ability of and CD34⁺c-kit⁺ cell levels in the Notch1^−/− YS were maintained. These results suggest the different roles of Notch signaling in the YS and P-Sp regions.

Notch1-independent and Notch1-dependent¹²⁷ phases of definitive hematopoiesis were revealed using Notch 1^−/− ESC differentiation cultures. Notch1 KO did not alter primitive or definitive colony-forming abilities or hematopoiesis in ESC cultures. However, in the FL of Notch1^−/− ESC chimera embryos, Notch1^−/− ESCs failed to contribute to the HSC pool. These data indicate that erythro-myeloid progenitor production is Notch signaling-independent, while HSC production requires Notch signaling (Notch-dependent). Is Notch signaling required for lymphoid cell production?

Traditionally, lymphoid potential has been tied to HSC potential. However, as reviewed above, it is evident that lymphoid potential, particularly in the embryo, is not always linked to HSCs. Because Notch signaling is indispensable for HSC emergence, Notch signaling independence indicates HSC independence; thus, we hypothesized that HSC-independent hematopoiesis can be identified by Notch signaling independence. Our laboratory examined the details of Notch signaling-independent hematopoiesis using Rbpj KO ESCs and embryos¹²⁸. Rbpj KO ESCs displayed delayed CD45⁺ cell production, reduced numbers of colony-forming cells, and reduced expression of essential hematopoietic genes, including Runx1, Tal1, Gata2, and Gata3, and the Notch target gene Hes1. Notch signaling is indispensable for T-cell lineage commitment¹¹³. When Flk1⁺ mesodermal cells from WT and Rbpj KO ESCs were cultured on OP9-DL1 stromal cells, the WT Flk1⁺ cells gave rise to CD4⁺CD8⁺ DP T cells, whereas the Rbpj KO ESCs failed to differentiate along the T-cell lineage and instead switched to produce CD19⁺ B cells at the CD4⁻CD8⁻CD25⁻CD44⁺ DN1 stage. Rbpj KO ESC-derived Flk1⁺ cells produced a percentage of CD19⁺ B cells on OP9 cells similar to that of WT Flk1⁺-derived B cells. When these CD19⁺ B cells were transferred to NSG neonates, both WT and Rbpj KO ESC-derived B cells engrafted and matured into peritoneal B-1a and B-1b cells but not B-2 cells. The B-cell production of the E9.5 WT YS and P-Sp cells was not altered by a γ-secretase inhibitor (which inhibited Notch signaling), and Rbpj deletion in the ECs gave rise to B cells. These results indicate that B-1 cell development is Notch-signaling-independent; thus, it is HSC-independent.

As we reported that cells with B-1 potential gain B-2 potential via AGM-EC culture, which replicates a developmental niche that fosters HSC maturation and activates Notch signaling, we also induced Notch signaling using doxycycline-inducible Notch-ICN ESCs¹²⁸. In this case, doxycycline addition induced production of B-2 progenitors from ESCs. Therefore, fine tuning of Notch signaling seems to be required for the development of B-1, B-2 cells and HSCs.

The addition of a γ-secretase inhibitor to OP9-DL1 cultures prevented the T-cell lineage commitment of E9.5 WT YS and P-Sp cells and directed lineage switching of these cells to the B-cell lineage, similar to the results of the Rbpj KO ESC OP9-DL1 culture. Because the DN1 population contains early thymic progenitors (ETPs) that retain B-cell and myeloid potential in addition to T-cell potential¹²⁹, it is assumed that ETPs or MPPs can be produced in the absence of Notch signaling. An understanding of the detailed mechanisms that segregate HSC-independent and HSC-dependent blood cell development under Notch signaling regulation is key to producing HSCs from pluripotent stem cells.¹³⁰

Mast cell development

Mast cells (MCs) are a unique cell lineage residing in mucosal and epithelial tissues throughout the body and are involved in various immune-mediated conditions and processes, including allergies, infections, inflammation, and cancer progression^131–134. The developmental origin of MCs has been a long-standing unresolved question¹³⁵. Despite more than 150 years of hematology history, evidence that MCs belong to the blood lineage was reported only approximately 40 years ago. MCs were originally discovered by Paul Ehrlich more than 100 years ago (in 1877) and were believed to represent a mesenchymal lineage.¹³⁶ Kitamura’s group challenged this theory at the end of 1970, demonstrating that MCs arise from BM cells by transferring BM cells into irradiated MC-deficient mice¹³⁷. Since then, the concept that MCs emerge from BM HSCs has become the accepted paradigm in the field.

In 1983, Sonoda et al. reported that E9.5 YS possesses higher MC potential than the embryo body at E9.5 and that the YS and FL at E11.5 have MC repopulation capacity¹³⁸. Along with advances in FACS, which allows prospective cell isolation, the identification of MC progenitors in BM has been pursued and reported by several groups. Around the year 2000, Rodewald et al. reported that the BM Lin⁻Thy1^lokit^hi population contains MC progenitors¹³⁹. Later, the population containing MC progenitors was narrowed down to a Lin⁻Sca^lokit^hiCD150⁻β7integrin⁺ population¹⁴⁰, and the bifurcation point separating the MC and basophil lineages was clarified by an in vitro study¹⁴¹. Since the discovery of MC progenitors in adult BM, there has been speculation that HSC-derived MC progenitors leave the BM, circulate in the blood, and distribute themselves across peripheral tissues to complete terminal differentiation.

However, one of the confounders regarding the origin of MCs is that although MCs can be produced from adult BM LSK cells in in vitro culture very easily^141,142, postnatal BM cells do not repopulate MCs upon standard transplantation¹³. These contradictory results between in vitro and in vivo assays have raised controversy regarding the origins of MCs. Successful MC repopulation by BM transplantation requires Kit^W-sh mutant mice as recipients. Kit^W-sh mutant mice lack MCs throughout their bodies but have normal hematopoiesis^13,140,141. It is possible to regenerate MCs through the transplantation of adult BM into lethally irradiated Kit^W-sh recipients, but MCs are not regenerated when adult BM is transplanted into nonirradiated Kit^W-sh hosts. Thus, it seems that MC repopulation by adult BM HSPCs requires both the depletion of host MCs and irradiation, which is known as stress-induced HSC repopulation. In addition, we need to be more cautious in interpreting the data; the MC potential shown by the in vitro study and the actual contribution to MCs shown by the in vivo study should be separately considered.

Recent progress in developmental immunology has clarified that various tissue-resident immune cells are generated independently of HSCs, including brain microglia, various tissue-resident macrophages^9,143, peritoneal B-1a cells²⁶, some γδ-T cells^5,116, and MCs^144,145. These findings are mainly based on the use of an inducible fate-mapping system originally developed in 2007⁷³ combined with transplantation studies. In mouse embryos, multiple waves of hematopoiesis arise from ECs, including primitive erythropoiesis and microglia production at E7.5, definitive-type EMP production at E8.5–9.5⁴⁵, and HSC emergence from hemogenic ECs at E10–11.

Based on two recent studies using multiple fate-mapping models, it has been concluded that mast cell-poiesis (MC-poiesis) is mostly independent of HSCs and that MCs have dual embryonic origins^144,145. These reports used Cdh5CreERT2, CSF1RCreERT2, and Runx1MER/Cre/MER models to demonstrate the existence of multiple waves of MC production in the embryo. The early wave of MCs is derived from Cdh5⁺ ECs mainly in the YS, possibly through EMPs. Subsequently, the second wave of MCs develops mainly from the ECs in the E10.5 AGM region labeled by Cdh5⁺ cells¹⁴⁴ or E9.5 Runx1-positive cells¹⁴⁵. However, Cdh5- or Runx1-based fate-mapping assays at E10.5 or E9.5 not only label HSCs but also label all blood lineages, particularly at E9.5^143,144. Therefore, it remains unclear whether the second wave of MCs is produced directly from ECs (HSC-independent) or via the first HSCs produced in the AGM region. One of the recent growing concerns about the developmental hematology field is whether HSC-independent MPPs can contribute to postnatal hematopoiesis^97,107,146, challenging the traditional stem cell dogma. We and another group reported that HSCs and MPPs are produced independently in the AGM region at E10.5–11.5^97,146, raising the question of whether these progenitors marked by E10.5 Cdh5⁺ possess the capacity for MCs are MPPs or HSCs with MC-potential (a different property from adult-type HSCs).

Currently, it is widely accepted that adult BM HSCs cannot reproduce MCs, and recently, E12.5 FL kit⁺β7⁺ cells were identified as transplantable MC progenitors (through intrauterine transfer)¹⁴⁵. Since we have been working to identify the first-emerging HSCs in the AGM region and FL by utilizing neonatal NSG recipients, we tried to address the question of whether there is MC potential in these first-emerging HSCs. We tested the MC potential of E11.5 AGM pre-HSCs (Lin⁻VECad⁺kit⁺EPCR^bright), E12.5 FL HSCs (Lin⁻CD48⁻EPCR⁺LSK), E14.5–15.5 FL HSCs, and adult BM HSCs (lin⁻CD48⁻CD150⁺LSK).¹⁴⁷ The results clearly demonstrated that MCs were repopulated in the peritoneal cavity by E11.5 AGM pre-HSCs and E12.5 FL HSCs but not by E14.5–15.5 FL HSCs (Fig. 4A). As few as 6 sorted E11.5 AGM pre-HSCs achieved prominent chimerism, indicating that the MC potential of HSCs dramatically diminished from E11.5 to E14.5. Because our transplantation assay utilizing NSG neonates (or adults) never eradicates host MCs, host MC depletion (induced by irradiation or use of a Kit^W-sh recipient) is not a required element for the in vivo MC repopulation addressed above. Instead, it is important to precisely identify HSCs with MC potential and MC progenitors in the embryo^145,147.

Figure 4. MC-producing activity evaluated by transplantation and fate-mapping assays.

A. Sorted pre-HSCs and HSCs were transplanted, and the chimerism (%) was measured in different HSC age groups (left panel). Right panel: the relationship between transplanted HSC numbers and MC chimerism (n=4–12, modified from Yoshimoto et al.¹⁴⁷). B. The Tomato ratio of peritoneal MCs, B-2 cells, and brain macrophages to HSCs at various examined time points after TAM injection into P2 iFgd5 mice. C. The Tomato⁺ percentage in the peritoneal MCs of iCdh5 mice was measured, and the kinetics were plotted for each TAM group (n=3~9 at each point, modified from Yoshimoto et al.¹⁴⁷). D. The Tomato⁺ ratios of various peritoneal cells and brain MGs to HSCs were determined and then plotted by the day of TAM injection into iCdh5 mice (n=3–7).

Based on our results, all positive cases of MC engraftment by E12.5 HSCs were accompanied by multilineage reconstitution. Additionally, in E11.5 pre-HSC transplantation assays, we observed only B-cell and MC reconstitution in the recipient peritoneal cavity (unpublished data), suggesting that HSC-independent progenitors in the E11.5 AGM region also contribute to MC-poiesis. Tsuruda et al. recently reported the HSC-independent origins of MCs in the E10.5 AGM and YS¹⁴⁸. Although E14.5 FL HSC failed to repopulate MCs, we also confirmed that E12.5 FL MPP and E14.5 FL MNCs repopulated peritoneal MCs. Therefore, while recent reports have demonstrated the presence of MC progenitors in E12.5–14.5 local tissues (skin, tongue, etc.) and E11.5 FL^144,145, a large portion of MC progenitors exist in the FL until E14.5 and then distribute themselves in systemic tissue before birth.

To observe the contribution of HSCs to the MC pool in unperturbed conditions, an HSC lineage-tracing model is essential. While HSC-specific labeling is still challenging, we used the iFgd5 model¹⁰⁹, which has shown better labeling efficiency than other models, including Pdzkip1-, Krt-18-, Hox-b5- and Vwf-based models^149–152. First, we tried to label E12.5 FL HSCs by injecting TAM at E12.5; this resulted in a low labeling efficiency of HSCs when analyzed at E14.5, but the Tomato+ cells were limited to the LSK cell population, with comparable Tomato⁺ percentages in HSCs and MPPs.¹⁴⁷ After injecting TAM at P2, we observed that the percentage of Tomato⁺ cells in the HSC compartment reached nearly 90 (60–100%). We were thus able to evaluate the contribution of postnatal HSCs to physiological MC-poiesis. TAM injection into P2 iFgd5 neonates revealed that the physiological MC production by postnatal HSCs was less than 5% (Fig. 4B). These data indicate that HSCs contribute least to the peritoneal MC pool, in line with the results of HSC transplantation assays, resolving the longstanding controversy.

We also determined the MC potential of ECs in the embryo. To understand the timing of MC production from HECs, we administered a single injection of TAM into pregnant iCdh5 mice from E7.5 to E11.5 and examined the Tomato⁺ percentage in MCs at various time points after birth (Fig. 4C). We also determined the Tomato⁺ cell percentage of BM HSCs and calculated the Tomato ratios for peritoneal MCs, macrophages, brain microglia, and B-cell subsets compared to BM HSCs (Fig. 4D). Because iCdh5 labels ECs, the ratio close to 1 indicate that the targe cell population and HSCs were produced by ECs at a similar timing, while a ratio higher than 1.0 or much less than 1.0 indicates the target cell population and HSCs were produced at different time points (thus, HSC-independent). The Tomato⁺ ratio was the highest (>10) at E7.5 and progressively declined at each TAM injection day, which was almost the same decline curve as that of microglia (known for completely HSC-independent production) (Fig. 4D). The Tomato⁺ ratio of peritoneal B-1a cells, another HSC-independent lineage, also showed a milder reduction after each TAM injection day compared to that of MCs. As peritoneal macrophages are reported to have dual origins (fetal-derived and HSC-derived)¹⁵³, TAM injection at early time points (E7.5) resulted in a higher Tomato⁺ ratio (>>1.0), indicating an HSC-independent origin, while HSC-derived macrophages were evident after TAM injection into P2 iFgd5 mice (Fig. 4B). Fig. 4C indicates the actual Tomato⁺ cell percentage in the gated MC population based on the day of TAM injection into the iCdh5 model. TAM administration at E9.5 induced higher Tomato⁺ cell percentage of MCs than TAM administration at E7.5 or E8.5 (approximately 70% at E9.5 and approximately 50% at E7.5 or E8.5 at the age of one year). This indicates that E9.5 ECs are the main producers of peritoneal MCs. Importantly, TAM injection at E10.5 labeled approximately 30% of MCs at one year old, a substantially decreased value compared to that derived from E9.5 EC labeling. Our results indicate that ECs from E7.5 to E9.5 are the major origin of peritoneal MCs, while ECs from E10.5−11.5, possible pre-HSCs, also contribute to the MC population. These results also confirmed that the MC repopulation capacity of E10.5–11.5 pre-HSCs shown by our and Tsuruda’s transplantation experiments is certainly involved in physiological MC-poiesis in the long term^147,148.

A large portion of ECs labeled at E7.5 start to produce the earliest blood progenitors, including embryonic erythrocytes, macrophages, and MCs. In this process, a portion of labeled ECs exit to form these blood lineages, which provides the first wave of hematopoiesis. The remaining ECs continue to produce various blood cells and terminate their hematopoietic activity by E11. Li et al. reported the second and the third waves of hematopoiesis marked by E8.5 Csf1r and E9.5 Runx1 positive-lineages, respectively,¹⁴⁵ while Gentek et al., described the second wave marked by E10.5 iCdh5¹⁴⁴. Based on our novel results from combined fate-mapping and transplantation assays, we propose four waves of MC-poiesis: the first wave from E7.5 to E8.5 ECs, the second wave from E9.5 ECs, and the third wave from E10.5 EC, and the minimal fourth wave from HSCs.

Taken together, the detailed observations from our laboratory and other groups discussed above allow a model of MC development to be formulated, as shown in Fig. 5. The first early wave of MC progenitors is produced from ECs at E7.5; these cells then migrate to the FL at E12.5 and colonize tissues directly. E9.5 late EMPs abundantly produce MC progenitors that seed the FL. Whether this second wave of MC progenitors directly colonize the local tissue remains unknown. The third wave derived from E10.5–11.5 ECs includes MC production from HSC-independent MPPs (wave 3a) and from pre-HSCs/HSCs that possess MC potential and seed the FL (wave 3b). This MC potential of HSCs is transiently detectable by transplantation assays (in stress conditions) and minimally detectable in postnatal life (in unperturbed conditions). Hematopoietic progenitors possessing MC potential in the E14.5 FL include HSC-independent MPPs and β7-integrin⁺ MC progenitors. The E10.5–11.5 AGM region contains a variety of HSC-independent MPPs capable of regenerating innate and adaptive immune cells, including macrophages, MCs, B-1 B cells, and αβ T cells. Moreover, our transplant data showed the presence of multilineage repopulating HSCs with MC and B-1a cell reconstitution ability. Thus, it is assumed that the earliest HSCs emerging from ECs in the E10.5/11.5 AGM region possess the ability to regenerate multiple tissue-resident innate immune cell lineages but lose the innate program during HSC maturation in the FL, acquiring a self-renewal/long-term repopulating capacity as definitive HSCs by E15.5. The mechanism for this dramatic change in HSC capacity remains unknown and needs to be clarified. Although BM MC progenitors are thought to be one of the providers of postnatal MC replenishment based on the current concept, it can be assumed that they are derived from HSC-independent fetal progenitors and remain in the BM long-term in postnatal life. The relationship between BM HSCs and MC progenitors should be elucidated by various fate-mapping models.

Figure 5. Multiple layers of MC-poiesis.

Production of MC progenitors by HECs starts at E7.5, mainly in the extraembryonic mesoderm. Most of the cells exit from ECs, and some go to the FL; the remaining cells relocate to local tissues (wave 1). E9–10 HEC activity is observed mainly in the AGM, and various HSC-independent MPPs contribute to many lineages, including MCs. A major portion of these cells moves to the FL (wave 2). The first-emerging HSCs arising around E11.0, which are indistinguishable from HSC-independent MPPs, possess efficient MC-producing capacity (wave 3a). These cells move to the FL and mature into adult-type HSCs with loss of the innate program (wave 3b). (Modified from Yoshimoto et al.¹⁴⁷)

Conclusion: Updating the concept of fetal hematopoiesis

With the recent accumulated evidence generated using advanced technologies that include lineage-tracing mouse models, cellular barcoding techniques, clonal analyses, and single-cell transplantation assays, we propose a revised concept of fetal-derived immune cell development and an updated model of layered immune system development (Fig. 6). We discussed the ontogeny of MCs, as a typical HSC-independent cell lineage, and showed that there are many types of immune cells that arise during the fetal period in different waves, independent of HSCs, which constitute the layers of the immune system in adults. These lifelong fetal-derived immune cells include not only tissue-resident innate immune cells but also adaptive immune B and CD4⁺/CD8⁺ T cells. Thus, it may be time to consider the definition of HSCs in terms of their lifelong hematopoiesis in unperturbed conditions. The established blood differentiation tree diagram in which BM HSCs remain at the apex should be revised, and we propose an updated lineage tree (Fig. 7). Recent technologies, including single-cell multiomics and DNA recording technologies, will further uncover the unappreciated road maps of immune cell development.

Figure 6. The updated layered immune system theory.

Mast cell progenitors, embryonic lymphoid progenitors (eLPs), and embryonic multipotent progenitors (eMPPs) are produced from hemogenic endothelial cells and seed the fetal liver. HSCs are produced as a final wave of EC-derived hematopoiesis, seed the liver, and expand in the FL. While eMPP- and eLP-derived lymphopoiesis are dominant in the first 4 weeks of life, HSC-derived lymphopoiesis gradually replaces them, but fetal-derived lymphocytes persist in aged mice. Created with BioRender.com.

Figure 7. The updated lineage tree.

Our working model based on our data and Fig. 6 is depicted. Mast cells and B-1a cells directly differentiate from HECs, followed by the production of eMPPs and eLPs, which produce almost all lymphoid cells. HSCs are produced as a final wave of EC-derived hematopoiesis and gradually initiate HSC-derived hematopoiesis after birth. Created with BioRender.com.