Layered immune system development in mice and humans

It is fitting to begin the introduction to this volume of Immunological Reviews by discussing a 1986 review in this journal from Herzenberg and colleagues which introduced the concept of layered immune system development.¹ That contribution summarized functional differences between conventional B cells, which participate in adaptive immune responses and constitute the majority of B lymphocytes in mouse tissues, and a minor population of Ly-1 B cells that are innate-like effectors. Ly-1 B cells are now referred to as B-1 B cells while conventional B cells are designated as B-2 B cells. Based on data showing that neonatal bone marrow cells reconstituted B-1 B cells in irradiated recipients but adult bone marrow could not do so, B-1 and B-2 B cells were proposed to be distinct developmental lineages. This view was later reinforced by data from Hardy and Hayakawa, who discovered B-1 B cells in Leonore Herzenberg's laboratory,² showing that fetal liver pro-B cells could reconstitute cells with a B-1 B cell phenotype in immunodeficient mice while adult bone marrow cells failed to do so.³

The developmental distinctions between B-1 and B-2 B cells were the basis for a now widely cited 1989 commentary in Cell by Leonore and Leonard Herzenberg in which they presented the basic tenets of layered immune system development.⁴ Their layered immune system hypothesis proposed that the various types of lymphocytes that constitute the adult immune system developed in waves from distinct progenitors that emerged at different times during development. B-1 B cells emerged in the first, most primitive wave of development along with fetal erythrocytes and selected γδ T cells. A subsequent wave produced the so-called Ly-1 B sister population. The B-1 B cells in these two waves would now be referred to as B-1a and B-1b B cells which are known to function differently.^5-7 A third wave generated conventional B and T cells from self-replenishing progenitors in postnatal bone marrow throughout life. Studies conducted over the past 30 years have validated many aspects of the layered immune system hypothesis and suggested that the layering of immune system development is more extensive than initially envisioned by the Herzenbergs. The contributions that form this edition of Immunological Reviews provide a comprehensive summary of the evidence for developmental layering of many if not all lineages of the innate and adaptive immune system.

There is no general consensus regarding the number of waves of immune and hematopoietic system development that occur in the fetus and adult. However, the emergence of hematopoietic stem cells (HSCs) and lymphoid and myeloid progenitors in mice can be considered in the context of three broad developmental windows.

1. Pre HSC hematopoiesis: It has been recognized for decades that the extra-embryonic yolk sac is a site of early hematopoiesis and that red blood cells develop in that tissue prior to the emergence of HSCs,^{8, 9} which in the mouse occurs at embryonic day (E) 10.5. However, progenitors with myeloid potential also arise in the early yolk sac, and it is now established that their macrophage progeny are retained long-term in the adult as brain microglia, liver Kupffer cells, skin Langerhans cells, and alveolar macrophages.^{10, 11} The erythro-myeloid progenitors (EMPs) that arise as early as E8.25 in the yolk sac also generate mast cells, some of which are maintained in the adult. The contributions from Kobayashi and Yoshimoto¹² and from Chia et al.¹³ address the origin of mast cell progenitors during this early wave of yolk sac hematopoiesis as well as at later stages of development. In addition to discussing the origins of mast cells in different waves of hematopoiesis, the comprehensive chapter from Chia et al. reviews the impact of mast cells on various developmental and pathologic processes. Their review is a must read for those interested in mast cell biology. The Herzenberg's 1989 perspective⁴ proposed that many perplexing aspects of myeloid development can be clarified when viewed from the perspective of layered hematopoietic development, and the information contained in these reviews indicates they were correct.

Additional studies in mice surprisingly revealed that in addition to erythroid and myeloid cells, progenitors in the pre-HSC yolk sac can generate αβ and γδ T cells and B-1, but not B-2, B cells. Many of these revelations were initially made by Momoko Yoshimoto in collaboration with Mervyn Yoder and subsequently in her own independent laboratory and are summarized in the review by Kobayashi and Yoshimoto.¹² The degree to which B and T cells derived from this initial wave of lymphopoiesis contribute to fetal and/or adult immunity remain to be determined.

2. Mid-gestation hematopoiesis: Multiple hematopoietic populations are present in the mid-gestation fetus as discussed by Soares-da-Silva et al.¹⁴ and other reviews herein. These include HSCs, thymus seeding progenitors, B-1 and B-2 progenitors, and innate lymphoid progenitors (ILCs). The possibility that ILCs, and those in the ILC3 family in particular, emerge in separable waves of development are discussed in detail in the chapter from Van de Pavert¹⁵

It is not always easy to define the origin of the various progenitors present in the mid-gestation fetus, as some may have been generated in the pre-HSC period described above while others are likely the progeny of HSCs that emerge from hemogenic endothelium during this mid-gestation period. Further complicating the picture are recent studies concluding multi-potential progenitors (MPPs) arise from hemogenic endothelium in a wave(s) of development distinct from that of HSCs.^{16, 17} A major challenge will be to resolve the number of waves of lymphoid and myeloid development that arise within narrow time frames during mid-gestation and catalog the various stem and progenitor populations that are produced. The reviews from Montecino-Rodriguez and Dorshkind¹⁸ and Soares-da-Silva et al.¹⁴ which define several distinct waves of B and T cell development are relevant in this regard.

3. Adult hematopoiesis: A distinguishing characteristic of this period is the production of conventional B and T cells, while the generation of various innate-like lymphocytes either does not occur or is highly attenuated. For example, although γδ T cells can be generated in the adult thymus, the production of those that utilize the Vγ3 family is restricted to fetal hematopoiesis. Furthermore, the potential of adult marrow stem and progenitor cells to produce B-1a B cells is limited as discussed in reviews from Montecino-Rodriquez and Dorshkind¹⁸ and Koybayashi and Yoshimoto.¹² The latter reviews discuss the controversy regarding postnatal B-1 B cell development in light of studies showing the potential of adult marrow to generate those cells. The issue is not whether B-1 B cells, and B-1a B cells in particular, can develop from precursors in adult marrow but the efficiency with which they do so, and those reviews make the point that B-1 output from adult precursors is significantly attenuated compared to fetal progenitors. It is tempting to speculate that the few B-1 B cells generated in adult marrow are the progeny of a few fetal stem and progenitor cells that survive long-term after birth.

It has been suggested that adult bone marrow hematopoiesis does not adhere to the canonical model of hematopoiesis, which proposes that HSCs at the head of the hematopoietic hierarchy are responsible for all adult blood cell production. Kobayashi and Yoshimoto¹² review recent results from their group and the Carmago laboratory¹⁶ indicating that the fetal derived MPPs that arise distinct from HSCs as discussed above are retained in adult bone marrow long-term and make a significant contribution to adult lymphopoiesis for up to 2 years after birth. These recent findings are interesting in view of earlier barcoding studies indicating that progenitors, and not HSCs, are responsible for most day to day blood cell production.^{19, 20} However, this model has been challenged.²¹ These observations indicate that additional studies are needed to clarify how steady state blood cell production in the adult is sustained.

The results discussed above were obtained from analysis of mice, but there is evidence that layering of immune system development also occurs in humans as presented in three reviews in this issue of Immunological Reviews. Sanchez et al.²² discuss single cell sequencing results that support the existence of several distinct functional waves of γδ T cell development in the human fetal thymus. They also review the role of Lin28b and additional intrinsic and extrinsic factors in driving the fetal γδ T cell receptor repertoire in human γδ thymocytes. The contribution from Tabilas et al.²³ reviews evidence that the functions of CD8 T cells in adults is linked to when they were produced. For example, CD8 T cells in neonatal mice have rapid innate-like functions and those in adult animals have slower adaptive characteristics. Their chapter provides an overview of single cell sequencing results which support the existence of distinct ontogenic populations of CD8 T cells in humans. Burt and McCune²⁴ present a detailed review that also supports the layered development of human CD8 as well as CD4 T cells.

However, whether the layering of immune system development in humans is as extensive as in mice is unclear. For example, it has been challenging to identify human B-1 B cell progenitors. Until it is possible to do so, whether these cells emerge in distinct waves, if at all, in humans remains to be determined. Sun et al.²⁵ discuss the difficulty in definitively fate-mapping lymphoid cells in humans back to their developmental origin. As a result, they point out that, despite suggestive literature, it is difficult to conclude that human yolk sac progenitors give rise to lymphocytes in an HSC independent manner as has been shown in mice.

In fact, studies of early human lympho-hematopoiesis are challenging, because it is not easy to obtain human fetal tissues, particularly at precisely timed stages of development as discussed by Burt and McCune.²⁴ In addition, they point out that the sophisticated lineage tracing models that have been used in mice, discussed in the various chapters herein and the review from Van de Parvet¹⁵ in particular, are not available. That is why various in vitro models, including the generation of HSCs and progenitors from embryonic or induced pluripotent stem cells and systems such as the artificial thymus organ (ATO) cultures^{26, 27} will be important for defining the developmental potential of fetal and adult derived stem and progenitor cells as further discussed by Sun et al..²⁵ The ATO model is a particularly powerful system as it generates the full range of human T differentiation, including production of mature CD4⁺ and CD8⁺ T cells, and allows comparisons of development from fetal and adult hematopoietic stem and progenitor cells (HSPCs). For example, in contrast to postnatal thymopoiesis, terminal deoxynucleotidyl transferase (TdT) expression is absent during T cell differentiation from pluripotent stem cells in ATOs, reflecting a previously reported property of fetal lymphopoiesis.²⁸

If multiple populations of stem and progenitor cells with distinct developmental potential arise in different waves of fetal and adult hematopoiesis, it is reasonable to assume that they will exhibit differences in gene expression. The review from Montecino-Rodriguez and Dorshkind¹⁸ discusses how the genetic regulation of mouse B and T cell development within and between waves of fetal and adult lymphopoiesis differs. There are multiple waves of murine thymopoiesis as discussed in detail by Soares-da-Silva et al.¹⁴, and MacNabb and Rothenberg²⁹ present a masterful review that compares the transcription factor networks involved in T cell differentiation in the fetal and adult mouse thymus. Their review is distinguished by an elegant genetic analysis of genome-wide transcription data that has become available in recent years. As noted above, the single cell human sequencing studies have also revealed differences in gene expression in waves of human γδ T cell development as discussed by Sanchez et al.²²

At least two models can be formulated to explain the layering of immune and hematopoietic system development (Figure 1). The “Multiple Stem/Progenitor Cell model” proposes that multiple types of stem and progenitor cells with distinct developmental potential arise in overlapping waves of development. For example, Beaudin et al. identified a transient fetal HSC with the potential to produce innate-like lymphocytes (eg B-1 B cells, Vγ3⁺ γδ T cells) distinct from adult HSCs which primarily produced conventional B and T cells.³⁰ Their data provide support for the Herzenberg's prediction that “several types of hematopoietic stem cells that have evolved sequentially and function at specified times during development” exist. It is interesting, particularly in view of the detailed discussion by Soares-da-Silva et al.¹⁴ about when stem cell expansion in the fetus occurs, that the fetal HSCs described by Beaudin et al. were more proliferative than adult stem cells.

An alternative “Molecular Layering model” is presented in the review from Burt and McCune.²⁴ Based on single cell sequencing of human cord blood HSPCs and T cells, those investigators concluded that the stem and progenitor cell pool in the fetus is relatively homogeneous and that, at least from the perspective of T cell development, “the transition from human fetal to adult T cell identity and function does not occur due to a switch between distinct fetal and adult lineages. Rather, recent evidence from single cell analysis suggests that during the latter half of fetal development a gradual, en masse progressive transition occurs at the level of hematopoietic stem-progenitor cells (HSPCs) which is reflected in their T cell progeny”. This model is consistent with results from a mouse study by Kristiansen et al.³¹ who used a cellular barcoding approach to track individual HSPCs in the fetus and adult and showed that stem cells in the fetus could generate innate-like B-1 and selected γδ T cells and this potential was lost in the adult. These differences were linked to the expression of the Lin28b RNA binding protein, which is expressed in fetal and downregulated in adult HSCs. Interestingly, re-expression of Lin28b in adult HSCs allowed them to now produce innate like B and T cells. These observations raise the question of what regulates changes in gene expression in a fetal HSPC as it takes up residence in the adult? One possibility, as discussed by Burt and McCune,²⁴ is that HSPCs possesses an internal clock that coordinates changes in gene expression over time. Alternatively, changes in the environments stem cells occupy may trigger changes in gene expression.

In some cases, it is possible to explain the existence of particular fetal progenitors with either model. For example, our group defined a murine B-1 B cell restricted progenitor which we discuss in detail in our review in this volume.³² These progenitors are primarily present in fetal tissues and only at very low levels in adult bone marrow.³² If the multiple stem/progenitor cell model is considered, B-1 progenitors could arise directly, such as in the early yolk sac or in the mid-gestation embryo, in a stem cell independent manner. They, along with B-2 B cells and various myeloid progenitors could also be the progeny of fetal MPPs or HSCs. From the perspective of the molecular layering model, B-1 progenitors could be the progeny of a homogenous population of fetal HSCs that express Lin28b. However, as the expression of Lin28b and other fetal lineage specifying genes is downregulated, stem cells no longer generate B-1 and other innate-like progenitors or do so with significantly reduced efficiency. However, in other cases only the multiple stem/progenitor model can explain experimental results. For example, if distinct populations of HSCs and MPPs that arise independent from one another exist, their origin is difficult to explain based on the molecular layering model.

When attempting to reconcile data, it is important to consider that the two models are not mutually exclusive; in this regard, the data that allowed formulation of a particular model did not exclude other possibilities. In addition, the types of stem and progenitor cells that arise in mice and humans may not be identical. As noted, despite an extensive literature describing the preferential prenatal development of B-1 B cells in mice, fetal B-1 development in humans has not been demonstrated, and whether stem cell independent yolk sac lymphopoiesis occurs is unclear. So, while the molecular layering model may not explain all of the cell types that arise during mouse fetal development, it may be sufficient to do so in humans.

Many of the mutations that underlie various hematopoietic malignancies arise in utero, raising the question: do the properties of fetal progenitors, such as high proliferation rates and distinct patterns of gene expression predispose to the development and progression of particular leukemias? The review from Mendoza-Castrejon and Magee³³ presents a comprehensive analysis of how layered immunity may instruct cell fates that underlie leukemogenicity. As those authors note, “if we can identify mechanisms that connect layered immunity to layered leukemogenicity, we can potentially identify age-specific therapeutic vulnerabilities to treat pediatric leukemia patients.” This review is noteworthy in that it presents a detailed discussion of leukemogenic mutations and how these may trigger leukemia at selective ages.

It is now evident that distinct differences between fetal and adult hematopoiesis exist and that the emergence of different immune effectors occurs in waves. However, there is still much to be learned before consensual models, which may differ between species, can be formulated. How many waves of hematopoietic development exist? Which stem and progenitor cells function only transiently in the fetus and neonate and which are maintained long-term? Further studies that increase the understanding of how steady state adult hematopoiesis is sustained are also needed. One of the challenges in reconciling existing data is that different experimental approaches, as well as an inconsistent terminology when defining waves, were used, and some studies focused on mice while others utilized human cells. Furthermore, the results from lineage tracing studies are not always easy to interpret. Extensive discussion and collaborations between laboratories will be important as the field moves forward, and the editors hope the issues raised in this issue of Immunological Reviews will stimulate such interactions.

The authors declare no conflicts of interest.