Heterogeneity and origins of myeloid cells

Structured Abstract

Purpose of review:

Myeloid cells – granulocytes, monocytes, macrophages and dendritic cells (DCs) – are innate immune cells that play key roles in pathogen defense and inflammation, as well as in tissue homeostasis and repair. Over the past 5 years, in part due to more widespread use of single cell omics technologies, it has become evident that these cell types are significantly more heterogeneous than was previously appreciated. In this review we consider recent studies that have demonstrated heterogeneity among neutrophils, monocytes, macrophages and DCs in mice and humans. We also discuss studies that have revealed the sources of their heterogeneity.

Recent findings:

Recent studies have confirmed that ontogeny is a key determinant of diversity, with specific subsets of myeloid cells arising from distinct progenitors. However, diverse microenvironmental cues also strongly influence myeloid fate and function. Accumulating evidence therefore suggests that a combination of these mechanisms underlies myeloid cell diversity.

Summary:

Consideration of the heterogeneity of myeloid cells is critical for understanding their diverse activities, such as the role of macrophages in tissue damage versus repair, or tumor growth versus elimination. Insights into these mechanisms are informing the design of novel therapeutic approaches.

Introduction

Granulocytes, monocytes, macrophages and dendritic cells (DCs) are collectively known as myeloid cells. Most are produced by hematopoietic stem cells (HSCs), although some arise early in development prior to the appearance of HSCs. In recent years, studies employing single-cell transcriptomic (scRNAseq) and multiparametric protein (flow cytometry and CyTOF) profiling have revealed notable heterogeneity within the myeloid lineages. In this review, we will consider how heterogeneity arises when myeloid cells originate from distinct tissues (e.g. yolk sac, fetal liver, bone marrow, or spleen) or are produced at different times (e.g. during fetal development, the neonatal period, or adulthood), and how microenvironmental cues also underlie myeloid subset diversity between organs and even within them (illustrated in Figure 1). We will also consider the functional relevance of myeloid cell heterogeneity.

Figure 1. Sources of myeloid cell heterogeneity.

(A) Sites of myelopoiesis. Early in development, prior to the emergence of hematopoietic stem cells (HSCs), macrophages are produced by erythromyeloid progenitors (EMPs) in the yolk sac and populate the developing embryonic tissues. Later, HSCs arising in the AGM region and stored in the fetal liver produce monocytes, neutrophils and DCs, and monocyte-derived macrophages are supplied to tissues and organs. After birth, HSCs are stored in the bone marrow and continue to produce myeloid cells throughout life. Myelopoiesis can also occur in the spleen, especially under inflammatory conditions, and to a lesser degree in inflamed tissues. In adults, tissue-resident macrophages have diverse origins – some remain yolk sac-derived, others are fetal monocyte-derived, and some are constantly replaced by newly recruited monocytes. (B) Myeloid cell heterogeneity can reflect distinct progenitor origins in the bone marrow, premature release of immature cells (especially under inflammatory conditions), and the influence of diverse microenvironmental cues between and even within organs.

The Heterogeneity and Origins of Neutrophils and Other Granulocytes

Among the granulocytes, neutrophils are the best studied (1) and the most abundant cell type in the bone marrow, where they can be found at different stages of differentiation from hematopoietic progenitors. They are typically released into the circulation as mature neutrophils, but under inflammatory conditions immature neutrophils may be found in the circulation. At the end of their lifecycle, neutrophils return to the bone marrow.

Recent studies have demonstrated the sequential development of neutrophils from early to intermediate and late precursor cell stages with specific surface marker and transcriptomic signatures (Figure 2). An early committed neutrophil progenitor (proNeu1) expressing Ly6C and CD81 but lacking CD11b was recently characterized in the LKS⁻ CD34⁺ FcγR^hi fraction of mouse bone marrow (2), downstream of oligopotent Granulocyte-Monocyte Progenitors (GMPs), which produce both monocytes and neutrophils. Neutrophil development from proNeu1 then proceeds via proNeu2, which have begun to express CD11b and represent a transitional stage that gives rise to pre-neutrophils (preNeu) (2). preNeus have also been further subdivided into preNeu 1, 2 and 3, which are thought to represent hierarchical stages of neutrophil differentiation (3). preNeus subsequently differentiate into non-proliferative immature neutrophils (immNeu), which mature in the bone marrow to yield mature neutrophils (mNeus) (4). Human equivalents of proNeus and preNeus have also been described (2, 5).

Figure 2. Neutrophil differentiation and diversity.

Neutrophils are derived from granulocyte-monocyte progenitors (GMPs, which also yield monocytes) via neutrophil-committed proNeu1 progenitors, then proNeu2 cells that begin expressing CD11b, 3 stages of preNeus, immature neutrophils (immNeu) and finally mature neutrophils (mNeu). Blood neutrophils include PMNa (arise from immNeu and mNeu), PMNb (mNeu-derived), and PMNc (derived from both PMNa and PMNb).

Three subsets of mature neutrophils have recently been described in bone marrow and blood: PMNa, PMNb and PMNc (6). PMNa are enriched in S100a8, S100a9 and Mmp8 gene expression, whereas PMNb express interferon-stimulated genes. PMNc have characteristics of senescent cells and are enriched in genes for ribosome biogenesis, cytoplasmic translation, post-transcriptional regulation and LPS-mediated signaling pathways. In the steady-state, PMNa arise from both mNeu and immNeu, PMNb from mNeu, and PMNc from both PMNa and PMNb (Figure 2). In contrast, E. coli infection suppresses differentiation of immNeu to PMNa, and immNeu mainly differentiate into mNeu. Neutrophils can also originate from different bone marrow sites. For instance, skull and vertebral bone marrow niches have been reported to specifically supply neutrophils to the meninges (7**).

Phenotypically diverse neutrophils can be found in healthy tissues, where they can acquire tissue-specific phenotypes under the influence of tissue-derived signals (8), but neutrophil heterogeneity is even more evident under disease conditions, such as inflammation, infection or cancer, including subsets that promote tumor growth and others that are anti-tumorigenic (9–11). Interestingly, a live imaging study recently demonstrated variation in neutrophil shape and movement that permitted the definition of “behavioral landscapes” of neutrophil subsets during inflammation (12**).

Two subsets of neutrophils have been identified within the red pulp of the mouse spleen: an immature, immobile Ly6G^int population and a more differentiated and mobile Ly6G^hi population (13). Upon Streptococcus pneumoniae infection, Ly6G^hi neutrophils pull bacteria off the surface of red pulp macrophages to kill them, while Ly6G^int neutrophils proliferate to increase the number of mature neutrophils. During S. aureus infection, neutrophils with the ability to recruit additional neutrophils to fight the infection enter the lymph nodes via L-selectin (14).

Neutrophils with high expression of MHCII and the ability to activate CD4 T cells ex vivo have also been observed in uninfected lymph nodes (15). Moreover, during inflammation, neutrophils have been reported to transdifferentiate into cells with features of both neutrophils and dendritic cells, called PMN-DCs (16). They have higher expression of pattern recognition receptors and kill fungal cells more efficiently than canonical neutrophils, but they can also trigger an efficient adaptive response against fungi after antigen pulsing by inducing Th1 and Th17 responses.

To better understand heterogeneity among neutrophils, a single phenotypic spectrum ordered chronologically and named “neutrotime” has recently been suggested (17*). The authors propose that phenotypic divergence among neutrophils arises from deviation from a specific position within the neutrotime continuum due to the influence of microenvironmental cues, and that each neutrophil maintains features from the point of deviation in neutrotime and acquires features according to the sequence of signals encountered. This mechanism provides an opportunity to generate a highly diverse neutrophil repertoire without a requirement for committed developmental branches or cell subsets. Therefore, neutrophil heterogeneity could result from both intrinsic cell programming by transcription factors during differentiation and the influence of external signals.

There is also accumulating evidence that other granulocytes (eosinophils, basophils and mast cells) are similarly heterogeneous (18–20). For instance, like macrophages (discussed in more detail below), mast cells, which are also tissue-resident, have dual developmental origins from both yolk sac progenitors and fetal liver/bone marrow HSCs, and mast cells in different tissues express common as well as tissue-specific gene signatures (21).

The Heterogeneity of Monocytes and Macrophages

Macrophages are the earliest immune cells to emerge during development, with production originating in the yolk sac where they arise from erythromyeloid progenitors during primitive hematopoiesis and begin to populate the tissues of the developing embryo (Figure 1). Once definitive hematopoiesis is established in the embryo itself, monocytes are produced by HSCs, which are initially maintained in the fetal liver and later in the bone marrow. Some resident macrophages of yolk sac or fetal monocyte origin are rapidly or gradually replaced over time, whereas other persist for decades or even a whole lifetime. Thus, macrophage heterogeneity in tissues is, at least in part, a reflection of their developmental origins.

The function of macrophages can also be tissue-dependent (22, 23). Although they share many key features, macrophages in different tissues, and even within the same tissue, can have distinct gene expression profiles due to the influence of microenvironmental cues. Some microenvironmental signals are relatively stable, such as those that define tissue identity, whereas others may reflect responses to infection or injury. Recent studies have also shown that monocytes and macrophages retain “memory” of previous exposures to microbial and inflammatory stimuli, which persists due to epigenetic and metabolic changes (24). Thus, within a tissue, the current state of each macrophage reflects its ontogeny and tissue location, as well as its past and current experiences.

Additional insight into these concepts was provided by a recent analysis of mouse tissue-resident macrophages that revealed 3 broadly conserved subsets in diverse organs, including the heart, liver, lung, kidney, and brain, albeit with different relative abundances in each tissue (25**). Macrophages expressing TIMD4/LYVE1/FOLR2 (TLF⁺ macrophages) have embryonic and early fetal origins and are maintained by self-renewal, whereas CCR2⁺ macrophages are monocyte-derived and constantly replaced. The third subset lacks TIMD4, LYVE1, FOLR2 and CCR2, but expresses high levels of MHCII and appears to be more slowly and only partially maintained by monocyte recruitment. Interestingly, TLF⁺ macrophages are more variable across tissues and exhibit more sex differences in gene expression than the other two subsets, which might impact their function. Another recent analysis of human monocytes, macrophages and DCs from healthy tissues as well as inflamed tissues and tumors, revealed broad conservation of subsets across tissues, albeit with some tissue-specific features, including 3 subsets of macrophages in lung, colon, liver, breast, stomach, and pancreatic tumors (26**). The study also suggested that most tumor-associated macrophages are monocyte-derived.

Within an organ, micro-anatomic differences can underlie macrophage diversity. For instance, in the mouse sciatic nerve, two subsets of macrophages with similar developmental origins occupy distinct niches in the epineurium and endoneurium and respond differently to nerve injury (27). Similarly, in a mouse model of nonalcoholic steatohepatitis (NASH), diet-induced changes in liver macrophage subsets include the accumulation of monocyte-derived cells that resemble embryo-derived Kupffer cells (KCs) and occupy the KC niche in hepatic sinusoids, as well as a monocyte-derived subset with a distinct epigenetic landscape that is located around the central and portal veins (28). Another recent study demonstrated that KCs receive key signals from hepatic stellate cells, whereas lipid-associated macrophages are induced by lipid exposure near the bile ducts (29**).

Heterogeneity has also been observed among monocytes. 3 major subsets of hierarchically-related monocytes have previously been characterized in mice and humans – classical or inflammatory monocytes (Ly6C^hi in mice, CD14⁺ CD16⁻ in humans) that enter tissues and give rise to macrophages and monocyte-derived DCs (moDCs), intermediate monocytes (Ly6C^int in mice, CD14⁺ CD16⁺ in humans), and non-classical or patrolling monocytes (Ly6C^lo in mice, CD14⁻ CD16⁺ in humans) that remain in the circulation to maintain the vasculature. However, recent studies have demonstrated additional heterogeneity within these subsets. For instance, one CyTOF study of mouse blood monocytes revealed 6 classical, 3 intermediate and 11 non-classical monocyte clusters (30).

Studies of bone marrow monocytes have shed some light on the role of ontogeny in monocyte heterogeneity. In mouse bone marrow, CXCR4 can be used as a marker to distinguish between pre-monocytes (CXCR4^hi) and mature monocytes (CXCR4^lo) (31). Moreover, 2 independent pathways originating from GMPs and monocyte-DC progenitors (MDPs) yield classical monocytes with distinct neutrophil-like and DC-like gene expression signatures (NeuMo and DCMo, respectively) in steady-state mouse bone marrow (32, 33). Both GMP- and MDP-derived classical monocytes can yield macrophages, but only MDPs give rise to moDCs (Figure 3A), and production of monocytes via the GMP and MDP pathways can be independently controlled by microbial stimuli (33). Monocytes with neutrophil and DC gene signatures have also been observed in human and mouse lung tumors (11).

Figure 3. Monocyte and DC differentiation and diversity.

(A) Monocytes arise from GMPs and MDPs, which also produce neutrophils and cDCs/pDCs, respectively. Classical monocytes produced by GMPs have a neutrophil gene signature (NeuMos), whereas their MDP-derived counterparts (DCMos) express MHCII and can give rise to moDCs. (B) cDCs (cDC1s and CDC2s) arise from CDPs via pre-cDCs, although DC3 cells (a cDC2 subset) are thought to arise independently of CDPs from an early myeloid progenitor. pDCs arise from CDPs via pre-pDCs or from the lymphoid lineage via BLPs.

GMP – granulocyte-monocyte progenitor; MDP – monocyte-DC progenitor; MP – monocyte progenitor (GMP-derived); cMoP – common monocyte progenitor (MDP-derived); CDP – common DC progenitor; CLP – common lymphoid progenitor; BLP – B cell-biased lymphoid progenitor.

Additional heterogeneity emerges when monocytes are activated by microbial components and inflammatory mediators in the circulation or in tissues. For instance, a subset of classical monocytes expressing the IL-7 receptor subunit CD127 has been observed in the bronchoalveolar lavage fluid of COVID-19 patients and in the circulation and synovium of rheumatoid arthritis patients (34). CD127^hi monocytes are less inflammatory than their CD127^lo counterparts due to IL-7-induced STAT5-dependent chromatin remodeling, which suppresses inflammatory gene expression.

The Heterogeneity and Origins of Dendritic Cells

Dendritic cells (DCs) can be classified as three main subtypes: monocyte-derived DCs (moDCs, discussed above), plasmacytoid DCs (pDCs) and professional antigen-presenting conventional DCs (cDCs). pDCs are notably efficient at producing type I interferon during viral infections, but they can also present antigen to activate T cells. This functional diversity has recently been reported to reflect sequential phases of activation of the same cells in different locations within the mouse spleen (type I IFN production in the marginal zone, then antigen presentation in the T cell zone), rather than functional specialization of different subsets of pDCs (35). cDCs can be subdivided into two major subsets based on cell surface markers, the transcription factors that control their differentiation, and their functions (36). cDC1s (mouse – XCR1; human – XCR1, CLEC9A, CADM1) depend on IRF8 and BATF3 for their differentiation and have the capacity to present and cross-present antigens to CD8⁺ T cells and thereby induce cytotoxic responses against intracellular pathogens and tumors. cDC2s (mouse – CD11b, CD172a; human – CD1C, CD172a) share some genes with monocytes, differentiate under the control of IRF4, and present antigen to CD4⁺ T cells, thereby playing key roles in immune defense against extracellular pathogens and allergens.

Moreover, two subsets of cDC2s have been reported in human peripheral blood – DC2 and DC3 – in addition to cDC1s, pDCs, DC precursors, and DCs of monocytic origin (37, 38). Additional cDC2 heterogeneity has been reported in human and mouse lymphoid tissues, which possess cDC2 cells expressing T-bet (cDC2A) that are absent from the blood and express transcripts associated with tissue repair, as well as cDC2 cells expressing RORγt (cDC2B) that are transcriptionally related to blood DC3 and have an increased proinflammatory potential (39). cDC2A and cDC2B appear to acquire their distinct phenotypes in the periphery in response to microenvironmental signals. A CD14⁺ cDC2 subset related to DC3s that exhibits pro-inflammatory functions has also been reported to accumulate in the blood of systemic lupus erythematosus patients (40).

Subpopulations displaying hybrid phenotypes have also been described, including a subset of mouse splenic DCs hypothesized to represent a transitional state between pre-DCs to cDCs, and human Axl⁺ Siglec6⁺ DCs (AS DCs), which have cDC2 and pDC features and are related to mouse transitional DCs (tDCs) that accumulate in the lung during influenza infection (37, 41, 42*). Additionally, inflammatory cDC2s (inf-cDC2s), which migrate efficiently to draining lymph nodes and prime CD4⁺ and CD8⁺ T cell responses in several inflammatory disease models, have similarities with monocytes and cDC1s (43), and merocytic DCs (mcDCs), a subset with the ability to reverse T cell anergy when presenting peptides derived from apoptotic bodies, exhibit a unique transcriptomic and metabolic signature with characteristics of both cDC1s and cDC2s (44). In contrast, mature DCs enriched in immunoregulatory molecules (mregDCs) have been observed in human and mouse non-small-cell lung cancers and can arise from both cDC1s and cDC2s upon uptake of tumor antigens (45).

pDCs and cDCs originate in the bone marrow from common DC progenitors (CDPs), which develop from MDPs (Figure 3B). cDCs derive from committed pre-cDC precursors that leave the bone marrow and traffic through the bloodstream to seed lymphoid organs and peripheral tissues, where they finally differentiate into DCs. Pre-cDCs include already committed pre-cDC1s and pre-cDC2s (46). Pre-pDCs have also been identified, but while they had previously been thought to originate from CDPs, recent data also indicates a lymphoid origin via a Ly6D⁺ B cell-biased lymphoid progenitor (BLP) with differentiation independent of the myeloid lineage (47). Recent studies also provided evidence that DC3s develop along an IRF8^low trajectory from a DC3-restricted progenitor that arises from an early myeloid progenitor, in contrast to pDCs, cDC1s and DC2s, which derive from IRF8^high progenitors (38, 48).

Interestingly, as discussed above, moDCs have distinct origins from cDCs and pDCs, but a scRNAseq study reported 7 clusters of moDCs differentiated in vitro from human peripheral blood CD14⁺ monocytes, including 5 that resembled cDC2s (three DC2-like and two DC3-like) (49*). Thus, cells with distinct origins can ultimately have similar characteristics, presumably due to overlapping differentiation programs.

Conclusion

Collectively, a growing body of literature has begun to reveal the mechanisms underlying myeloid heterogeneity, including distinct developmental origins and diverse microenvironmental signals. Additional mechanisms that require further attention include the impact of sex differences – both sex chromosomes and sex hormones, as well as aging – both cell-intrinsic mechanisms such as clonal hematopoiesis, and the extrinsic influence of the more inflammatory microenvironment during “inflammaging”. Moreover, plasticity is a key feature of myeloid cells so phenotypically distinct cells, especially those that arise under disease conditions, may represent different and potentially reversible activation states rather than distinct subsets. It may also be possible to override the impact of origins and memory of prior exposures. Thus, one important priority for the field is to define how subset identity is maintained at the level of chromatin organization and whether deleterious features could be manipulated by selective targeting for therapeutic purposes in disease-specific contexts.

Key points.

• Single cell profiling has revealed that neutrophils, monocytes, macrophages and DCs are more heterogeneous than previously appreciated.

• Heterogeneity is in part a reflection of diverse origins, including spatial and temporal variation in subset production during development and after birth, and distinct progenitors and differentiation programs.

• Heterogeneity can also arise due to variation in microenvironmental signals between and within organs, and during homeostasis versus in response to infection, inflammation or tumors.