Hematopoiesis and Cardiovascular Disease

Abstract

A central feature of atherosclerosis, the most prevalent chronic vascular disease and root cause of myocardial infarction and stroke, is leukocyte accumulation in the arterial wall. These crucial immune cells are produced in specialized niches in the bone marrow, where a complex cell network orchestrates their production and release. A growing body of clinical studies has documented a correlation between leukocyte numbers and cardiovascular disease risk. Understanding how leukocytes are produced and how they contribute to atherosclerosis and its complications is therefore critical to understanding and treating the disease. In this review, we focus on the key cells and products that regulate hematopoiesis under homeostatic conditions, during atherosclerosis, and after myocardial infarction.

Introduction

Cardiovascular disease (CVD) is the leading cause of death worldwide ¹. Despite major advances in health education, drug therapy, and interventional treatments, an estimated 17.9 million people worldwide died from CVD in 2016, representing 31% of all global deaths that year ¹. In most cases the underlying disease is atherosclerosis, a chronic inflammatory process in the vessel wall characterized by progressive accumulation of lipoproteins and leukocytes. Ruptures or erosions of atherosclerotic plaques lead to fatal events such as myocardial infarction and stroke, which together account for 85% of all cardiovascular deaths ¹.

Atherosclerosis has long been considered a cholesterol storage disease in which cholesterol and thrombotic material are passively deposited in the artery walls ^2,3. Over the last three decades, however, it has become clear that immune cells and inflammatory processes are critical to atherosclerosis and its complications, leading to the modern understanding of atherosclerosis as a lipid-driven inflammatory disease ^4–7. Various immune cells have been identified as crucial contributors to atherosclerotic lesion initiation and progression as well as the course and outcome of its major complications ^4–7.

Immune cells arise from hematopoietic stem cells (HSCs) in a process called hematopoiesis. Every day, healthy adults produce ~4–5 × 10¹¹ new blood cells to maintain homeostatic levels ⁸. HSCs give rise to all blood cell types and physiologically reside in a specialized bone marrow microenvironment referred to as the hematopoietic stem cell niche ^9,10. As a result of a tightly controlled proliferation and differentiation process, mature leukocytes exit the bone marrow in response to various stimuli, enter the circulation, and reach their target tissues via adhesion and diapedesis. While the hematopoietic supply of inflammatory immune cells is crucial for CVD development, CVD in turn strongly affects hematopoiesis. Common cardiovascular risk factors such as hyperlipoproteinemia, arterial hypertension, and diabetes mellitus considerably alter hematopoietic processes. Likewise, atherosclerosis and myocardial infarction induce strong effects on hematopoiesis. Here, we review recent discoveries regarding the cellular composition and molecular regulation of the hematopoietic stem cell niche and hematopoietic differentiation during steady-state and cardiovascular disease.

Hematopoiesis

Hematopoiesis is the process by which a small pool of self-renewing pluripotent hematopoietic stem cells (HSCs) produce red blood cells, platelets, and all leukocyte types. Only 1 of about 10,000 bone marrow cells is an HSC ^11,12 and these cells are comparatively quiescent with fewer than 5% of them being in cell cycle at any point in time. Adult humans have an estimated 3,000–10,000 HSCs with an estimated division rate between once every three months to once every three years ¹³. The HSC cell pool is divided into self-renewing long-term HSCs and transiently self-renewing short-term HSCs. To maintain the stem cell pool, some HSCs proliferate via asymmetric division in which one daughter cell remains an HSC and only the other differentiates into a more restricted progenitor with limited self-renewal capacity. These restricted progenitor cells proliferate robustly and give rise to most cells under steady-state conditions ^14–16. To maintain homeostasis and ensure adequate blood cell production under steady state and stress conditions, the balance between HSC self-renewal and differentiation is strictly controlled by a complex network of HSC-intrinsic mechanisms, such as transcriptional regulation, epigenetic modification and metabolic adaptation, as well as by extrinsic factors ^9,10 including long-distance signals from outside the bone marrow ^17,18 and local signals produced in the bone marrow microenvironment. Due to its crucial role in maintaining HSCs, this local microenvironment has been termed the “hematopoietic stem cell niche”, a concept first proposed in 1978 by R. Schofield ¹⁹.

The hematopoietic stem cell niche

Location and anatomy of hematopoietic niches

During embryogenesis, hematopoiesis occurs in so-called blood islands within the yolk sac before shifting to the spleen, liver, and lymph nodes. Once the bone marrow has developed, hematopoietic stem cells seed the newly formed niche, where they begin to produce cells in a process called definitive hematopoiesis ²⁰. In response to strong hematopoietic stress, hematopoiesis can transiently occur in extramedullary sites including the adult spleen and liver ^9,21,22.

Bone surfaces are covered by the heavily vascularized and innervated periosteum ^23,24. Among the innervations, sympathetic nerve terminals enter the bone marrow alongside arteries and continue down to the arterioles ^17,18, where they form a structural network with neural-glial antigen- (NG2) and nestin- (Nes) positive perivascular stromal cells called the “neuro reticular complex” ^25–27. While sensory nerves also innervate bone and bone marrow ^18,28, the parasympathetic nervous system does not appear to broadly innervate the bone marrow ^29,30. In long bones, a central artery typically enters the marrow cavity through a nutrient canal, branches into smaller ascending and descending arteries, and eventually extends into thin-walled arterioles that run near the endosteum following the long axis of the bone ^25,31. These transitional vessels connect the arterioles with the extensive sinusoidal network that drains into a central venous sinus, which in turn empties into nutrient veins that leave the marrow via nutrient canals ^32,33.

The perivascular HSC niche

The discovery of histological markers that unambiguously identify HSCs in tissue sections was a critical step for analyzing bone marrow niche microanatomy and cellular composition ^34,35. HSCs stain positive for CD150 and negative for CD48 and CD41, enabling their identification with a two-marker protocol ^34,36–38. Histological analyses of bone marrow and spleen revealed that almost all CD150⁺CD48^–CD41^–Lineage^– cells reside within five cell diameters of a sinusoid and many of them are localized directly adjacent to the sinusoid ^34,39.

The identification of additional highly specific genetic HSC markers, including homeo-box b5 (Hoxb5) ⁴⁰ and α-catulin (Ctnnal1) ⁴¹, led to the development of α-catulin-GFP and Hoxb5mCherry knock-in mice, which further improved our understanding of the HSC niche. A marker combination of α-catulin-GFP and KIT (the receptor for stem cell factor (SCF)) has been used for deep confocal imaging of optically cleared bone marrow, allowing for 3D reconstruction of large bone marrow segments ⁴¹. This approach revealed that ~80% of dividing and non-dividing HSCs reside in close proximity to a sinusoid and nearly all HSCs are in contact with leptin receptor⁺ (LEPR⁺) and CXC chemokine ligand 12 ^high (CXCL12^high) niche cells. The remaining HSCs are located near arterioles (10%) and transition zone vessels (10%) with only few HSCs residing adjacent to the endosteum ⁴¹. Overall, there is ample evidence that HSCs reside in perivascular niches primarily around sinusoids and that the majority of these HSCs are quiescent ⁴².

Identifying and characterizing niche cells

The earliest approach to functionally identify which cell types form the HSC niche was to ablate specific cell populations and analyze the effects on HSCs. Such interventions typically activated and/or depleted HSCs ⁴², making it difficult to distinguish between direct and indirect effects the ablated cell population had on HSCs. To overcome these issues, mouse models were engineered to specifically trace and deplete individual niche factors from defined cell populations. These models, combined with highly specific HSC markers, advanced imaging techniques, as well as functional analyses of the marrow cells by transplantation analysis, enabled the identification of cellular niche constituents, niche factors, and niche factor receptors. Among the non-hematopoietic cell types that constitute the bone marrow niche are adipocytes, neurons and glial cells, osteoblastic cells, perivascular mesenchymal stem/stromal cells (MSCs), and endothelial cells. However, hematopoietic cells, including megakaryocytes, macrophages, and T cells, can also influence hematopoiesis.

Adipocytes

Accumulating evidence suggests that adipocytes are negative regulators of HSC function. First, over the course of human life, hematopoietic red bone marrow is progressively replaced by fatty yellow marrow with far less hematopoietic activity ⁴³. Second, the adipocyte-derived protein adiponectin reduces hematopoietic progenitor proliferation in culture ⁴⁴ and hematopoietic stem and progenitor cell (HSPC) numbers negatively correlate with adipocyte content of different bones – adipocyte-free thoracic vertebrae versus adipocyte-rich tail vertebrae – in mice ⁴⁵. Third, a recent study has shown that an adipocyte progenitor negatively affects hematopoiesis and bone healing ⁴⁶. However, adipocytes may also be important to the niche’s protective function, as has been shown during dietary restriction ⁴⁷.

Neurons and glial cells

Sensory and sympathetic nerves that innervate the bone marrow can regulate hematopoiesis ^24,48,49. Sympathetic nerve terminals release noradrenaline that stimulates β3 adrenergic receptors (ADRβ3) on perivascular MSCs. The noradrenergic stimulation attenuates production of the hematopoietic retention factor CXCL12 in MSCs, leading to the release of HSCs, progenitors, and mature leukocytes from the bone marrow ⁵⁰. This axis controls circadian oscillations of HSC mobilization as well as acute and chronic stress and emergency responses ^27,48,50,51. Importantly, catecholamines can also activate adrenergic receptors on various hematopoietic cells, including human HSPCs, thereby regulating their migration and engraftment behaviors ⁵². Non-myelinating Schwann cells that encircle nerve fibers also influence hematopoiesis by inducing HSC quiescence via transforming growth factor-β (TGFβ) signaling ⁵³.

Osteoblastic cells

Osteoblastic cells were the first non-hematopoietic cells described as direct regulators of the hematopoietic stem cell niche ^54,55, although recent evidence suggests that osteoblasts’ effect on HSCs might not be direct. Specifically, imaging analyses found few HSC in contact with osteoblasts, and studies using genetic and pharmacological tools to decrease or increase osteoblasts frequently had no acute effect on HSC numbers ^9,10. Furthermore, conditional deletion of osteoblast-derived CXCL12 and SCF, factors crucially required for HSC maintenance ^56–61, had very limited impact on HSC function and numbers ^62,63. Finally, studies have questioned the importance of the adhesion molecule N-cadherin, which was once thought to be the key mediator of HSC-osteoblast interaction. Such studies could neither confirm its expression on HSCs nor detect any effect on hematopoiesis, HSC numbers, and function following N-cadherin deletion from osteoblastic cells and HSCs ⁹.

Nevertheless, there are still compelling arguments that osteoblasts have an indirect role in HSC regulation. When conditional deletion of osterix prevents chondrocytes from differentiating to osteoblasts, hematopoiesis is almost completely abrogated in the HSC-rich metaphysis ^64–66. Likewise, transplanted mesenchymal progenitors can create donor-derived ectopic bones that recruit host-derived blood vessels and HSCs in vivo ^67,68, suggesting that mature osteoblastic cells may be important contributors of structural components required to form hematopoietic niches. Osteoblastic cells also seem to be involved in maintaining downstream hematopoietic progenitors, primarily from the lymphoid lineage ^10,69.

Perivascular mesenchymal stem/stromal cells (MSCs)

The location of HSCs around blood vessels motivated extensive analyses of the stromal cells in this region. Bone marrow mesenchymal stem cells are rare stromal cells, with the ability to self-renew and differentiate into osteoblasts, chondrocytes, and adipocytes ^67,68,70. Mesenchymal stromal cells, the descendants of mesenchymal stem cells, are abundant around blood vessels throughout the bone. In mice, perivascular MSCs expressing typical marker proteins such as nestin-GFP ²⁷, leptin receptor ⁶², PRX-1-Cre ⁶³, osterix-Cre ⁶³, inducible Mx-1 Cre ⁷¹, and CXCL12-GFP ⁵⁹ can regulate HSCs. An equivalent in humans are CD146-positive skeletal stem cells that synthesize the HSC niche factors CXCL12 and SCF ⁶⁸. Overall, there is convincing evidence that MSCs have a crucial role as constituents of the perivascular HSC niche.

The first peri-sinusoidal MSCs shown to co-localize with HSCs were so-called CXCL12-abundant reticular (CAR) stromal cells ⁵⁹. CAR cells are adipo-osteogenic progenitors that express high amounts of the HSC maintenance factors CXCL12 and SCF ⁷² as well as the lymphoid progenitor and B cell maintenance factor interleukin 7 (IL-7) ⁷³. These cells’ importance for maintaining HSCs was demonstrated in studies ablating CXCL12-positive bone marrow cells, which led to HSC depletion ⁷². Self-renewing Nestin⁺ MSCs are also essential niche cells that associate with HSCs and highly express HSC maintenance genes including SCF and CXCL12 ²⁷. In the bone marrow, Nestin⁺ MSCs densely surround arterioles, can be found around the sinusoids, and receive direct sympathetic innervation ^25,48,51,74. Advanced 3D imaging of the whole bone marrow space identified rare Nes-GFP^high cells that are positive for the pericyte marker NG2 ²⁵ or myosin heavy chain 11 (MYH11) ⁷⁵ and closely associate with arterioles and quiescent HSCs as well as Nes-GFP^low cells located around the sinusoids that overlap with LEPR⁺ cells and CAR cells ²⁵.

Importantly, studies have clarified that mesenchymal stromal cell populations identified and described using markers such as CAR, LEPR-Cre, NG2-Cre, PRX1-Cre, Nes-GFP, and SCF-GFP substantially overlap (reviewed in ¹⁰).

Endothelial cells

Endothelial cells are important niche constituents and form the so-called vascular HSC niche. ³⁴. Endothelial cells produce multiple HSC function-regulating factors including CXCL12, SCF, pleiotrophin, and Notch ligands ^{62,63,76–80}. Different approaches to manipulating endothelial cell function using conditional deletion of proteins (e.g. SCF ⁶², CXCL12 ^63,81, gp130 ⁸²) in Tie2-Cre or Cdh5-Cre mice, knockout mice (e.g. E-selectin^−/−) ⁸⁰, or blocking antibodies (e.g. VEGFR2 signaling) ⁸³ all considerably affected HSC function and numbers. In co-culture experiments endothelial cells promote HSC maintenance ⁸⁴ and bone marrow sinusoidal endothelial cells even support HSC expansion ^78,85.

Based on differential expression of podoplanin (PDPN) and SCA1, bone marrow endothelial cells can be divided into arteriolar endothelial cells (SCA1^highPDPN⁻) and sinusoidal endothelial cells (SCA1⁺PDPN⁺) that fulfill distinct roles in HSC maintenance. One example is the almost exclusive production of endothelial cell-derived SCF by arteriolar endothelial cells ⁸⁶.

In addition to producing crucial HSC maintenance factors, endothelial cells affect HSC function in other, more physical ways. For example, blood vessel permeability may correlate with nearby HSCs’ ROS levels, which in turn determine HSC quiescence. HSCs located around less-permeable arterioles have lower ROS levels and are therefore more quiescent, while HSCs around the leakier sinusoids show higher ROS levels promoting their differentiation and migration ⁸⁷.

Megakaryocytes

Megakaryocytes promote HSC quiescence ^88–91. 3D imaging of the bone marrow space revealed an HSC subset closely associated with megakaryocytes, and megakaryocyte depletion reduced HSC quiescence ⁸⁸. Megakaryocytes might regulate HSC quiescence by secreting CXCL4, TGFβ, and thrombopoietin (THPO) ^88–92. Indeed, TGFβ1 and CXCL4 release may be an important mechanism by which megakaryocytes control myeloid progenitor proliferation and restore HSC quiescence during emergency hematopoiesis ⁹³. HSCs biased towards platelet formation (von Willebrand factor (vWF)-GFP⁺) are spatially associated with megakaryocytes that regulate their activity, suggesting the existence of a feedback loop by which megakaryocytes regulate their own production ⁹⁴. Overall, megakaryocytes seem to form a second HSC quiescence-promoting niche that is distinct from the arteriolar niche that also promotes HSC quiescence ^25,53.

Monocytes, Macrophages, Neutrophils, and T cells

Macrophages promote HSC retention in their bone marrow and spleen niches ⁹⁵. These effects are mediated either by direct macrophage-HSC interactions or indirectly via stromal cells ^96–98. In the spleen, for instance, red pulp macrophages use vascular cell adhesion molecule 1 (VCAM1) to retain HSCs ⁹⁹. Furthermore, macrophages also influence the HSC niche by helping clear “aged” neutrophils (CD62^lowCXCR4^high) from the circulation. In this process, phagocytic macrophages modulate niche activity by liver x receptor (LXR) signaling, which regulates CXCL12 expression on MSCs and promotes the circadian release of HSPCs into the circulation ¹⁰⁰. Besides macrophages, changes in sympathetic activity also contribute to the circadian rhythmicity of HSPC release ^48,51,74. Osteoclasts, which are specialized giant cells that resorb bone, have also been implicated as niche regulators ^101–103. Neutrophils likewise fulfill important roles in the bone marrow niche. Upon granulocyte colony-stimulating factor (G-CSF)-induced sympathetic stimulation, neutrophils produce prostaglandin E2, which acts on osteolineage cells to retain HSPCs in the bone marrow ¹⁰⁴, and neutrophil-derived tumor necrosis factor α (TNFα) accelerates vascular and hematopoietic recovery of the bone marrow after irradiation ¹⁰⁵. A recent study identified an orexin-receptor⁺ pre-neutrophil population in the bone marrow that produces colony stimulating factor-1 (CSF1), which accelerates myelopoiesis and atherosclerosis during sleep fragmentation due to reduced orexin signaling from the hypothalamus ¹⁰⁶. Finally, lymphoid cells also control niche processes. FOXP3⁺ regulatory T cells have been shown to promote allogeneic HSC survival after transplantation by secreting interleukin 10 (IL-10) and generally seem to provide the niche with immune privilege ^107–110.

Molecular regulators of the stem cell niche

Stem cell factor (SCF) is a crucial cytokine for HSC survival, proliferation, and differentiation as well as for chemotactic HSC migration and retention of HSCs and progenitors in the bone marrow ^61,111–114. SCF exists in a soluble and membrane-bound forms, both of which activate the c-KIT receptor (CD117) on HSCs ^115–123. Mice lacking membrane-bound SCF (Sl/Sld) have very low HSC numbers and exhibit severe anemia ^60,124. In mice with chimeric spleens containing a mixture of S1/S1d and wild-type stromal cells, normal hematopoiesis was only observed in close proximity to the wild-type cells, thereby revealing the importance of locally produced SCF in forming the HSC niche ¹²⁵. SCF is primarily produced by leptin receptor⁺ (LEPR) mesenchymal stromal cells in the perivascular space of bone marrow sinusoids as well as by sinusoidal endothelial cells, and using LEPR-Cre and Tie2-Cre to delete SCF from these cell types indeed lowered HSC levels ^62,75,77,126.

Another key niche factor required for HSC maintenance and retention is the chemokine CXCL12 (SDF-1), which binds to the CXC chemokine receptor type 4 (CXCR4) (SDF-1-receptor) on hematopoietic cells ^56–59,127. CXCL12 is also involved in HSC proliferation and transplanted HSC engraftment ^128,129, and conditional deletion of CXCL12 or CXCR4 from niche cells strongly reduces HSC numbers in the bone marrow ^59,127,130. The proliferation and retention of various hematopoietic progenitors also depends on CXCL12 ^{56,128,131,132}. Perivascular MSCs are the main sources of CXCL12, while its expression levels are 100-fold lower in endothelial cells and 1000-fold lower in osteoblasts ^{59,63,81,133,134}.

HSC maintenance also requires thrombopoietin (THPO), which acts via myeloproliferative leukemia protein (MPL) on HSCs ^135–138. In addition to its role in maintaining HSC, THPO promotes megakaryocyte and thrombocyte development ^139–141. The growth factor is mainly produced in the liver and kidney, with only minor production in the bone marrow under homeostatic conditions ^139,142. The relevance of locally produced versus liver- or kidney-derived THPO for HSC maintenance has not yet been evaluated.

Multiple other proteins that act locally in the bone marrow have been shown to regulate HSC function without being absolutely required for HSC maintenance. Among them, several factors that are dispensable to steady-state hematopoiesis but become important regulators of emergency hematopoiesis in response to injury or infection, including fibroblast growth factor 1 and 2, angiogenin, angiopoietin-like protein 3, interleukin 6, Notch 2, pleiotrophin, interleukin 3, and others (reviewed in ⁴²). Among important long-range signals reaching the bone marrow niche via the circulation are corticosterone, catecholamines, and toll-like receptor (TLR) ligands.

Effect of aging on the HSC niche

Aging strongly affects the HSC niche. Characteristics of aged HSCs include reduced regenerative potential ^143–146, myeloid bias upon transplantation ^145–149, enhanced mobilization and reduced homing ^143,150,151 as well as an altered distribution pattern within the bone marrow ^152,153. These changes of HSC functions seem to mainly depend on cell intrinsic mechanisms such as DNA damage, cell polarity defects, altered metabolism and transcriptional and epigenetic profiles ^143,145,146, reviewed elsewhere ^154,155. However, cell extrinsic mechanisms also seem to play a role as the engraftment of transplanted young HSCs is better in young recipient than in old recipients ^145,156. HSC extrinsic mechanisms include increased CCL5 production promoting myeloid bias of HSCs ¹⁵⁶ as well as decreased osteopontin expression by bone marrow stromal cells ¹⁵⁷.

With ageing the bone marrow space also undergoes anatomical changes including a shift from osteolineage to adipolineage differentiation of MSCs, which results in decreased bone formation and accumulation of adipocytes ^158,159. Changes in the vasculature include a reduced frequency of arterioles and transitional vessel, decreased arteriolar innervation, higher leakiness and ROS levels, as well as reduced expression of HSC maintenance factors ^76,153,160. The importance of bone innervation is underlined by surgical denervation experiments in young mice that recapitulated many aspects of normal HSC ageing.

The niche in context

The emergence and rapid development of single cell sequencing techniques made possible broad molecular definition and characterization of bone marrow niche cell populations. Two recent single-cell resolution studies analyzing the transcriptional landscape of the bone marrow niche revealed unexpected cellular heterogeneity and determined cellular sources of important niche factors ^161,162. Tikhonova et al. performed single-cell RNA sequencing on pre-sorted vascular (VEcad-cre⁺), perivascular (LEPR-cre), and osteoblast (Col2.3-cre) cell populations from the mouse bone marrow in steady state and hematopoietic stress conditions (5-FU i.p.) ¹⁶¹. Analyses at baseline identified two endothelial, four perivascular, and three osteolineage subpopulations as well as a small cycling cell cluster and the relevant factors these subsets produce. Stress induced considerable changes in niche cell transcriptional profiles: perivascular cells were skewed adipocytic, vascular Notch delta-like ligands and E-selectin were attenuated, and proliferation of all niche subpopulations increased from 0.7% at baseline to 5.4% after chemotherapy. Using droplet-based single-cell RNA sequencing of murine bone and bone marrow stroma, Baryawno et al. identified 17 stromal cell subsets, including previously undescribed mesenchymal, pericyte, endothelial and fibroblast subpopulations that express distinct genes involved in regulating hematopoiesis ¹⁶². These analyses revealed previously unknown differentiation trajectories for osteolineage cells; production of the key niche factor CXCL12 by a fibroblast subset; a novel, distinct cluster of arterial endothelial cells with particularly high niche factor expression; and molecular distinctions among Nestin, LEPR, and NG2-expressing niche populations.

As a complement to single-cell RNA sequencing approaches, mass cytometry (CyTOF)- based single-cell protein analyses have recently identified 28 stromal cell subsets in the steady-state bone marrow, of which 14 subsets expressed factors that regulate hematopoiesis ¹⁶³. While radiation conditioning induced a loss of LEPR⁺ and Nes⁺ putative niche cells and many other subsets, a CD73⁺NGFR⁺ subpopulation was preserved and supported HSPC engraftment and acute recovery of hematopoiesis. These findings raise questions about the importance of some stromal subpopulations (e.g. LEPR⁺ and Nes⁺ cells) currently though to play a dominant role, mostly based on non-inducible Cre-derived results, at least under stress conditions. Another important finding of this study is that endogenous leptin receptor and nestin production correlate poorly with reporter gene expression in the commonly used Cre and reporter lines ¹⁶³. Finally, a recent study described an Apelin+ bone marrow endothelial cell population that regulates steady-state hematopoiesis and responds to signals from HSPCs to mediate regeneration of the vascular niche after irradiation injury ¹⁶⁴.

During the last decade, our understanding of the bone marrow stem cell niche and its cellular and molecular composition has expanded rapidly (Figure 1). Genetically engineered mouse strains, advanced imaging techniques, computational analyses, and single-cell RNA sequencing approaches have shed light on the complexity and heterogeneity of stem cell niches in the bone marrow. However, important technical considerations that arose with technical progress include the specificity and recombination efficacy of the different Cre lines and the compensatory effects that might occur following deletion of a specific cell type or a factor from a specific cell type.

Figure 1: The hematopoietic stem cell niche.

HSCs mainly reside in the bone marrow, where the bulk of hematopoiesis occurs during adulthood. Bone structures that form and protect the medullary cavity are densely innervated and vascularized with the periosteum lining the outer surface and the endosteum coating the inner surface. The endosteum forms the interface between bone and bone marrow and consists of bone-lining cells including osteoblasts that form bone substance and osteoclasts that resorb it. Arteries reach the bone marrow via nutrient canals, branch into smaller arteries and eventually into thin-walled arterioles. Transitional vessels connect arterioles with an extensive sinusoidal network that drains into a central venous sinus, which in turn empties into nutrient veins that leave the marrow via nutrient canals. Sympathetic fibers follow the arteries into the bone marrow, where they form a structural network with neural-glial antigen- (NG2) and nestin- (Nes) positive perivascular stromal cells. HSCs primarily reside in perivascular niches around fenestrated sinusoids as well as around less-permeable arterioles, while lymphoid progenitors reside in an osteoblastic niche closer to the endosteum. Overall, the bone marrow niche is formed by a complex network of non-hematopoietic as well as hematopoietic cell types that produce multiple factors orchestrating maintenance, proliferation and differentiation of HSCs and their progeny. The best-studied and apparently most relevant niche cell types are endothelial cells (sinusoidal and arteriolar ECs) and perivascular mesenchymal stromal cells (Ng2-CreER⁺, LEPR⁺/CAR cells) that produce crucial HSC maintenance factors such as SCF and CXCL12. Other cell types involved in niche formation include osteoblasts, osteoclasts, sympathetic nerve fibers, non-myelinating Schwann cells, adipocytes, megakaryocytes, macrophages and regulatory T cells (Treg). Abbreviations: NG2 - neuron/glial antigen 2; HSC - hematopoietic stem cell; Treg - regulatory T cell; LEPR - leptin receptor; CAR cell - CXCL12-abundant reticular cell; EC - endothelial cell; CXCL12 - CXC chemokine ligand 12; SCF - stem cell factor; OPN - osteopontin; IL7 - interleukin-7; TGFβ - transforming growth factor-β; CXCL4 - CXC chemokine ligand 4; DARC - duffy antigen receptor for chemokines; TNF - tumor necrosis factor; vWF - von Willebrand factor; PTN - pleiotrophin.

The hematopoietic differentiation tree

Conceptual update of the tree model

The hematopoietic differentiation tree was long imagined to be strictly hierarchical. Recent studies, however, have significantly challenged this conceptualization, particularly its clear demarcation of stem and progenitor cells, the strict time points for cell fate choices, and order of differentiation steps.

The first hematopoietic tree models suggested an early separation between lymphoid and other (myeloid, erythroid, megakaryocytic) lineages, followed by several branching points so that each lineage differentiated from multipotent to bipotent to unipotent progenitors. Subsequent modifications of the tree included a longer association between myeloid and lymphoid lineages ^165–167, earlier separation of the megakaryocytic lineage ^168,169, and definition of multipotent progenitor subsets ^170,171. In addition, several studies described heterogeneity within the HSC population ¹⁷². In recent years, new technologies provided further evidence for a dynamic system comprising heterogeneous cell populations that gradually progress from one to the other, with a high grade of flexibility to respond to changing demands.

The hematopoietic stem cell population at steady state and under stress

In the bone marrow, approximately 1 in 10,000 cells is a transplantable hematopoietic stem cell ^11,12. HSCs were traditionally defined by multipotency and self-renewal capacity, while progenitors were defined by restricted lineage and lack of prolonged self-renewal capacity. While HSC can be repetitively transplanted and can recover the entire hematopoietic system, progenitors are typically lost 2–3 weeks post transplantation ¹⁶⁵. The transcriptional program of HSCs has unique cellular and metabolic properties, including a quiescent, glycolytic state with low mitochondrial activity and low protein synthesis levels compared to other hematopoietic cells (reviewed in ¹⁷²). Specific stress-response and control mechanisms render HSCs less susceptible to metabolic stress as well as to DNA and protein damage ¹⁷². In contrast to HSCs, progenitor populations strongly proliferate and show high metabolic activity requiring oxidative metabolism and mitochondrial activity.

Over the last decade, different HSC subpopulations have been identified according to their capacity for self-renewal. Using transplantation assays, HSC capable of engrafting for more than 16 weeks upon both primary transplantation and at least a second round of transplantation are defined as Long-Term (LT) HSCs ^34,173–176. Multipotent cells that only transiently engraft but differ with respect to durability and graft stability are defined as Intermediate (IT) HSCs ¹⁷⁷, Short-Term (ST) HSCs, or Multipotent Progenitors (MPP) ^170,171,173.

Label retention studies indicate that the time an HSC spends in quiescence positively correlates with its self-renewal capacity ^178–182. In mice, division frequencies vary between once monthly to twice yearly, depending on the HSC subset ^{13,178,180,182}. A recent study estimated that LT-HSCs asynchronously undergo four symmetric self-renewal divisions during adulthood to expand their pool but irreversibly lose long-term regenerative capacity at the fifth division ¹⁸². Quiescent HSCs that have been activated by stress can return to the quiescent state ^178,180.

In the steady state, a low number of HSCs are released into the circulation in a circadian pattern and some HSCs also reside in the lungs and spleen ^74,183,184. So far, little is known about either potential lineage biases or differentiation and clonal expansion potentials of HSCs in these extra-medullary niches.

Various stressors, including DNA damage, inflammation, infections, metabolic stress, obesity, and psychosocial stress, directly influence HSC functions ¹⁷². Furthermore, HSC and progenitor subsets differentially react to stress. An example of subtype-specific stress responses is enhanced GMP production by MPP2 and MPP3 subsets that dynamically reorganize themselves and activate a progenitor self-renewal network under conditions of emergency myelopoiesis ^46,170.

HSC heterogeneity and cell fate decisions

While HSCs are defined by their capacity to differentiate into all blood cell types, different HSC and MPP subpopulations demonstrate heterogeneous lineage output ^170,171. For example, HSC subsets have varying relative myeloid and lymphoid lineage outputs ^185,186, and one subset is characterized by the expression of the platelet markers CD41 and vWF ^168,169 and primarily gives rise to megakaryocytes ^168,169, bypassing the multipotent progenitor stages ^{16,168,169,187–190}. Though the molecular mechanisms underlying this non-stochastic HSC heterogeneity are not well understood, micro-environmental cues and long-range signals are likely regulators.

Since cell fate decisions occur at the individual cell level, functional and biochemical assays must be conducted at single-cell resolution. In the last decade, multiple studies have used in vivo transplantation assays in mice as well as in vitro models with individual human progenitors to explore conceptual questions about lineage differentiation routes. For instance, and contrary to the classical common myeloid progenitor (CMP) – common lymphoid progenitor (CLP) model, lymphoid and myeloid lineage segregation is not the earliest fate decision; rather, lymphoid and myeloid potential are preserved until the stage of so-called lymphoid-primed multipotential progenitors (LMPPs) ^166,167. Furthermore, the different branches house various additional subsets, mainly with bi- or uni-lineage potential (reviewed in ¹⁷²).

Single-cell RNA sequencing has been successfully used to characterize the transcriptomes of HSC as well as diverse progenitor populations ^191,192. Various algorithms have been developed to understand cells’ differentiation cascades based on gradual changes in their single-cell transcriptomes during this process ¹⁷². These analyses uncovered differentiation trajectories in which the position of each individual cell is determined by its transcriptional snapshot. For example, a recent study analyzing human HSPC using single-cell RNA sequencing with computational analyses and in vitro single-cell assays indicated early lineage restriction ¹⁹³. The authors suggest HSCs continuously acquire lineage biases without passing through discrete multi- and bipotent stages, so that unipotent cells emerge directly from a “continuum of low-primed undifferentiated hematopoietic stem and progenitor cells’ (CLOUD-HSPCs)“ ¹⁹³. Earlier reports of numerous unipotent progenitors within compartments that are multipotent at the population level support these data ^187,191,194. That said, in contrast to these single-cell RNA sequencing-based data, HSPC can be clearly split into functionally distinct subpopulations based on surface marker combinations. There is ongoing discussion about the reasons for these seemingly incompatible results, including decoupling of mRNA and protein expression; epigenetically determined functional heterogeneity; more gradual changes in functionality, as suggested by functional assays of purified cells; and inefficacy of single-cell RNA sequencing-based methods in separating closely related cell types ¹⁷². Future research should determine the extent of continuity versus distinctness in early hematopoiesis.

To study HSC and progenitor behaviors in their physiological microenvironments, individual cells can be tagged before transplantation (reviewed in ¹⁷²). Though their tagging methods vary, these studies generally support the lineage biases and restrictions demonstrated in single-cell transplant assays. In a transplantation setting, a small number of HSCs gives rise to the vast majority of differentiated cells. New technologies such as barcode labeling can track the lineage output of individual HSCs and progenitor cells during this process ^14,172. Such studies indicate that during unperturbed steady-state hematopoiesis, MPPs mainly contribute to the myeloid lineage and that hematopoiesis is almost completely driven by cells from the MPP populations. Furthermore, the studies suggest that during steady-state hematopoiesis, megakaryocytes can arise independently from the other lineages and that LT-HSCs actively contribute to the megakaryocyte output ^172,189.

Hematopoietic differentiation trajectories

As our understanding of stem cell heterogeneity and early hematopoiesis grows, so does our knowledge of the various differentiation trajectories cells can follow from the stem cell compartment to the mature blood cell pool. Current models indicate that the erythroid-megakaryocytic lineage separates directly from biased cells within the stem cell compartment, thus making it the earliest division in the hematopoietic tree. Three MPP populations with different progeny potentials have been identified: MPP2, MPP3, and MPP4. From the MPP stage a cell can differentiate into either a common myeloid progenitor (CMP) or a lymphoid-primed multipotential progenitor (LMPP) ^166,167. LMPPs mainly differentiate into common lymphoid progenitors (CLPs), which give rise to the lymphoid lineage (B cells, T cells, natural killer (NK) cells, innate lymphoid cells (ILCs)) but still retain the potential to differentiate into granulocyte-monocyte progenitors (GMPs). Thus, the myeloid and lymphoid lineages are connected longer than previously assumed. CMPs give rise to GMPs and monocyte-dendritic cell progenitor (MDPs). GMPs further differentiate into granulocyte progenitors (GPs) and common monocyte progenitors (cMoPs) that give rise to neutrophils and monocytes, respectively. MDPs differentiate into common dendritic cell progenitors (CDPs) that give rise to dendritic cells. A recent study used fate mapping via the expression history of Ms4a3, which is specifically and transiently expressed in bone marrow GMPs, to trace monocyte-derived cells. This study found that MDPs give rise to monocytes directly, bypassing the cMoP stage, and that MDPs do not arise from GMPs ¹⁹⁵. Altogether, these studies have enriched our understanding of hematopoiesis and the various differentiation pathways that progenitors can follow (Figure 2).

Figure 2: The hematopoietic differentiation tree.

Mature immune cells arise from hematopoietic stem cells in a process called hematopoiesis. Over the last decades, various models have been developed and continuously adapted in light of novel scientific insights to best visualize this differentiation process. Recent technical and conceptual breakthroughs that have been incorporated into the above model of the differentiation tree, include greater heterogeneity within the HSC population, less clear demarcation of stem and progenitor cell populations, earlier and more dynamic cell fate choices, as well as a much more gradual rather than stepwise differentiation of cells through the heterogeneous tree compartments with highly dynamic reactions to changing demands. The dashed line represents a trajectory of a hematopoietic cell during its differentiation process with red dots indicating transcriptional snapshots captured by single-cell RNA sequencing during this process. Overall, it is difficult to visualize all aspects of hematopoiesis in a 2-dimensional tree model and various partly controversial attempts to depict this have been published. Abbreviations: HSC - hematopoietic stem cell; LT - long term; IT - intermediate term; ST - short term; MPP - multipotent progenitors; LMPP - lymphoid-primed multipotential progenitor; CMP - common myeloid progenitor; CLP - common lymphoid progenitor; MEP - megakaryocyte-erythrocyte progenitor; EoBP - eosinophil-basophil progenitor; GMP - granulocyte-monocyte progenitor; GP - granulocyte progenitor; cMoP - common monocyte progenitors; MDP - monocyte-dendritic cell progenitor; CDP - common dendritic cell progenitors; NK - natural killer cells; ILCs - innate lymphoid cells; cDC - conventional dendritic cell

Hematopoiesis and atherosclerosis

In recent decades, we have come to understand that atherosclerosis is a multifactorial inflammatory disease that crucially depends on inflammatory leukocyte supply from the bone marrow niche and extramedullary sites ^4–7. Innate immune cells such as monocytes and macrophages as well as neutrophils in particular are decisive contributors to the initiation, progression, and destabilization of atherosclerotic lesions.

The healthy arterial wall

In the healthy arterial wall, resident macrophages dwell in the adventitia where they renew via local proliferation. In mice, arterial resident macrophages mostly derive from circulating monocytes, many of which accumulate immediately after birth ¹⁹⁶. Studies in mice with labeled CD11c⁺ cells indicate that dendritic cells also reside in the arterial wall, as well as in heart valves ^197,198, but it is still unclear how distinct these cells are from macrophages. While arterial resident macrophages likely conduct general macrophage tasks such as tissue homeostasis and pathogen clearance, their artery-specific functions are only beginning to be appreciated ¹⁹⁹.

Leukocytes in atherosclerosis

Macrophages are among the most abundant and functionally relevant cell types in atherosclerotic lesions, where they are responsible for plaque growth and destabilization ^{196,200–202}. These phagocytic cells are part of the innate immune system and populate most tissues in both steady state and disease conditions. Depending on lineage, phenotype, and host tissue, macrophages fulfill various functions including disposing pathogens and foreign material, presenting antigens, regulating hematopoiesis, resorping bone, iron recycling, synaptic pruning, electric coupling, and regulating temperature, among others ^203,204. Proatherogenic conditions lead to the deposition of cholesterol-rich lipoproteins, such as low-density lipoprotein (LDL), in the intima that become oxidized, triggering macrophage accumulation and inducing their differentiation programs ^4–7. In an attempt to remove the oxidized lipoproteins via scavenger receptors, macrophages become lipid-laden foam cells that eventually die in the lesions, fueling the growth of the necrotic core ²⁰⁵, which is an area within the plaque in which lipids and dead cells accumulate. Macrophages also remove dead or dying cells in a process called efferocytosis, and insufficient efferocytosis has been implicated as an important pathomechanism in atherosclerotic lesion growth ²⁰⁶. As important protease producers, macrophages contribute to the destabilization of the fibrous cap that covers the necrotic core. Furthermore, macrophages attract and activate other immune cells via the production of pro-inflammatory cytokines. Intravital imaging of large arteries in mice has shown that macrophages in plaques are highly motile: in addition to entire cells migrating, stationary cells extend and retract their dendrites, an action referred to as “dancing on the spot“ ²⁰⁷. The importance of macrophages and their pro-inflammatory actions to atherosclerosis and its complications is undeniable, and the recent large-scale CANTOS trial has provided proof-of-concept evidence that interfering with inflammatory pathways, specifically the macrophage product interleukin 1β (IL-1β), can indeed reduce cardiovascular events ²⁰⁸.

Aside from monocytes and macrophages, neutrophils also play important roles in atherosclerotic lesions. Activated by chemokines (e.g. platelet-derived CCL5) neutrophils are recruited into atherosclerotic plaques, where they produce proteins such as cathelicidins, myeloperoxidase (MPO), cathepsin G and other proteases as well as reactive oxygen species (ROS), which together accelerate leukocyte recruitment along with accumulation and oxidation of LDL ^209–212. The release of neutrophil extracellular traps (NETs) is another important mechanism leading to plaque progression via stimulation of plasmacytoid dendritic cells ²¹³ and macrophages ²¹⁴ as well as to plaque destabilization via the NET component cytotoxic histone H4 that perforates vascular smooth muscle cell membranes thereby inducing cell lysis ²¹⁵. Metalloproteinases released by neutrophils also contribute to plaque destabilization by degrading the extracellular matrix, which eventually may trigger plaque erosions and ruptures. For further discussion we refer the reader to a recent review ²¹⁶.

Monocyte recruitment into plaques

Blood monocyte numbers robustly correlate with atherosclerosis progression and cardiovascular event rate ^217–222, and in animal models atherosclerosis does not develop without monocyte influx ^223–225. Monocytes are divided into at least two functionally distinct subsets that can be distinguished by expression of the glycoprotein Ly6C in mice and CD14/CD16 in humans ²²⁶. Ly6C^high monocytes (CD14⁺⁺CD16⁻ and CD14⁺⁺CD16⁺ in humans) have a short life span and accumulate in inflamed tissues, where they can differentiate to macrophages ^226,227. Ly6C^low monocytes (CD14⁺CD16⁺⁺ in humans) have a longer life span and exhibit vascular patrolling behavior to enable early response to infections and endothelial damage ^228,229. Fate mapping studies convincingly demonstrated that Ly6C^low monocytes derive from Ly6C^high monocytes in the blood ^{195,230–232}. Human classical monocytes (CD14⁺⁺CD16^–) similarly give rise to intermediate monocytes (CD14⁺⁺CD16⁺) that further develop into non-classical monocytes (CD14⁺CD16⁺⁺) ²³³.

During atherogenesis, Ly6C^high monocytes bind to activated endothelium, migrate into the atherosclerotic lesion, and differentiate into macrophages. Monocyte recruitment depends on chemokine/chemokine receptor interaction (e.g. CCR2/MCP-1, CCR5/CCL2/CCL5, CX3CR1/fractalkine) as well as adhesion molecule expression on activated endothelial cells (e.g. VCAM-1, P-selectin etc.) reviewed in ^7,234. The relevance of Ly6C^low monocytes in atherosclerosis is less clear, although we do know they can ingest oxidized lipoproteins ²³⁵, do not directly contribute to the plaque macrophage population ^232,236,237, and may be important in endothelial protection ²³⁸.

As additional players in monocyte recruitment, platelets bind to activated endothelium ²³⁹, which induces production of macrophage-attracting chemokines and adhesion molecules ²⁴⁰. Platelet activation results in degranulation, chemokine release, increased P-selectin expression, and formation of monocyte-platelet aggregates ^241–245. Neutrophils have several possible functions, including promoting monocyte accumulation in plaques by producing cathelicidin that interacts with formyl-peptide receptor 2 on monocytes, leading to integrin activation ^211,212.

Plaque macrophage proliferation

In addition to monocyte recruitment, lesional macrophage numbers are determined by local macrophage proliferation within the plaque. Parabiosis and fate mapping analyses revealed that local proliferation dominates macrophage accumulation, particularly in established atherosclerotic lesions ²⁰², while monocyte recruitment has much larger contributions to developing lesions ^246,247. According to BrdU tracking experiments in ApoE^−/− four-month-old mice, high turnover of lesional macrophages can renew the entire plaque macrophage population within one month ²⁰². While little is known about factors that regulate macrophage proliferation in plaques, GM-CSF has been implicated in proliferation in early lesions ²⁴⁷, whereas type 1 macrophage scavenger receptor class A (Msr1) seems to play a role in advanced lesions ²⁰².

Based on fate mapping experiments, it has been suggested that macrophage-like cells in the plaque may derive from smooth muscle cells. An estimated 18% of macrophages in human plaques seem to originate from smooth muscle cells, but the functions of these macrophages remains unclear ²⁴⁸. The relevance of vascular smooth muscle cell transdifferentiation to macrophage-like cells is still under discussion ^248,249. Overall, monocyte production from hematopoietic progenitors in medullary and extramedullary sites, monocyte recruitment into atherosclerotic lesions, and local proliferation of macrophages all contribute to the progression of atherosclerotic lesions ^250–252 (Figure 3). Whether and how monocyte-derived macrophages functionally differ from locally produced macrophages requires further study.

Figure 3: Atherosclerosis.

Atherosclerotic plaque formation begins with the accumulation of low-density lipoprotein (LDL) particles in the intima of large arterial blood vessels. Common atherosclerotic risk factors accelerate this process. Within the intima, LDL particles are oxidatively modified, which renders them immunogenic and triggers an early inflammatory response, including endothelial cell activation. Upregulation of endothelial adhesion molecules and release of pro-inflammatory chemokines leads to the recruitment of inflammatory monocytes. Monocytes in the intima differentiate into macrophages that take up oxidized LDL via scavenger receptors and eventually become lipid-laden foam cells. The proinflammatory milieu in the intima also triggers the migration of quiescent smooth muscle cells (SMCs) from the media into the intima, where they proliferate and produce excessive extracellular matrix (ECM), including proteoglycans and collagens. Some SMCs may even trans-differentiate into macrophage-like cells. Cell recruitment, cell proliferation, and ECM production fuel atherosclerotic plaque growth. Ongoing cell death and impaired efferocytosis (removal of dead/dying cells) lead to the formation of an ever-growing lipid-rich necrotic core that sustains inflammation. Necrotic core growth and production of proteases, such as matrix metalloproteinases (MMPs) by activated macrophages that degrade collagens in the fibrous cap set the stage for plaque ruptures and plaque erosions that are responsible for the life-threatening complications of atherosclerosis.

Fate of plaque macrophages

Mechanisms that reduce plaque macrophage numbers may include macrophage death and exit from plaques as well as reduced monocyte recruitment ^6,253,254. Methods that can image cells in the mouse arterial wall in vivo were an important technical breakthrough to address cell motility and cell-cell interactions ^207,255. Such methods may enable visual analyses of macrophages fates in plaques (death versus exit) and the pathways they might take (blood versus lymph). Although macrophage departure from atherosclerotic lesions has not be visually confirmed, there is evidence that dendritic cells exit the aortic wall of healthy mice in response to infectious stimuli ¹⁹⁷. Whether lipid-laden macrophages depart similarly and whether this is a plaque regression mechanism still remains to be imaged.

Atherosclerosis accelerates hematopoiesis

Over the course of atherosclerosis progression, monocyte levels in the blood progressively increase ²²⁴. These circulating monocytes accelerate atherosclerosis, which in turn reinforces immune cell production and release from the bone marrow, thus fueling a feedforward cycle. The involved pathways include catecholaminergic signaling on bone marrow niche cells as well as hematopoietic stem and progenitors cells, all of which express adrenergic receptors. Elevated catecholamine release from the adrenal glands as well as elevated noradrenaline release from sympathetic terminals in the bone marrow contribute to this process ^74,50,256. Other danger signals such as TLR ligands also regulate HSC functions via different transcription factors, including PU.1, Egr-1, Irf8, Klf4, Mafb, and C/EBPα ²⁵⁷, which can induce myeloid bias of HSC, leading to increased myeloid cell production.

Myeloid cell release into the circulation is tightly controlled by several bone marrow intrinsic and extrinsic signals. Chemokines/chemokine receptor interactions are an important regulatory mechanism. C-C chemokine receptor type 2 (CCR2) blockade in mouse models of atherosclerosis, for instance, reduces monocyte release from the bone marrow and strongly decreases plaque burden ²⁵⁸. Major mechanisms involved in neutrophil release include interactions between CXCL12 and CXCR4, as well as between CXCL1/2 and CXCR2 ²⁵⁹.

Cardiovascular risk factors also directly influence the stem cell niche to induce increased HSPC proliferation, myeloid-biased differentiation, release into the circulation, and extramedullary hematopoiesis ²⁶⁰. Various mechanisms control HSC proliferation and differentiation to myeloid cells. For example, increased plasma lipid levels impair reverse cholesterol transport in HSPCs, leading to higher myeloid cell production ^{224,225,261,262}. Proteoglycan-bound ApoE at the cell surface of HSPCs promotes adenosine triphosphate-binding cassette (ABC) transporter ABCA1/ABCG1-mediated cholesterol efflux, which reduces proliferation, blood monocytosis, and atherosclerosis. In ApoE-deficient HSPCs, cholesterol efflux pathways are disrupted and the accumulating cholesterol raises the expression of the common beta subunit (CBS) of the interleukin 3/granulocyte-macrophage colony-stimulating factor (IL-3/GM-CSF) receptor, thereby rendering the cells hypersensitive to IL3/GM-CSF signaling, which in turn induces phosphorylation of ERK1/2 and STAT5. Overall, this HSPC-intrinsic mechanism supports proliferation and leukocytosis, consequently accelerating atherosclerosis ^{22,224,261,262}.

While we are beginning to identify pathways through which risk factors accelerate leukocyte production in CVD, many questions remain to be answered before these insights can be translated into novel therapies. Although several significant CVD risk factors, such as arterial hypertension, smoking, and aging, are all associated with considerable leukocytosis, the underlying mechanisms remain unknown. Acute cardiovascular events like MI or stroke also have major impacts on hematopoiesis via sympathetic activation, glucocorticoid release, and production of various proinflammatory molecules (IL-1β, TLR ligands, chemokines) ^7,48,256,263. The overall hematopoietic response can be summarized as a massive expansion of myelopoiesis driven by catecholamines, inflammatory cytokines, damage-associated molecular patterns (DAMPs), and TLR ligands, at the expense of a strongly suppressed lymphopoiesis, which is mainly induced via corticosterone signaling on lymphocyte progenitors.

Formation of extramedullary hematopoietic niches

Under healthy steady state conditions, only a few hematopoietic progenitors are detectable outside the bone marrow ^264,265. In inflammatory diseases such as atherosclerosis, however, myeloid cell production can partly shift to other lymphoid organs, particularly the splenic red pulp ^{22,256,262,266,267}. Within this extramedullary niche myeloid-biased HSCs give rise to significant numbers of monocytes and neutrophils that can be released into the circulation to eventually enter atherosclerotic lesions ²².

Little is known about factors that influence the extramedullary retention of HSPCs and the formation of niches outside the bone marrow that maintain HSC proliferation and differentiation, but sphingosine 1-phosphate (S1P) signaling might be involved ^265,268. HSPC-extrinsic factors that potentially accelerate splenic myelopoiesis include GM-CSF-producing innate response activator (IRA) B cells, which develop in the spleen in response to infections or atherosclerosis ^269–271, and high-fat diet-induced interleukin 23 (IL-23), which triggers production of the HSC-mobilizing factor G-CSF ²⁷². Atherosclerotic risk factors and disease-promoting lifestyle factors, such as stress, lack of exercise, or poor sleep, also influence HSC functions ^50,106,273 and are reviewed in detail elsewhere ²⁷⁴. Finally, recent studies have documented the importance of clonal hematopoiesis in cardiovascular risk ^{260,275–277}.

Hematopoiesis and myocardial infarction

Myocardial infarction (MI) and post-MI heart failure are accompanied by profound inflammatory immune activation that shapes and is shaped by the course of the disease. Here we discuss our growing knowledge of the immune system’s role in the healthy heart as well as in myocardial infarction and ischemic heart failure.

Leukocytes in the healthy heart

Healthy mouse heart tissue is populated by at least an order of magnitude more leukocytes than regular skeletal muscle; among the leukocytes, macrophages represent the biggest fractions ^{204,278–281}. Macrophages are also present in the healthy human myocardium ²⁸², forming a dense network between the cardiomyocytes ^{204,280,281,283,284}.

Cardiac macrophages are heterogeneous. While all macrophages are CD45⁺CD11b⁺F4/80⁺ CD64⁺MERTK⁺, their relative expression of MHCII, CCR2, and Ly6C differs. CCR2^– macrophages can be divided into MHCII^high and MHCII^low subsets while only a minor population of CCR2⁺ macrophages expresses MHCII. The smallest subset is a Ly6C-expressing macrophage population ^204,280,281. Available fate mapping evidence suggests that CCR2^– and Ly6C⁺ macrophages arise from cells that enter the heart during embryogenesis and self-renew postnatally to sustain these populations without monocytic input ^280,283. CCR2^– cells are found throughout the myocardium adjacent to developing arteries. In contrast, the smaller CCR2⁺ subset in adult mice most likely derives from recruited monocytes and resides near the endocardium. Although several studies challenged these seemingly clear ontogenic demarcations, the weight of evidence suggests that only a minority of cardiac macrophages in the healthy heart derive from circulating monocytes ^196,285,286. Additional leukocyte sources, under specific conditions, include pericardial fluid and pericardial adipose tissue ^287,288.

We are only beginning to understand macrophages’ tissue-specific functions in the heart, though they likely ingest invading pathogens, dying cells, and matrix components ^280,283. Determining cardiac macrophages’ functions may be particularly important for diseases in which resident cardiac macrophages die and are replaced by monocyte-derived macrophages that might not necessarily acquire tissue-specific tasks from the resident cells. In addition, cardiac resident macrophages themselves may neglect tissue-specific tasks when exposed to an inflammatory milieu. Cardiac resident macrophages may be involved in heart development, coronary maturation ²⁸⁹, adaptation to increased tissue strain, training and pregnancy-induced cardiac remodeling, and regulation of autonomic innervation ²⁹⁰. A recent study showed that macrophages form connexin 43-containing gap junctions with cardiomyocytes in the conduction system of the heart ²⁸². These gap junctions electrically couple cardiomyocytes with macrophages and influence conduction, especially via the AV node where macrophages reside in high numbers. Indeed, macrophage depletion in CD11b DTR mice induced progressive AV block as well as milder conduction abnormalities in atria and ventricles ²⁸².

During aging, the cardiac leukocyte population undergoes significant changes, including lower overall macrophage numbers, potentially higher percentage of monocyte-derived macrophages, and fewer granulocytes ^278,285,291. In older mice, phenotypic changes in macrophages such as increased IL-10 production ^291,292 contribute to myocardial fibrosis and diastolic dysfunction. However, the implications of these changes for cardiac diseases remain largely unclear. Further details on leukocyte functions and intercellular communication pathways in the steady-state heart have recently been reviewed ^281,290.

Leukocytes and myocardial infarction

Atherosclerotic lesions develop over decades and progressively destabilize as dying myeloid and foam cells fuel necrotic core growth and proteases thin out the fibrous cap. Eventually, atherosclerotic plaques may either rupture, meaning a break in the fibrous cap exposes highly thrombogenic necrotic core material to the blood stream, or erode, as a loss of plaque endothelial cells uncovers thrombogenic subendothelial proteins, such as collagens. In the coronary arteries, the resulting thrombus may acutely occlude the vessel and cause downstream ischemia. Epidemiological data suggest that acute stressful events trigger plaque ruptures and erosions. Mechanistically, stress-induced elevations of blood pressure, heart rate, and coronary flow elevate shear stress at the plaque cap. The immediate contribution of immune cells to either rupture or erosion are not known, but extracellular matrix-destabilizing proteases synthesized by neutrophils and macrophages drive the destabilization process that precedes the actual event. The resulting myocardial infarction is a life-threatening complication that can, depending on infarct size and location, lead to severe arrhythmias, heart failure, or even cardiac rupture ^4–7,293.

Ischemic injury rapidly triggers a series of immune responses that begin within the resident cardiac leukocyte population, including mast cell degranulation ²⁹⁴, production of inflammatory factors (IL-1, IL-6, TNF, CCL2) by cardiomyocytes and macrophages ^295–297, GM-CSF release from fibroblasts ²⁹⁸, and endothelial cell activation. In response to these inflammatory mediators and the progressive cell death of cardiomyocytes and resident cardiac macrophages, the bone marrow and extramedullary sites produce and release large numbers of neutrophils and inflammatory monocytes that enter the circulation and reach the infarcted myocardium ^7,21,267. Strong clinical evidence exists for the occurrence of post-MI leukocytosis ^222,299 and a robust association between leukocytosis and cardiovascular morbidity and mortality ^300,301.

The early phase after MI is dominated by the recruitment of millions of neutrophils and inflammatory monocytes (Ly6C^high in mice and CD16^low in humans) that differentiate into macrophages with high proteolytic and phagocytic activity ^{232,302–304}. During the first several days, these cells scavenge dead and dying cells along with debris ^305,306 and fuel inflammation by producing IL-1, TNF, and IL-6 ³⁰⁷. Neutrophil numbers decrease after 3 days, and after 1 week they have largely disappeared from the infarcted myocardium. Experimentally depleting neutrophils, which are important for cardiac healing, elevates cardiac fibrosis, impaired function, and heart failure ³⁰⁸. Unlike neutrophils, monocytes migrate to the heart for several days. Comparatively low numbers of non-inflammatory monocytes also enter the heart.

Around day 4, the initial inflammatory clean-up phase transitions into a reparative healing phase with rapidly decreasing neutrophil numbers and a phenotypic macrophage switch towards reparative Ly6C^low macrophages that mainly contribute to angiogenesis and fibrosis by producing factors such as IL-10, TGFβ and vascular endothelial growth factor (VEGF) ^232,302. The reasons why resident macrophages die and are replaced by monocyte-derived macrophages are not well understood but may include a need for high numbers of macrophages with alternative phenotypes.

Macrophage subsets in the heart were generally identified based on the surface expression of CCR2, Ly6C, and MHCII. With the emergence of single-cell RNA sequencing, these cell population definitions are being re-evaluated. Indeed, single-cell RNA sequencing analyses of sorted CD45⁺ infarct cells have identified up to seven transcriptionally distinct subsets. On day 4 after MI in mice, one macrophage subset is activated via the interferon regulatory factor 3(IRF3)-type 1 interferon pathway, which induces the transcription of genes that typically respond to viral infections ³⁰⁶. Interferon-based activation of these macrophages that were termed “IFNICs” seems to be detrimental in MI, as pathway blockade via global IRF3 or type 1 interferon receptor knockout improves recovery and survival after MI in mice ³⁰⁶. However, IFNICs may have other beneficial functions during MI healing. While the other six subsets’ functional distinctiveness remains to be investigated, these single-cell RNA sequencing data provide additional evidence for the complexity of cardiovascular macrophage biology and the insufficiency of the M1/M2 model ^260,309. Single-cell approaches, including CyTOF and RNA sequencing, are very effective in identifying new leukocyte subsets in cardiovascular tissues, but it is equally important to evaluate their functional differences ^200,201,310.

Leukocytes and ischemic heart failure

The balanced sequence of inflammatory and reparative actions by various contributing immune cell populations determines the mechanical outcomes of the myocardial infarct, including scar size, thickness, and stability ^290,311,312. Only after several weeks do cardiac leukocyte populations return to baseline levels. While the myocardial tissue that died during the infarct turns into a scar, the surrounding myocardium undergoes intense remodeling that is critical for the development of ischemic heart failure with reduced ejection fraction (HFrEF). In addition to stromal cells, leukocytes are also involved in this process. Different leukocyte classes are recruited to the remote myocardium starting shortly after MI and peaking around day 10 post-MI ^313–316. In the surviving areas, macrophage pools expand due to resident macrophage proliferation and monocyte recruitment ^267,315. If heart failure develops, the increased sympathetic tone fuels myeloid cell production in the bone marrow via altered expression of CXCL12 and other niche factors and leads to leukocytosis in the blood ^220,315. During remodeling, macrophages in the remote myocardium also change their expression of key enzymes and factors, including reduced production of MPPs and increased production of IL-1β, TNF, as well as VEGF ³¹⁵. Overall, recruited macrophages seem to negatively influence the post-MI remodeling process ³¹⁵. Recent reviews provide detailed information about the roles played by different leukocyte populations in myocardial infarct healing and remodeling ^260,290,317.

Myocardial infarction accelerates inflammatory hematopoiesis

If the patient survives, a myocardial infarction still has dramatic consequences. Within the first year after an index MI, the frequency of stroke, secondary MI, or death is around 8–12% despite optimal medical treatment ^318–321 and the rate of other ischemic events is even higher ^322–324. Understanding the reasons for this phenomenon is crucial for developing successful secondary prophylaxis strategies. While ongoing exposure to risk factors is one important aspect, another is MI-induced long-lasting changes in the immune system. Given the short lifespan and rapid turnover of leukocytes in inflamed tissues, a continuous supply of newly produced immune cells is crucial for cardiovascular disease progression. Several different pathways have been suggested to directly accelerate inflammatory hematopoiesis in MI.

Dead and dying cells in the infarcted myocardium release various inflammatory mediators, including IL-1, IL-6, TNF, CCL2 ^{295–297,325}, and GM-CSF ²⁹⁸, that act on their respective receptors on hematopoietic and niche cells in the bone marrow to promote hematopoiesis and leukocyte release. The infarcted myocardium also releases damage-associated molecular patterns (DAMPs), including DNA, S100A8-S100A9, high-mobility group protein B1 (HMGB1), and ATP, that circulate and activate toll-like receptors on HSPCs, thereby triggering production of innate immune cells ^260,290,326. Such external alarm signals activate intracellular signal cascades leading to the production of multiple transcription factors, including C/EBPα, Egr-1, Irf8, Klf4, Mafb, and PU.1, that activate HSC proliferation and/or induce myeloid bias ²⁵⁷.

Another important mechanism that accelerates hematopoiesis in response to MI is the attenuation of quiescence-promoting niche factors, such as CXCL12 (SDF-1), in the bone marrow. Similar to other stressors, MI mobilizes the sympathetic nervous system, which activates β3 adrenergic receptors on niche cells leading to decreased CXCL12 production ²⁵⁶. The extent to which MI-induced sympathetic activity directly influences HSPCs via their adrenergic receptors is not known. We do know a population of lineage⁻ c-kit⁺ sca-1⁺ CD150⁺ CD48⁻ HSCs, which are referred to as SLAM HSCs ³⁴, that increase proliferation within 48h after myocardial infarction ³²⁶. A subset of SLAM HSCs that express the chemokine receptor CCR2, which is regulated by the co-transcription factor Mtg16, show the highest proliferation rate ³²⁶. CCR2⁺ HSCs are myeloid biased, migrate to extramedullary sites, and give rise to MDPs and CMoPs, which eventually become monocytes ²⁶⁴ that leave the bone marrow in response to the chemokines CCL2 ³²⁷ and CCL7 ³²⁸. Bone marrow mesenchymal stem cells and peripheral B cells contribute to CCL2 and CCL7 production ^{296,327–329}. In addition to mature leukocytes, the bone marrow also releases HSPCs, which seed into extramedullary sites like the spleen, where they initiate extramedullary hematopoiesis ²⁶⁷. In response to myocardial infarction, the spleen also releases huge numbers of inflammatory leukocytes and overall contributes as much as 50% of all myeloid cells that enter the ischemic myocardium within the first day ²¹. Splenic monocyte release depends on angiotensin-2 signaling ^21,330.

After a few days, HSPCs released from the bone marrow in response to increased sympathetic activity ^256,331 seed the spleen, where they help form an extramedullary hematopoietic niche ²⁶⁷. Splenic mechanisms supporting extramedullary niche formation might include CD169⁺ macrophages that retain HSPCs via VCAM-1 ⁹⁹. While experimentally proven in mice with atherosclerosis and MI ^22,256,262, splenic hematopoiesis in humans has so far only been indirectly shown via increased ¹⁸F-FDG uptake in PET imaging after acute MI ^332,333. Overall, heightened systemic inflammation after MI increases monocyte production in the bone marrow and spleen. In ApoE^–/– mice, the resulting blood monocytosis leads to increased recruitment not only to the ischemic myocardium but also to preexisting atherosclerotic plaques. Via this mechanism, a first myocardial infarct accelerates atherosclerosis and might increase the likelihood of secondary ischemic events ²⁵⁶.

In agreement with these studies, clinical data indicate higher levels of innate immune system activity in patients with recurrent cardiovascular events ³³⁴. Translational evidence to confirm experimental animal observations in humans mostly derives from blood analyses and imaging, due to the limited availability of in vivo hematopoietic organ samples. Clinical studies revealed patients with acute MI have more circulating HSPCs in their blood, very similar to the observations in mice post MI ^256,331,335. Imaging studies described higher ¹⁸F-FDG-uptake in non-culprit plaques and hematopoietic tissue activation after myocardial infarction ^332,336,337. Similar to MI-induced acceleration of atherosclerosis in mice ²⁵⁶, coronary imaging in humans also showed accelerated lesion growth post MI ³³⁸. Furthermore, a retrospective study reported metabolic activity in bone marrow and spleen, measured via ¹⁸F-FDG uptake, positively correlated with subsequent cardiovascular events ³³². While these studies provide first insights, larger prospective studies focusing on immune pathways are necessary to conclusively test translatability and relevance of mouse findings in patients with myocardial infarction. Nevertheless, despite important insights (Figure 4), both the signaling pathways that expand myelopoiesis after MI and the pathways that limit the hematopoietic response to restore steady state conditions after MI remain key areas of research.

Figure 4: Myocardial infarction.

Leukocytes play important roles in the steady-state myocardium as well as during and after myocardial infarction (MI). The healthy myocardium contains various mature immune cell subsets, with macrophages being the most abundant population fulfilling important maintenance tasks that we are only beginning to understand. In the event of acute myocardial ischemia, a massive reaction of basically all cell types in the infarcted area occurs within minutes, including TNF release by degranulating mast cells, cytokine-, growth factor- and chemokine production (e.g. CCL2 and IL-1β) by activated and dying macrophages and fibroblasts, as well as release of damage-associated molecular patterns (DAMPs) by dead and dying cardiomyocytes, all of which lead to a strong expression of adhesion molecules, such as VCAM1 and selectins on endothelial cells. In consequence, the first day after an MI is characterized by a massive influx of neutrophils and monocytes from the blood stream into the myocardium. These innate immune cells secrete matrix metalloproteinases (MMPs) and are phagocytically active to clear dead cells and debris, and they also release inflammatory cytokines that sustain the inflammation. Around day 4, the inflammatory clean-up phase passes over to a reparative phase with disappearance of neutrophils and appearance of reparative macrophages that produce IL10, TGFβ, and VEGF to promote regression of inflammation, angiogenesis, fibrosis, and scar formation. Abbreviations: TNF - tumor necrosis factor; DAMPs - damage-associated molecular patterns; VEGF - vascular endothelial growth factor; TGFβ - transforming growth factor-β; IL10 - interleukin 10

Conclusions and Future Directions

Research conducted in recent decades has demonstrated that atherosclerosis, from its early initiation to its major complications, crucially depends on inflammatory leukocyte supply from the bone marrow niche. Technical and conceptual breakthroughs, including identification of highly-specific HSC markers, genetically engineered mouse strains for cell fate tracking and cell type-specific manipulations, advanced imaging techniques, and single-cell multiomics have dramatically expanded our knowledge of the cellular and molecular composition and interactions of leukocyte niche cell networks in the bone marrow, as well as in atherosclerotic plaques and infarcted myocardium. While we can now appreciate the complexity of such processes at unprecedented detail, we are still far from fully understanding such pathways, much less successfully targeting them therapeutically.

Future studies will need to focus on further unraveling the heterogeneity of bone marrow niche populations, identifying and characterizing additional niche factors, determining the timing of cell fate decisions, and tracking cells’ precise differentiation trajectories and relative contributions to mature leukocyte production. In the context of atherosclerosis, further work is needed to understand cell behavior and fate in atherosclerotic lesions as well as ischemic and remodeling myocardium, uncover MI-induced epigenetic changes in HSPCs, identify additional mechanisms through which cardiovascular risk factors accelerate atherosclerosis, and determine tissue-specific macrophage functions. This is an exciting area for the hematology of cardiovascular disease, with many discoveries yet to come.

Non-standard Abbreviations and Acronyms

ADRβ3

β3 adrenergic receptor

ApoE

apolipoprotein E

BM

bone marrow

CAR

cellCXCL12-abundant reticular cell

CCR

C-C chemokine receptor

CDP

common dendritic cell progenitor

CLP

common lymphoid progenitor

cMoP

common monocyte progenitor

CMP

common myeloid progenitor

CSF1

colony stimulating factor-1

CVD

cardiovascular disease

CXCL

CXC chemokine ligand

CXCR

CXC chemokine receptor

DAMP

damage-associated molecular pattern

G-CSF

granulocyte colony-stimulating factor

GMP

granulocyte-monocyte progenitor

GP

granulocyte progenitors

HFrEF

heart failure with reduced ejection fraction

HSC

hematopoietic stem cells

HSPC

hematopoietic stem and progenitor cell

IL

interleukin

ILC

innate lymphoid cell

IT-HSC

intermediate-term hematopoietic stem cell

LDL

low density lipoprotein

LEPR

leptin receptor

LMPP

lymphoid-primed multipotential progenitor

LT-HSC

long-term hematopoietic stem cell

MDP

monocyte-dendritic cell progenitor

MPO

myeloperoxidase

MPP

multipotent progenitor

Nes

Nestin

NG2

neural-glial antigen 2

NK cell

natural killer cell

MI

myocardial infarction

MSC

mesenchymal stem/stromal cell

ROS

reactive oxygen species

SCF

stem cell factor (SCF)

ST-HSC

short-term hematopoietic stem cell

TGFβ

transforming growth factor-β

THPO

thrombopoietin

TLR

toll-like receptor

TNFα

tumor necrosis factor α

VCAM1

vascular cell adhesion molecule 1

VEGF

vascular endothelial growth factor

vWF

von Willebrand factor