The origins and non-canonical functions of macrophages in development and regeneration

The immune functions of macrophages (MPs) were first highlighted by Elie Metchnikoff, who was awarded the Nobel Prize in Physiology or Medicine in 1908. He developed the now commonly accepted theory that phagocytic cells, such as MPs, monocytes (MOs) and neutrophils, fight against pathogens (reviewed by Cavaillon, 2011). Subsequently, during the 20th century, a lineage relationship between circulating blood MOs and the MPs found in inflamed tissues was discovered, first in 1930 in the amphibian larvae model and later in 1939 for mammals (Cavaillon, 2011). This was soon followed by myriad studies of macrophage activation and functions in inflammation.

In its most reductionist definition, inflammation is described as a process that is triggered in response to viral or bacterial infections. However, it is now commonly accepted that sterile inflammation (non-pathogen-induced inflammation) also occurs in multiple settings, including during development and in afflictions such as autoimmune disease and cancers, and even in aging (Jackaman et al., 2017). Moreover, it is clear that any form of tissue damage invariably triggers sterile inflammation, influencing outcomes spanning from functional restoration to fibrosis depending on the efficiency of the regenerative process. Importantly, MPs are now thought to play key roles in modulating such processes, and recently developed molecular tools have provided strong evidence for their ability to execute tissue-specific tasks beyond the response to pathogens (Davies et al., 2013). In this Review, we describe some of these non-canonical and organ-specific functions of MPs. We begin by providing an overview of the populations of MPs that exist and discussing their developmental origins. We then highlight how such MPs function in the context of tissue development, homeostasis and repair.

In summary, at least three developmentally distinct types of MPs exist in the body: yolk sac (YS)-derived tissue-resident MPs, fetal liver-derived tissue-resident MPs, and infiltrating bone marrow-derived MPs. All of these MPs emerge at different points during development and adulthood (Fig. 1) and play key roles in tissue development, growth, homeostatic maintenance and remodeling (reviewed by Epelman et al., 2014b).

During embryogenesis, MPs are detected in most organs and tissues including the brain, heart, liver and skeletal muscle. The progeny of these first tissue-resident MPs persist for the life of the organism. The precise ontogeny of embryonic MPs is still debated, but it has been elegantly demonstrated that at least one subset arises before the appearance of the first multipotent hematopoietic stem cells (HSCs) (Ginhoux and Guilliams, 2016). For example, during early embryogenesis in mice, ‘primitive’ hematopoiesis takes place in the extra-embryonic YS [embryonic day (E)6.5 to E11] and gives rise to MPs (Fig. 1), which are the only ‘white’ hematopoietic cells found in the embryo at this time point. Later, from E8.5 to E12, immature HSCs emerge from the aorta-gonad-mesonephros (AGM) and give rise to other early immune lineages (which do not include lymphoid cells) (Orkin and Zon, 2008). These immature HSCs eventually colonize the fetal liver (between E10-P1). Soon after birth (E19-P1), fetal liver HSCs migrate to the bone marrow and give rise to mature HSCs that are capable of generating all blood lineages, including MPs, in a process called ‘definitive’ hematopoiesis (E19-P1) (Fig. 1) (Palis and Yoder, 2001; Orkin and Zon, 2008). In addition to the fetal liver, the placenta has been revealed as a site of active hematopoiesis (between E10-E13) (Dzierzak and Robin, 2010). At each of these stages, MP precursors are generated and seed a specific subset of organs, often establishing local lineages that will last the organism's lifetime.

During adulthood, MOs/MPs originate in the bone marrow via a differentiation cascade that comprises multiple cellular intermediates. According to a classical model, HSCs give rise to common myeloid progenitors (CMPs) that generate additional progenitors that are further restricted in their developmental potential, i.e. granulocyte-monocyte progenitors (GMPs) and megakaryocyte-erythrocyte progenitors (MEPs) (Fig. 2A). However, this classical model has been challenged for years, and alternative models in which all other myeloid cells (neutrophils, eosinophils, mast cells), in addition to MOs/MPs, also come from GMPs has been suggested (Kierdorf et al., 2015; Ginhoux and Guilliams, 2016; Yamamoto et al., 2018) (Fig. 2A). Even today, the debate is not settled. For example, based on single cell RNA sequencing, Drissen and colleagues recently proposed that mast cells, basophils and eosinophils arise from a different, earlier progenitor (Gata1⁺) than do neutrophils and MOs (Flt3⁺) (Drissen et al., 2016; Sarrazin and Sieweke, 2016) (Fig. 2B). In parallel, the existence of MEPs has been questioned by reports demonstrating that erythrocytes and megakaryocytes can directly derive from HSCs, both during fetal development and during adulthood (Adolfsson et al., 2005; Notta et al., 2016) (Fig. 2B).

In conclusion, our understanding of the hematopoietic lineage is evolving fast, thanks to rapid progress in single cell-based techniques, and has highlighted that the MP lineage can arise at multiple points during development and adulthood (Figs 1 and 2).

Tissue-resident MPs are the most abundant immune cell present in most tissues, yet at resting state they are still rare compared with parenchymal cells. In 1968, Van Furth and Cohn proposed that tissue-resident MPs originate from circulating blood MOs (van Furth and Cohn, 1968). This hypothesis has been accepted for many years without more thorough investigation. More recently, however, the development of new murine models allowing the specific labeling of MO/MP populations in vivo (Box 1) has allowed researchers to explore the ontogeny and roles of the different subsets of MOs and MPs found in tissues.

For example, mice that express GFP from the CX3CR1 locus, which encodes the receptor for the chemokine CX3CL1, have provided a tool for following MOs and MPs within blood and tissues (Jung et al., 2000; Geissmann et al., 2003). The subsequent development of CX3CR1-creER mice has allowed lineage tracing studies, confirming that at least a subset of tissue-resident MPs is established during embryogenesis and self-renews in peripheral tissues during adulthood (Yona et al., 2013; Salter and Beggs, 2014). Furthermore, by using CCR2-KO mice, in which MOs lose their responsiveness to CCL2 and therefore fail to infiltrate tissues, Hashimoto and colleagues showed that blood-derived MPs are not required for maintaining lung-, spleen- and peritoneum-resident MP homeostasis (Hashimoto et al., 2013). Other lineage tracing models, such as Kit-Mer-Cre-Mer, Tek-Mer-Cre-Mer, Csfr1-Mer-Cre-Mer or Runx1-Mer-Cre-Mer, specifically target YS-derived hematopoietic cells. In these models, activation of Cre recombinase at E7 targets the YS hematopoietic system but at E8.5 and beyond targets the ‘definitive’ hematopoietic system. This allowed independent validation of the existence of two waves of erythro-myeloid progenitors contributing to tissue-resident MOs (Ginhoux et al., 2010; Hoeffel et al., 2012; Hoeffel and Ginhoux, 2015).

Additional studies have clarified that progenitors derived from the YS give rise to MPs in the brain (microglia), and to a fraction of tissue-resident MPs in other organs such as the heart, lung (alveolar MPs), liver (Kupffer cells), spleen, bone (osteoclasts) and skin (Langerhans cells) (Merad et al., 2002; Guilliams et al., 2013; Hashimoto et al., 2013; Jakubzick et al., 2013; Epelman et al., 2014b; Molawi et al., 2014; Gomez Perdiguero et al., 2015). Thus, with the exclusion of microglia, tissue-resident MP populations in other tissues comprise cells of multiple developmental origins (Fig. 1), at least some of which are maintained locally from early development onwards.

A correct number of tissue-resident MPs needs to be maintained during adulthood. Although the self-renewal of tissue-resident MPs is complex, tissue specific and not yet fully understood, we know that the long-term maintenance of tissue-resident MPs is essential (Aziz et al., 2009; Jenkins and Hume, 2014). Indeed, the disruption of MP numbers or their inflammatory state can lead to severe defects in organ function during life, leading to early aging (Linehan and Fitzgerald, 2015).

During adult life, a percentage of tissue-resident MPs, variable from tissue to tissue, is continuously renewed by blood-derived MPs. However, whether these cells should be considered as ‘tissue-resident’ MPs is a matter for debate. In addition, this process does not occur to the same extent in all tissues. For example, microglia are capable of self-renewal in their tissue of residence and are completely independent of definitive hematopoietic progenitors for their maintenance, and circulating bone marrow-derived myelomonocytic cells only enter the CNS in the context of disease (Ajami et al., 2007).

Tissue-resident MPs also display distinct molecular characteristics depending on their tissue of residence, but the links between them are unclear (Ayata et al., 2018). For example, do resident MPs in different tissues come from the same type of progenitor during development, suggesting that their specialization takes place due to environmental signals present in the target tissue? Or do specific progenitors exist, with different predispositions for infiltrating specific organs, suggesting that at least some of the reported differences are cell autonomous? In this context the existence of a niche, composed of tissue-specific cells, which could induce MP specialization has been proposed (Gautier and Yvan-Charvet, 2014; Guilliams and Scott, 2017). The fate of resident MPs following catastrophic damage also appears to vary depending on the tissue. For example, following ablation in the CNS, microglia are restored from a small number of surviving YS-derived cells without contribution from the bone marrow (Elmore et al., 2014; Huang et al., 2018). In contrast, YS-derived cells in the heart, which form a sizable subset of tissue-resident MPs, are replaced by blood-derived MOs (Epelman et al., 2014b). It is unclear whether these ‘new’ tissue-resident MPs have the same function as the embryo-derived ones (Blériot et al., 2015; Machiels et al., 2017). For further consideration about tissue-resident MP ontogeny, fate and self-renewal, we refer the reader to recent reviews (Sieweke and Allen, 2013; Ginhoux and Guilliams, 2016; Kierdorf and Dionne, 2016).

In summary, the ontogeny and behavior of tissue-resident MPs has started to be unlocked over the past decade, revealing that many of these cells originate during embryonic development. Although many questions remain, the combination of single cell-based technologies and new specific markers for adult tissue-resident cells should rapidly lead to their answers.

In addition to the tissue-resident MPs discussed above, MPs can arise from MOs that are present within the blood and that infiltrate into tissues. Two types of MOs, classified according to their expression of the marker Ly6C, are found in the circulation in both mice and humans. Their ratio in the blood is dynamic and changes in response to various stimuli (e.g. bacterial or virus infection, tissue damage and sterile inflammation). Ly6C⁺ MOs originate directly from bone-marrow progenitors, whereas Ly6C⁻ MOs likely arise from circulating Ly6C⁺ MOs (Yona et al., 2013). Functionally, these two populations are different. The Ly6C⁺ MO subset responds to damage by infiltrating the tissue and then differentiating into MPs. In contrast, Ly6C⁻ MOs do not infiltrate tissues, but rather patrol the vascular system to detect and likely remove damaged endothelial cells within vessels (Geissmann et al., 2003; Carlin et al., 2013; Jakubzick et al., 2013).

After a long debate on the origins of blood-derived MPs found in damaged tissue, it is now commonly accepted that infiltrating MPs (in contrast to tissue-resident MPs) only arise from Ly6C⁺ MOs recruited from the bloodstream, and are associated with the scavenging of damaged tissue components, wound healing and tissue remodeling (Arnold et al., 2007; Mounier et al., 2013; Juban et al., 2018).

After tissue injury, MO infiltration is dramatic, and essentially equivalent in the majority of tissues, with the noticeable exception of the CNS, in which infiltration following sterile damage is limited because of the blood-brain barrier (Engelhardt and Ransohoff, 2012). Infiltration involves CCR2, the receptor for the tissue-damage induced chemoattractant CCL2. CCR2 is highly expressed in Ly6C⁺ MOs and is required for them to infiltrate tissues. Importantly, studies of CCR2-KO mice (Box 1), in which MOs fail to infiltrate in response to damage, have revealed that blood-derived MPs are required for the efficient regeneration of many tissues (Boring et al., 1997).

Functionally, blood-derived MPs are plastic cells that can adopt different transcriptional programs depending on the cytokines that are present in the microenvironment, assuming a range of functions spanning from the promotion of inflammatory and adaptive immune responses to their inhibition in favor of tissue regeneration (Epelman et al., 2014b). In sterile damage, these ‘polarization’ states are often not the result of classical, irreversible binary cell fate decisions but rather are sequentially acquired according to a well-defined temporal succession. Early after damage, Ly6C⁺ MOs enter the tissue and differentiate into pro-inflammatory MPs (classically called ‘M1’ MPs, see Box 2). These cells, which are Ly6C⁺CCR2⁺CX3CR1^midF4/80 (Adgre1)^mid (Table 1), scavenge apoptotic debris but also modulate tissue-specific stem cells through the secretion of multiple factors (Freire-de-Lima et al., 2000; Johann et al., 2006; Arnold et al., 2007). Next, these MPs transition to a pro-restorative status (or the ‘M2’ state, see Box 2 and Table 1). These Ly6C⁻CCR2⁻CX3CR1^highF4/80^high cells are associated with fibrosis formation, wound healing and tissue regeneration.

Historically, these phenotypes and related polarization states have been defined mostly based on in vitro results. It is therefore unclear how similar the cells found in damaged tissues are to the MPs that are generated in vitro by exposing bone-marrow progenitors to specific cytokine cocktails (Martinez et al., 2008). For this reason, and also reflecting a recent rejection of the M1/M2 terminology in the literature (Varga et al., 2013; Varga et al., 2016a,b), pro-inflammatory M1 MPs will be referred to hereafter as Ly6C⁺ MPs, and pro-restorative M2 MPs will be called Ly6C⁻ MPs (see Box 2).

Although significant differences are maintained between tissue-resident MPs and MO-derived infiltrating MPs, these are not obvious, and phenotypically the cells are very similar. This raises questions about the overlap in biological functions between these two populations (reviewed by Epelman et al., 2014b). As an example, in the bone marrow itself, MPs are required for correct hematopoiesis. Their depletion affects the HSC niche and induces HSC mobilization in the blood (Winkler et al., 2010; Gustafsson and Welsh, 2016). Moreover, MPs are also required to be in direct contact with erythroblasts in the bone marrow for proper erythropoiesis (Chow et al., 2013). These two examples demonstrate that tissue-resident MPs are crucial even for their own homeostasis. However, animals congenitally deficient for tissue-resident MPs have reduced survival, making the study of MP function during embryogenesis or postnatal growth a difficult task. For example, mice carrying a spontaneous mutation in Csf1 (op/op mice), a factor that is essential for the maintenance of mature MPs, present many abnormalities (Dai et al., 2002). This is owing to a lack of osteoclasts, which are multinucleated cells that are formed by fusion of MPs with each other and are required for bone resorption. As a result, bone deposition is excessive in these mice, cavities for bone marrow are not formed and hematopoiesis is deeply altered (Dai et al., 2002), making it hard to distinguish direct and indirect effects of the lack of MPs. Likewise, deletion of the hematopoietic-specific transcription factor PU.1 (Spi1) causes lethality in the late embryo or soon after birth. However, by using a conditional LysM-Cre PU.1 floxed model, it has been demonstrated that PU.1 is dispensable for HSC development, allowing the community to investigate hematopoiesis more thoroughly (McKercher et al., 1996; DeKoter et al., 1998; Dakic et al., 2005). Similarly, other conditional knockout mouse models (see Box 1) have aided the study of tissue-resident MPs, revealing how these cells arise and are maintained, and highlighting organ-specific roles for these cells during embryogenesis and early development.

Microglia, first described in 1932 (del Rio-Hortega, 1932), are the resident MPs of the CNS. Microglia are phenotypically identical to other MPs and share the ability to respond to damage and to phagocyte cell debris. However, as mentioned above, microglia have distinct developmental origins: although the timing of their colonization of the brain is still debated, and is likely to happen at multiple stages of embryonic life, it is now accepted that microglia derive mainly from the YS at E8.5-E9 (Alliot et al., 1999; Herbomel et al., 2001; Ajami et al., 2007; Chan et al., 2007; Ginhoux et al., 2010; Hashimoto et al., 2013; Salter and Beggs, 2014), before the origins of the first HSCs and of most adult MPs. Interestingly, microglia formation is independent of the CSF1/CSF1R axis, but rather depends on IL34, an alternative ligand for CSF1R (Erblich et al., 2011; Schulz et al., 2012).

Microglia have important functions during development and adult life (reviewed in detail by Hammond et al., 2018). As an example, a lack of microglia in Csf1r^−/− mice results in functional deficiencies, for example in olfaction processing, pointing to a requirement for these cells during development (Erblich et al., 2011). Indeed, the number of microglia increases significantly during brain development, owing to both local proliferation and additional waves of infiltration (Ginhoux et al., 2010; Erblich et al., 2011). In addition to playing a role in embryonic and fetal brain development, microglia are required for normal postnatal development, likely by modulating interneuronal connections through synaptic pruning (Graeber and Streit, 2010; Tremblay et al., 2010; Paolicelli et al., 2011; Fig. 3). This phenomenon is thought to be modulated by CX3CR1, which is expressed at high levels in microglia. Indeed CX3CR1-KO animals have a higher number of unconnected synapses, which is reminiscent of the immature brain and similar to what can be seen in neurodevelopmental disorders (Graeber and Streit, 2010; Tremblay et al., 2010; Paolicelli et al., 2011). Moreover, microglia are able to recognize CD47, which appears to modulate pruning of neuronal projections (Lehrman et al., 2018; Fig. 3). CD47-KO mice exhibit increased engulfment of neurons by microglia, resulting in a decrease in synaptic connections (Lehrman et al., 2018). Interestingly, the phagocytic activity of microglia is different depending on their location in the CNS, and depends on epigenetic changes likely acquired locally (Ayata et al., 2018).

Microglia are also involved in angiogenesis and vessel sprouting during brain development, acting via the secretion of VEGF (Fig. 3). This well characterized pro-angiogenic function is shared with other non-CNS resident MPs, and represents an important pro-cancer function of tumor-associated MPs (TAMs) (Fantin et al., 2010). Finally, by analogy with other tissue-resident MPs, microglia are thought to acquire different functions depending on the local microenvironment, a process often called polarization. Such proposed functions range from inflammatory roles, involving the release of reactive oxygen species (ROS), nitric oxide (NO) or TNFα, to pro-neurogenic roles. It has also been proposed that, in some cases, microglia activation can have a direct neurotoxic effect (Hellwig et al., 2013). However, the ability of microglia to polarize, and in particular to polarize to functional states that resemble those described for MPs in other tissues, is still controversial and based more on analogy than solid data (Wang et al., 2016).

Tissue-resident MPs in the heart comprise a mixture of cells from both YS and fetal liver origins (Epelman et al., 2014b). The self-renewal capacity of these cardiac-resident MPs is unclear and appears to depend on their anatomical location. Indeed, whereas some publications report a slow but progressive replacement of heart-resident MPs by blood-derived MPs with age (Epelman et al., 2014a; Molawi et al., 2014), others suggest that less than 1% of the MPs localized in the atrioventricular bundle are replaced (Hulsmans et al., 2017) (Fig. 4). More recently, it has been demonstrated that cardiac TIMD4⁺ MPs are self-renewing with little input from infiltrating MPs, whereas cardiac TIMD4⁻ MPs are almost fully replaced by infiltrating MPs (Dick et al., 2019). Whether cardiac MPs are required during normal development is still unclear. In contrast to the adult heart, the neonatal heart is able to fully regenerate, and the depletion of circulating MOs (using clodronate liposomes) blocks this regenerative capacity (Porrello et al., 2011; Aurora et al., 2014). This could be because of the absence of angiogenesis in MP-depleted mice (Aurora et al., 2014). Although the specific molecular mechanism underlying this observation is not known, the authors' suggestion that a lack of MP-derived VEGF may be crucial appears to be reasonable (Fig. 4). This function matches the pro-angiogenic function of microglia described above and, by analogy, also suggests that cardiac MPs are required for the establishment of a full blood coronary tree during development (Fantin et al., 2010). Lastly, it has been described that, at resting state, heart-resident MPs are particularly abundant in the atrioventricular node, leading to the proposal that they actively communicate with cardiomyocytes via GAP junctions to regulate cardiac rhythm (Hulsmans et al., 2017) (Fig. 4), again demonstrating the importance of tissue-resident MPs.

The skin is composed of three layers: the epidermis, which directly faces the external environment, the dermis, which contains hair follicles, and the hypodermis, which contains sweat glands, fat and blood vessels. Various populations of MPs and dendritic cells are known to be present in the skin but, owing to their phenotypic overlap, the distinction between them is ambiguous (Henri et al., 2010; Tamoutounour et al., 2013). Aside from dendritic cells, at least two types of tissue-resident MPs are present in the skin: Langerhans cells are mainly present in the epidermis, whereas cutaneous (or dermal) MPs are more present in the dermis (Di Meglio et al., 2011; Tamoutounour et al., 2013; Mojumdar et al., 2014). Of note, Langerhans cells share properties with both MPs and dendritic cells (e.g. self-renewal and T cell stimulation), making them an unclassified cell type (i.e. neither MP nor dendritic cell) (Doebel et al., 2017). In terms of their developmental origins, both Langerhans cells and cutaneous MPs are highly heterogeneous. Indeed, Langerhans cells are first derived from YS progenitors but are later gradually replaced by precursors from fetal liver hematopoiesis (Fig. 1) (Hoeffel et al., 2012). In contrast, cutaneous MPs appear to originate from the fetal liver and are then gradually replaced by circulating Ly6C⁺ MOs (Varol et al., 2009) (Fig. 1). Langerhans cells are also able to self-renew, but Ly6C⁺ MOs can be recruited to replace these cells after damage (Tamoutounour et al., 2013).

Until after birth, tissue-resident MPs (both Langerhans cells and cutaneous MPs) are the only immune cells to appear in the skin (Tamoutounour et al., 2013). So far, the function of tissue-resident MPs during development of the skin is unknown, however we hypothesize that they are required for extracellular matrix formation and/or deposition and skin layer differentiation. The number of these tissue-resident MPs in the skin also changes during normal hair cycles, suggesting roles for them during hair follicle development. For example, the number of Langerhans cells and cutaneous MPs increases in anagen (the active hair growth phase) and decreases in telogen (the resting phase), because of MP apoptosis (Osaka et al., 2007; Wang et al., 2017). Concomitantly, an increase in Wnts expressed by MPs is found locally, activating epithelial cells to initiate the anagen phase (Paus, 1998; Castellana et al., 2014) and suggesting that MPs modulate the activity of hair follicle stem cells. Consistent with this notion, it has been shown that Ly6C⁻ MPs secrete TGFβ1 to activate follicular stem cells, and thus hair regrowth, in the context of wound healing (Rahmani et al., 2018). Interestingly, this phenomenon is dependent on CX3CR1 but not CCR2 (Rahmani et al., 2018), suggesting it may not involve blood-derived, infiltrating cells. In addition, MPs can modulate hair cycle and hair growth after plucking via the secretion of TNFα (Chen et al., 2015).

Beyond their important functions in immunity, tissue-resident MPs in the spleen play crucial roles related to the spleen's function as the filter of the bloodstream (Fig. 5). Thus, it is perhaps not surprising that various subsets of these cells exist, segregated in the different functional areas of this organ. Splenic-resident MPs originate from the fetal liver, are Spi1 (PU.1-related transcription factor)-dependent and self-renew mostly independently of blood-derived cells (Kohyama et al., 2009; Hashimoto et al., 2013; Yona et al., 2013). Tissue-resident MPs within the red-pulp region of the spleen are CD11b (Itgam)^low, F4/80⁺ and CD206 (Mrc1)⁺ and have important functions in iron processing connected with their main function: the clearance of damaged red blood cells (Kohyama et al., 2009). In contrast, MPs in the white-pulp region express different markers (e.g. CD68 and MERTK) and mainly phagocytose lymphocytes (Kohyama et al., 2009). Defects in the function of these white-pulp MPs lead to B cell accumulation and auto antibody development, providing a link between their scavenger and immune roles (Khan et al., 2013).

The liver contains a specialized population of resident MPs called Kupffer cells, named after Karl Wilhelm von Kupffer who discovered them in 1876 (von Kupffer, 1876). Until recently, it was commonly accepted that, between E10.5 and E12, AGM-derived HSCs give rise to fetal liver MOs, which later differentiate into Kupffer cells (Gomez Perdiguero et al., 2015). However, this model was recently disputed. Indeed, it was shown that YS-derived erythro-myeloid progenitors are also able to generate a common circulating precursor, which colonizes the embryo from E9.5 and directly gives rise to Kupffer cells in a CX3CR1-dependent way (Schulz et al., 2012; Mass et al., 2016). After liver colonization, a distinct transcriptional program (i.e. different to that in YS-derived cells) is activated, specifically involving the transcription factor inhibitor of DNA binding 3 (ID3) and leading to the development of Kupffer cells that are endowed with self-renewal activity (Mass et al., 2016).

One important non-immune function of Kupffer cells is the modulation of metabolism in hepatocytes via PPARγ, preventing the pathogenic accumulation of lipids and thus hepatic steatosis (Desvergne, 2008; Kang et al., 2008; Odegaard et al., 2008). After injury, the main function of Kupffer cells is to clear dead cells and pathogens in order to protect the liver (Nguyen-Lefebvre and Horuzsko, 2015). Following acute damage (e.g. carbon tetrachloride injection or bile duct ligation), Kupffer cells (CD11b^lowF4/80⁺) are activated by damage-associated molecular patterns (DAMPs), upregulate Ly6C, an event that may be indicative of activation rather than polarization into M1, and contribute to the attraction of circulating MOs (by secretion of CX3CL1, CCL2 and CCL5) and other cells such as neutrophils, basophils or natural killer cells (via CXCL16) (Nguyen-Lefebvre and Horuzsko, 2015; Wynn and Vannella, 2016) (Fig. 5). Basophils are also of importance for liver repair, as they are a source of IL4, a pro-proliferative cytokine for hepatocytes (Blériot et al., 2015). Thus, appropriate coordination between resident Kupffer cells and cells infiltrating from the bloodstream is required for damage resolution and maintenance of liver function.

After completion of postnatal growth, organs and tissues need to maintain homeostasis for a prolonged period of time to preserve function and integrity. The involvement of blood-derived MPs in organ and tissue homeostasis is paramount. A detailed discussion of the specific roles of these MPs in every tissue is beyond the scope of this review. We therefore focus below on the roles of blood-derived MPs in the homeostasis, as well as the regeneration, of four well-studied tissues: skeletal muscle, cardiac muscle, the liver and skin (Fig. 6).

Despite the fact that skeletal muscle was one of the first tissues in which the crucial roles played by MPs in damage-induced regeneration were revealed, a developmental role for macrophages in this tissue has not been investigated. MPs are present within muscle at steady state and are found in the perimysium (connective tissue surrounding muscle fascicles) and the epimysium (fascia surrounding the muscle) (Saclier et al., 2013b). However, a lack of tools to discriminate tissue-resident versus infiltrating MPs in skeletal muscle has made it difficult to ascertain the precise role of tissue-resident MPs in development. In contrast, it is well established that infiltrating MOs are required for efficient skeletal muscle regeneration following acute damage (Fig. 6A). Deletion of CCR2 delays the infiltration of myelomonocytic cells after damage, causing a strong impairment of skeletal muscle regeneration and the appearance of diffused fibrosis, demonstrating that these cells are crucial for efficient healing (Warren et al., 2005). These findings were validated by depleting CD11b⁺ cells using the CD11b-DTR mice model (Box 1; Jung et al., 2000; Arnold et al., 2007).

Early after skeletal muscle damage, Ly6C⁺ MPs express various inflammatory markers (see below) but lose Ly6C within 48-72 h (Arnold et al., 2007; Mounier et al., 2013; Saclier et al., 2013a,b; Varga et al., 2016a,b). This phenotype switch, often called skewing, also occurs in multiple other tissues (e.g. heart, liver or kidney) (Sica et al., 2015) and is regulated at the DNA, RNA and protein levels throughout the regenerative process (Ruffell et al., 2009; Mounier et al., 2013; Varga et al., 2016a,b). Soon after infiltration, Ly6C⁺ MPs also express molecules such as TNFα (at high levels), IL6 and IL1β. Beyond their well-known roles as inflammatory mediators, these factors also act as mitogens, stimulating myogenic cell proliferation and inhibiting their differentiation/fusion into damaged fibers (Sonnet et al., 2006; Arnold et al., 2007; Saclier et al., 2013a,b). It is conceivable that, in the context of sterile damage, the non-inflammatory functions of these factors may be prevalent.

Another key function of Ly6C⁺ MPs, more in line with their classical role as scavengers, is the removal of debris via their phagocytic capacities. Phagocytosis also plays a role in triggering the transition of MPs to a pro-restorative phenotype through the activation of signaling pathways linked to metabolic control such as AMP-activated kinase (AMPK) (Mounier et al., 2013). As a result of skewing, the cytokines secreted by MPs change dramatically: inflammatory factors such as TNFα are downregulated whereas pro-regenerative factors such as IL10, IGF1, VEGF, and TGFβ are produced and stimulate muscle cell survival, differentiation, fusion and fiber growth (Fig. 6A) (Arnold et al., 2007; Mounier et al., 2013; Saclier et al., 2013a; Tonkin et al., 2015). More recently, it has been shown that the transcription factor PPARγ is required for such pro-restorative effects and, in particular, for secretion of the TGFβ family member growth differentiation factor 3 (GDF3). Indeed, suppression of either PPARγ or GDF3 leads to a delay in the anabolic processes taking place in the late phase of regeneration (myogenic cell fusion and myofiber growth) (Fig. 6A) (Varga et al., 2016a,b). Altogether these results demonstrate that the metabolic and functional states of MPs are highly correlated (Mounier et al., 2013; Varga et al., 2016a,b).

During the later stages of skeletal muscle regeneration, vessel remodeling is required to restore muscle function, which also appears to involve MPs (Zordan et al., 2014; Latroche et al., 2017). Ly6C⁻ MPs express matrix-remodeling enzymes such as matrix metallopeptidases (MMP) 2, 13 and 14, and VEGF, and are therefore good candidates to mediate vessel remodeling (Zordan et al., 2014). However, the precise roles of Ly6C⁺ and Ly6C⁻ MPs or even of tissue-resident MPs in damage-induced angiogenesis have yet to be addressed in depth.

Ly6C⁺ MPs also play a role in regulating fibrosis and fibrogenic differentiation during muscle regeneration. During early regeneration, stromal progenitors called fibro/adipocytic progenitors (FAPs) expand dramatically to establish a trophic environment but are then cleared, thus preventing their differentiation and the formation of fibrofatty infiltration. The clearance of these cells is mediated by Ly6C⁺ MPs through the secretion of TNFα (Lemos et al., 2015). In contrast, TGFβ, an anti-inflammatory cytokine mainly secreted by Ly6C⁻ MPs at later stages of regeneration, provides protection to FAPs from apoptosis (Lemos et al., 2015), highlighting how crucial the timing of the switch from Ly6C⁺ to Ly6C⁻ MPs is for the restoration of tissue function. This is also the case in the heart, where FAPs and MPs collaborate in the response to damage (Wang et al., 2015). Notably, this switch may be perturbed in the context of regenerative diseases such as Duchenne Muscular Dystrophy (DMD; see Box 3).

Cardiac damage, such as myocardial infarction, triggers massive inflammation. However, in this context, tissue-resident MPs die at the site of damage and the infiltration of new MPs that give rise to tissue-resident MPs is required for cardiac tissue repair (Heidt et al., 2014). As in other tissues, it is now accepted that damage recruits a unique wave of Ly6C⁺ MOs, followed by their skewing into a Ly6C⁻ profile (Fig. 6B; Hilgendorf et al., 2014, Pinto et al., 2014; Frangogiannis, 2015).

Consistent with the role of MPs in supporting cardiac angiogenesis during neonatal development, infiltrated MPs are capable of stimulating cardiac angiogenesis following perinatal ischemic damage (Aurora et al., 2014). However, in adult infarction, MPs mainly promote matrix deposition and support scarring (Fig. 6B) (Aurora et al., 2014). These differences in MP function in neonatal versus adult heart repair could be because of distinct phenotypes of MPs. Indeed, at least some tissue-resident MPs in the heart are of embryonic origin, and are phenotypically more similar to Ly6C⁻ MPs than to Ly6C⁺ MPs (Aurora et al., 2014). Again, depletion of tissue-resident or infiltrating MPs before or following cardiac injury has been performed in many ways. Whereas depletion of circulating MOs at an early point in repair negatively affects cardiac repair, depleting them late after myocardial infarction in fact improves cardiac recovery, highlighting the changing roles that these cells play at different stages of the response to damage (Van Amerongen et al., 2007; Heidt et al., 2014; Sager et al., 2016). Moreover, depletion of tissue-resident MPs using CX3CR1-DTR mice affects cardiac repair after myocardial infarction in the adult. This appears to be because of the massive infiltration of new MPs and the proliferation of the remaining tissue-resident MPs (Dick et al., 2019). To conclude, it appears that both tissue-resident MPs and infiltrating MPs have important functions in cardiac repair, and boosting one or the other systems at a precise time could be a potentially useful therapeutic tool.

Following skin damage, MPs are recruited and actively participate in wound healing (Fig. 6C). Ly6C⁺ MOs infiltrate the skin, providing immune protection against external pathogens, but also act on stromal cells, keratinocytes and endothelial cells (Wang et al., 2017). Again, depletion of circulating MOs using GM-CSF-KO, CD11b-DTR or CCR2-KO mice highlights their importance in this process (Fang et al., 2007; Mirza et al., 2009). Indeed, angiogenesis is strongly decreased owing to CD31⁺ cell apoptosis (likely because of decreases in TGFβ and VEGF secretion), strongly delaying wound healing (Lucas et al., 2010). Infiltration of circulating MOs (and other immune cells) to the site of healing is driven by activated keratinocytes and the secretion of several chemokines such as CXCL10, CXCL11, CCL5 and CCL20 (Banno et al., 2004). In return, MPs activate keratinocyte proliferation and reepithelization via the secretion of members of the epidermal growth factor (EGF) family (Hancock et al., 1988; Shirakata, 2005). Although this direct interaction has been demonstrated in vitro, it is not as clear-cut in vivo (Grose and Werner, 2004).

Interestingly, wound-induced macrophages acquire a profile that differs from the aforementioned Ly6C⁺/Ly6C⁻ progression that is observed in other tissues. Indeed, from day 1 to day 7 after wounding, MPs express CD206, arginase 1 and Ym1 (Chil3), which are markers typical of pro-regenerative cells that appear at later stages of regeneration in other tissues (Mounier et al., 2013). However, IL4 and IL13, which are cytokines described as being required for an alternative MP activation pathway, are not present in the wound, which suggest that other types of cytokines must be at play. Ly6C and TNFα are downregulated, whereas the secretion of TGFβ and VEGF is increased to accelerate the process of wound healing, as described above for muscle (Daley et al., 2010). TNFα secreted by MPs is important for the activation of hair follicle stem cells after damage (Wang et al., 2017). Indeed, whereas TNFα-deficient mice show delays in hair growth, TNFα actually stimulates the telogen-anagen transition required for hair growth (Wang et al., 2017). To conclude, during wound healing, MPs are the source of multiple cytokines and growth factors acting on hair follicle stem cells, endothelial cells and keratinocytes.

Infiltrating MPs play unique functional roles in both the resting and regenerating liver (Fig. 6D). For example, Ly6C⁺ MOs engulf stressed red blood cells in the bloodstream or the spleen, migrate to the liver and differentiate into Ly6C⁺ MPs in a process supported by CCL2, CCL3, and Csf1 released from Kupffer cells. These iron-charged MPs express high levels of ferroportin1, allowing them to dispense iron to hepatocytes (Gammella et al., 2014; Theurl et al., 2016).

As in most other tissues, MPs play key roles during regeneration in the liver, and undergo a skewing process essentially identical to that described for skeletal muscle, and likely just as important. CD11b-DTR-mediated myelomonocytic cell depletion at the early stages (the acute damage phase) of regeneration leads to reduced fibrosis, whereas depletion at late stages (the recovery phase) leads to the impairment of matrix degradation, inducing more fibrosis (Duffield et al., 2005). It has also been shown that blood-derived Ly6C⁻ MPs are present in large numbers during fibrosis resolution, which they likely promote through their secretion of MMP enzymes (Ramachandran et al., 2012). In CCR2-deficient mice, in which MO infiltration is reduced, damage-induced liver fibrosis is increased, as occurs in the case of cardiac and skeletal muscle, suggesting that loss of the matrix remodeling function of Ly6C⁻ cells has a dominant effect (Mitchell et al., 2009). As in most other tissues that have been analyzed, liver MPs also directly influence the function of tissue-specific stem/progenitor cells. For example, it was demonstrated that, in a model of biliary and hepatocyte regeneration, blood-derived MPs secrete Wnt ligands and remodel extracellular matrix to induce hepatocyte differentiation of progenitor cells (Boulter et al., 2012). After the resolution of inflammation, it is possible that blood-derived MPs replace the missing Kupffer cells, thereby maintaining a pool of tissue-resident MPs ready for further needs (Blériot et al., 2015).

The roles of MPs in disease are so varied that it is not possible to provide a full and comprehensive review here. However, as these cells are crucially important in many human afflictions, we have selected below a few examples, illustrating their non-immune functions in pathological conditions, and their interactions with tissue-resident stem cells in these contexts.

Type II diabetes, which is associated with obesity, is on the rise and accounts for a significant number of deaths every year. Obesity is associated with inflammation, and an accumulation of MPs in adipose tissue is thought to predispose to insulin resistance (Weisberg et al., 2003). In this context, it is of interest that high fat diet-induced insulin resistance is reduced in mice lacking the myelomonocytic integrin CD11b, and this correlates with an increase in adipose tissue-resident MP proliferation and an inhibition of MO recruitment from the circulation (Zheng et al., 2015). This suggests that Ly6C⁺ MPs freshly immigrated from the bloodstream may play a pathogenic role in Type II diabetes. Consistent with the idea that they modulate adipose metabolism, MPs have been proposed to contribute to thermogenesis regulation in brown adipose tissue by secreting catecholamine, although this controversial notion has recently been questioned (Nguyen et al., 2011; Fischer et al., 2017).

Obesity also causes adipocyte apoptosis and thus damages the signals that attract Ly6C⁺ MOs, leading to their significant increase in white adipose tissue. This may further contribute to adipocyte enlargement and apoptosis, perpetuating a vicious circle (Sun et al., 2011). As Ly6C⁺ MPs engulf dead adipocytes, they become engorged by fatty acids, resulting in the adoption of a highly inflammatory phenotype including secretion of TNFα and/or IL6. This increase in inflammatory cytokines in the blood stream results in chronic systemic inflammation and contributes to insulin resistance (Lumeng et al., 2007). Thus, increasing the skewing of MPs into a more mature phenotype (Ly6C⁻) may represent a promising therapy to treat obesity-induced chronic inflammation and its effects.

Tissue-resident MPs in the liver, Kupffer cells, are also involved in metabolic homeostasis and related pathologies. Kupffer cells secrete high levels of IL4 and IL13, possibly because of their low levels of PPARγ, a transcription factor that promotes the acquisition of an anti-inflammatory phenotype and the loss of Ly6C. The acquisition of this pro-inflammatory phenotype contributes to activating fatty acid oxidation by hepatocytes and eventually leads to liver failure due to lipid accumulation (Kang et al., 2008; Odegaard et al., 2008).

MPs also play a key role in the pathogenesis of autoimmune metabolic diseases. For example, pancreatic islets of Type II diabetic patients and rodent models are massively infiltrated by Ly6C⁺ MOs and MPs, likely attracted by cytokines produced locally by autoimmune T cells. In turn, IL1β, CCL2 and CX3CL1 released from these Ly6C⁺ MPs contribute to pancreatic dysfunction by recruiting new inflammatory TNFα-secreting Ly6C⁺ MOs, and possibly starting a vicious cycle (Ehses et al., 2007).

In summary, it appears likely that MPs mediate a number of the pathogenic effects that inflammation has on metabolism, and patients may benefit from treatments modulating the functional state of these cells.

The more we study MPs, the more we realize that the early simplistic views of these cells as ‘bug gobblers’ only describes a small portion of their core functions. Indeed, we now know that these cells act as a distinct disseminated tissue that is capable of influencing the local microenvironment as well as priming the systemic response to various types of damage. Acting via the carefully timed secretion of a variety of molecular signals that modulate the activity of tissue-resident cells, notably stem cells, MPs are involved in a large number of developmental processes. Importantly, they also play key roles in most events that lead to the reactivation of such processes, including during adult regeneration and in the context of disease. In line with these diverse functions, it is becoming apparent that MPs are far more heterogeneous and significantly more complex than can be captured by current classifications. The deployment of emerging technologies allowing high-throughput single cell analysis will be required to match phenotype to function and to identify interventions that may modulate them specifically.

We thank Andrew Hagner for inspiring discussions. Some images of cell types, organs and tissues were obtained from smart.servier.com and modified.