Opinion
Homeostasis in the mononuclear phagocyte system Stephen J. Jenkins1 and David A. Hume2 1
University of Edinburgh, Centre for Inflammation Research, Edinburgh EH16 4TJ, UK University of Edinburgh, The Roslin Institute and Royal (Dick) School of Veterinary Studies, Easter Bush Campus, Midlothian, EH25 9RG, UK
2
The mononuclear phagocyte system (MPS) is a family of functionally related cells including bone marrow precursors, blood monocytes, and tissue macrophages. We review the evidence that macrophages and dendritic cells (DCs) are separate lineages and functional entities, and examine whether the traditional view that monocytes are the immediate precursors of tissue macrophages needs to be refined based upon evidence that macrophages can extensively self-renew and can be seeded from yolk sac/foetal liver progenitors with little input from monocytes thereafter. We review the role of the growth factor colony-stimulating factor (CSF)1, and present a model consistent with the concept of the MPS in which local proliferation and monocyte recruitment are connected to ensure macrophages occupy their welldefined niche in most tissues. Introduction The MPS was originally defined as a family of cells derived from a pluripotent progenitor in the bone marrow and includes committed bone marrow progenitors, blood monocytes, and tissue macrophages; the latter comprising around 10% of total cells in every organ in the body [1,2]. MPS cells share many features, notably their phagocytic activity, but are also extremely plastic in their patterns of gene expression, defying identification based upon surface markers [3]. The proliferation and differentiation of MPS cells is controlled by macrophage CSF1 and interleukin (IL)-34, acting through a common receptor, CSF1R [4]. The traditional view of the MPS is that the major proliferative compartment is within the progenitors in the marrow, whereas blood monocytes provide the immediate precursors to replace tissue macrophages. This view has been challenged by several recent studies that, in effect, divide the MPS into separate cell lineages arising at different stages of embryonic development. The relationship of DCs with the MPS continues to be controversial; overlap of function and marker expression with monocyte-derived cells has made these cells difficult to delineate from macrophages, particularly in nonlymphoid tissues. Recent advances have provided evidence of a Corresponding author: Hume, D.A. ([email protected]). Keywords: macrophage; dendritic cell; proliferation; hematopoiesis; colony-stimulating factor 1. 1471-4906/ ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.it.2014.06.006
358
Trends in Immunology, August 2014, Vol. 35, No. 8
distinct precursor for DCs that would allow their definition as a separate haematopoietic lineage. It is therefore timely to ask whether these findings necessitate a complete revision of the MPS concept. The complex relation between DCs and macrophages One challenge to the concept of the MPS was the identification of the antigen-presenting DCs. Should these cells be included in the MPS, and if not, how can they be distinguished from macrophages? Efforts to define DCs have been based on function, molecular markers, and dependence upon growth and transcription factors. Each of the criteria are discussed below. The definition of a macrophage, or a mononuclear phagocyte, as a professional phagocyte is relatively straightforward and owes its origins to Metchnikoff. When first described, DCs were obviously distinct, nonphagocytic by definition, and corresponded to the interdigitating cells in T cell areas in lymphoid organs and migratory cells in afferent lymph. We refer to these cells defined by Steinman and Cohn as ‘classical’ DCs, while recognising that even this strict definition may in some circumstances combine several cell types of different developmental origin. With time, the definition of classical DCs in the mouse expanded to include phagocytes, and was subdivided into numerous functional subsets, including crosspresenting, migratory, myeloid, lymphoid, tolerogenic, and inflammatory or TipDCs, that differ in expression of surface markers including CD11c, CD8, CD103, F4/80, and CD11b, and may also differ between tissues [5]. DCs are believed to have a unique ability to present antigen to naı¨ve T cells; this has in turn led to the implicit view that any cell with antigen-presenting cell (APC) activity is a DC by definition. The alternative view is that subsets of macrophages can also present antigen to naı¨ve T cells [6]. This remains a divisive issue in immunology. As noted by Randolph [7], the question effectively divides the community into the macrophage and DC believers, who do not communicate because they disagree on basic principles. The view that classical DCs are unique as APCs requires a definition of a classical DC. That in turn depends upon the identification of markers to distinguish clearly all classical DCs from all macrophages. As reviewed previously, it is not clear that any surface marker fulfills this criterion, partly because no one marker is perfectly correlated with any other [3,6,7]. Recent systematic analysis of large gene expression arrays by the ImmGen Consortium identified molecular markers that distinguish DCs from a
Opinion set of prototypical macrophages, and similarly identified macrophage-specific markers, notably the Fc receptor CD64 and the signalling molecule Mer tyrosine kinase receptor (MerTK) [8,9]. Re-analysis of the same data, and a large meta-analysis of many datasets in the public domain for both mice and humans, using a network clustering approach, did not support the identification of these definitive DC or macrophage markers [10–12]. Instead, that analysis divided the myeloid APCs into two major clusters based upon selective expression of genes associated with endocytosis and lysosomal degradation. On the basis of this analysis, rather than distinguishing DCs from macrophages, the authors suggested a dichotomy between phagocytic APCs (antigen-presenting macrophages) and nonphagocytic APCs (classical DC). The phagocytic APCs included cells derived from the growth of bone marrow in CSF2 [granulocyte–macrophage (GM)-CSF], which have been termed bone-marrow-derived DCs (BMDCs). Many of the transcription factors associated with classical DCs, such as helix–loop–helix (HLH) transcription factor inhibitor of DNA binding or differentiation 2 (Id2), zinc finger and BTB domain containing 46 (zbtb46), nuclear factor, interleukin 3 regulated (Nfil3), and basic leucine zipper transcription factor, ATF-like 3 (BATF3), may act primarily as repressors of phagocytic differentiation (reviewed in [13]), and have been found to be functionally redundant for generation of CD8+ DCs in bone marrow reconstitution experiments [14]. So, one might take the view that a classical DC is the default option in MPS differentiation when the phagocyte gene expression cluster is repressed. Zbtb46 has been suggested as a definitive classical DC marker in several studies (e.g., [15]), but it does not fall within a DC-specific gene expression cluster. The proposed functional divide between macrophages and DCs as APCs has been further compromised with the recent recognition that blood monocytes can enter tissues and capture and transport antigen to lymph nodes without acquiring characteristics of classical DCs [16]. A separate question from the functional distinction is whether classical DCs, if they can be strictly defined, constitute a distinct cell lineage. Macrophages and granulocytes are clearly both phagocytes, and share a committed GM progenitor, but are commonly regarded as separate lineages, even though they may be interconverted [17]. Indeed, a recent paper claimed that granulocytes can also differentiate into DCs (or perhaps more accurately, active APCs) [18]. Classical DCs certainly share growth factors and committed progenitors with monocytes and macrophages. The differentiation of precursors of both macrophages and classical DCs is regulated by stem cell factor (SCF/c-kit), macrophage CSF (CSF1), IL-34, GM-CSF (CSF2), granulocyte CSF (CSF3), IL-3, and fms-related tyrosine kinase ligand. Flt3 ligand administration to mice produces a selective increase in the numbers of classical DCs in lymphoid and peripheral tissues [19], whereas CSF1 can expand tissue macrophage populations [20]. As discussed below, antibodies against the receptors for these factors are commonly used to identify progenitors, but there is considerable redundancy in their actions and the factors interact in complex ways. Mutations in CSF1 or IL-34, or their shared receptor (csf1r, CD115) cause a
Trends in Immunology August 2014, Vol. 35, No. 8
substantial reduction in particular tissue macrophage populations, including bone-resorbing osteoclasts, microglia, and Langerhans cells [21–23]. A CSF2, or CSF2R knockout (KO) had a selective effect on lung macrophage populations, corresponding to the human disease, alveolar proteinosis [24–26]. CSF1 also contributes to lung macrophage populations, as evident from the impact of the double KO with CSF2 [24]. Mutations of SCF/c-kit or Flt3L had severe effects on haematopoiesis, reflecting expression of their receptors on multipotent progenitors. Mutation of Flt3L produces a selective loss of both lymphoid and nonlymphoid tissue classical DCs [27,28], whereas CSF2R deletion produces a loss in many tissues of a subset of classical DCs defined by the CD103 (Itgae) surface marker. There was less effect of CSF2R deletion on DCs that express CD11b [25]. In mouse marrow, a high proliferative potential macrophage colony-forming unit requires costimulation with CSF2 or IL-3 plus CSF1 [29]. There are likely to be other factors that substitute for CSF2 or IL-3, because mice that lack the common receptors for these two factors still produce macrophage and classical DC progenitors [25]. Similarly, although CSF1R is expressed at low levels on progenitors, whereupon it can influence fate decision [30], it is not essential for monocyte generation [31]. Taken together, the data do not provide a definitive view of the separation of macrophage and DC lineages based upon growth-factor dependence, but do suggest a broad distinction between cells that depend upon Flt3L and/or GM-CSF versus those that depend upon CSF1 (or IL-34). Accordingly, the growth factor receptors, c-kit (CD117), Flt3 (CD135) and Csf1r (CD115) have been used as markers to purify bone marrow progenitor cells [32]. The DC/ macrophage lineages were unified through the identification of a shared clonogenic progenitor, the macrophage and DC precursor (MDP) [33] (Figure 1A). These cells, purified based upon coexpression of Csf1r and Flt3, gave rise to classical DCs, plasmacytoid DCs, monocytes, and macrophages following adoptive transfer [34]. The concept of ontogenically distinct precursors of DCs and macrophages arose upon identification of a committed dendritic cell progenitor, the common dendritic cell progenitor (CDP), in the bone marrow, that did not give rise to monocytes upon intravenous adoptive transfer, but was proposed to derive from the MDP [35] (Figure 1A). A circulating precursor that seeded classical DCs and not macrophages/ monocytes, termed the pre-DC, was subsequently identified following a similar adoptive transfer strategy [5], and indeed, fate-mapping studies indicate that classical DCs are not derived from mature monocytes but from this phenotypically distinct pre-DC (Figure 1A). Jakubzick et al. [36] first used a lysM-cre dependent reporter strain to demonstrate that resident classical DCs in lymphoid tissues are unlikely to derive from monocytes. More recently, Schraml et al. [37] have reported that the C-type lectin domain family 9, member A (CLEC9A) is expressed in CDP and pre-DC but not MDP or several conventional DC populations. Subsequent fate-mapping of CLEC9A-expressing cells using Clec9a-cre mice showed progeny within classical lymphoid and nonlymphoid tissue DC subsets, but not in blood monocytes, plasmacytoid DCs, nor in 359
Opinion
Trends in Immunology August 2014, Vol. 35, No. 8
Bone marrow (A)
Blood
Current model
GMP
LMPP
HSC
CDP
preDC
preDC
DC
MoP
Mono
Mono
MΦ
MDP
Model proposed here
(B)
preDC HSC
Tissue
LMPP
GMP
CDP
preDC
DC
Mono
MΦ
MDP MoP
Mono
(C) Possible alternave Key:
Flt3 CSF1R
LMPP
preDC
preDC
DC
LMPP
Mono
Mono
MΦ
TRENDS in Immunology
Figure 1. Models of differentiation of circulating DC and macrophage precursors from pluripotent progenitors in the bone marrow. (A) The CDP and MoPs are the direct ancestors of the pre-DCs and monocytes (Mono) in the bone marrow, respectively, and share a common origin in the MDP. (B) MDP is derived from CDP and gives rise to both pre-DCs and MoPs. (C) Fate decision is imprinted much earlier during haematopoiesis in individual LMPP. Abbreviations: CDP, common DC progenitor; CSF1R, colony stimulating factor 1 receptor; DC, dendritic cell; Flt3, fms-related tyrosine kinase; GMP, granulocyte-macrophage progenitor; HSC, haematopoietic stem cell; LMPP, lymphoidprimed multipotent progenitors; MDP, macrophage and DC precursor; Mono, monocyte; MoP, monocyte progenitor; MF, macrophage; pre-DC, precursor dendritic cell.
inflammatory monocyte-derived CD11c+ cells (i.e., classically activated macrophages) or tissue-resident CD11c+ macrophages such as alveolar and lamina propria macrophages. However, these findings do not necessarily imply that classical DCs and macrophages have distinct progenitors prior to circulating monocytes and pre-DCs, because the likelihood of recombination would increase in pre-DCs with level and duration of expression of the cre recombinase. Exactly how the precursors of pre-DCs and monocytes in the bone marrow are related and regulated is unclear [34], because the markers used to separate the putative CDP from the MDP populations overlap. It probably involves differential regulation of the key receptors, Csf1r and Flt3. Although CSF1 can apparently instruct fate in purified haematopoietic stem cells (HSCs), suggesting these cells probably express low level of Csf1r [30], Csf1r mRNA is low in mixed populations of kit-positive progenitors [38,39]. Indeed, the transcriptional events that lead to Csf1r induction in these progenitors have been analysed in detail [39]. Conversely, Flt3 is highly expressed at the mRNA level in the same c-kit-positive progenitor populations, and downregulated in common myeloid progenitors (www.biogps.org; [8]). This observation was confirmed in both mice and humans in the recent large promoter-based gene expression analysis by the FANTOM5 consortium [38]. One alternative interpretation that remains compatible with the available 360
data is that the CDP is actually a precursor of the MDP in the bone marrow and its fate is determined by whether it encounters Flt3L before it encounters Csf1 (Figure 1B). When this cell either enters the blood, or is adoptively transferred, it would presumably traffic to the sites of highest expression of Flt3L in lymphoid organs and tissues [38] (www.biogps.org). If this is the case, one key event in fate divergence in the bone marrow is the induction of Csf1r, and formation of an intermediate cell (the MDP, probably a subset of common myeloid progenitors) that expresses both receptors. The subsequent loss of Flt3 would then give rise to a committed monocyte progenitor (MoP) described recently by Hettinger et al. [40]. Conversely, retention of Flt3 and stimulation with Flt3L would direct formation of circulating precursors of DC (pre-DCs) and mature classical Flt3expressing DCs [5,41]; some of which nevertheless retain expression of Csf1r [42]. An alternative to the hierarchical commitment view of myeloid haematopoiesis derives from cellular barcoding of individual purified lymphoid-primed multipotent progenitors (LMPPs) [43]; progenitors thought to occur before the divergence of lymphoid and myeloid potency [44]. Adoptive transfer of tagged LMPPs followed by quantitative analysis of the frequency of progeny in B cells, lymphoid-tissue DCs, and monocytes has indicated that fate decisions are already imprinted in some individual progenitors. Output patterns
Opinion
Trends in Immunology August 2014, Vol. 35, No. 8
from tagged sibling LMPPs generated in vitro showed the same fate when adoptively transferred into separate mice, thereby discounting stochastic effect or effects of microenvironment niches after cell transfer. The high frequency of progenitors that gave rise to monocytes or classical DCs alone argues against an obligate linear relation between the MDP and CDP (Figure 1C). The conclusion of this discussion does not produce a consensus. A definition of DCs based upon function (DC=APC) cannot exist side-by-side with definitions based upon markers, lineage, or growth-factor dependence that clearly divide APCs into multiple classes. We take the view that the large majority of DCs are part of the MPS and the term DC should be strictly reserved for cells demonstrably derived from an Flt3+ pre-DC. More specifically, a monocyte-derived cell is a macrophage by definition, whether or not it acquires APC activity. Differentiation and turnover of monocytes and their recruitment into tissues Another challenge to the MPS concept is evidence that monocytes are not the immediate precursors of tissue macrophages. This concept is discussed in detail below but clearly warrants an appraisal of our current understanding of monocyte function. Monocytes in the peripheral blood of mice can be divided into two subsets of roughly equivalent numbers based upon the expression of lymphocyte antigen 6C (Ly6C) and chemokine CC receptor (CCR)2 and chemokine CX3C receptor (CX3CR)1 [45,46]. These subpopulations represent a differentiation series controlled by CSF1 with Ly6C+ cells emerging from the bone marrow via a CCR2-dependent mechanism [47] and Bone marrow
Blood
subsequently differentiating into Ly6C cells [31,48,49] (Figure 2). Ly6C monocytes were originally considered the immediate precursors of resident macrophages [46], whereas Ly6C+ cells were recruited under conditions of inflammation via CCR2 ligation, but the alternative proposal is that Ly6C cells represent a functional endpoint [49], the blood macrophage [50] that among other functions, patrol and monitor vascular endothelium [51] (Figure 2). Ly6C+ monocytes have an estimated half-life of 18 h compared to 5–7 days for the Ly6C subset [48,49], therefore, the majority of Ly6C+ cells must be lost to other (functional) endpoints, and thus, are more likely to be the direct precursors of tissue macrophages described in the original MPS model (Figure 2). The generation of the Ly6C monocyte population requires the early response gene, NR4a1 (Nur77) [50], but deletion of this gene does not alter the numbers of macrophages within tissues [52], further arguing that the Ly6C monocyte is an endpoint. A refinement of the MPS concept would be to propose that monocytes replenish some, but not all, tissue macrophages in the adult. This again returns to the question of nomenclature. Many mouse nonlymphoid tissues, such as lung interstitium, dermis, heart, liver, and kidney, contain populations of CD11b+ MHCII+ CD11c+ cells with short (7– 15 days) half-lives, sometimes referred to as ‘DCs’ [53]. Subsets of these cells express F4/80, Fc receptor (CD64) and MerTK [8,16,54,55] markers that have been identified as macrophage associated [8,11]. In contrast to the CD64negative populations, in all peripheral tissues examined except the kidney, these subsets are not tagged with the Clec9A–cre fate-mapping system described above, strongly suggesting they do not derive from pre-DCs [37]. In the gut Tissues
DC preDC
preDC
MΦ
MoP
Ly6C+
Ly6C+
Short-lived classical DC Gut MΦ Dermal MΦ Liver MΦ Kidney MΦ Peritoneal MΦ
Short-lived non replicang MΦ (oen CD11c+) Undifferenated monocytes migrang to dLN
Ly6C+
Ly6C–
Ly6C–
Bone marrow Mφ
Lung MΦ Liver Kupffer cells Peritoneal MΦ Langerhans cells
Long-lived selfrenewing MΦ (oen F4/80High)
TRENDS in Immunology
Figure 2. Peripheral tissues contain multiple populations of ontogenically distinct myeloid cells: a model for their relation to circulating precursors. Pre-DCs and Ly6C+ monocytes derive from the bone marrow, and circulate via the blood into the tissues. Pre-DCs continually replenish all tissue DC populations, with the exception of plasmacytoid DCs that exit from the bone marrow fully formed [37]. A minor proportion of Ly6C+ blood monocytes differentiate into blood Ly6C monocytes that function as vasculature macrophages [49,50]. Remaining Ly6C+ blood monocytes enter tissues and become either macrophages, for example in the gut and dermis [16,55,61] or remain undifferentiated, acquire antigen, and migrate to the draining lymph nodes [16]. Some tissue macrophages, including those in the bone marrow, are mainly maintained by self-renewal and longevity [16,49,64,93,95], even during inflammation [88,92,95,100]. In several tissues both self-renewing and continually replenished macrophages co-exist suggesting divergent functions. Abbreviations: DC, dendritic cell; MoP, monocyte progenitor; MF, macrophage; pre-DC, precursor dendritic cell.
361
Opinion where CD11b+ MHCII+ CD11c+ cells are more numerous, and possess a vacuolar cytoplasm resembling that of macrophages [56], they are uniformly positive for a Csf1r–EGFP transgene, completely depleted using anti-CSF1R antibody treatment [31], and appear predominantly to be continually replaced from the Ly6C+ monocyte population during homeostasis (reviewed by [57]). There is really no consensus among different authors as to whether these are called macrophages or DCs [37,58,59]. However, Cerovic et al. [60] have clearly separated them from the less numerous classical DCs, and demonstrated the contrasting expression of Csf1r and Flt3. The strongest evidence for their monocyte origin is provided by adoptive transfer of control Ly6C+ monocytes into Ccr2 / mice. Although engraftment of donor monocytes into control recipient animals rarely results in detectable populations in the tissues [56], the low frequency of endogenous circulating monocytes in Ccr2 / animals provides engrafted cells with a competitive advantage, following which they can be clearly tracked differentiating into the CD11c+ MHCII+ CD11b+ gut population [55,61]. Furthermore, Ccr2 / mice have fewer macrophages in the gastrointestinal tract [61], and monocytes injected directly into the colonic lamina propria acquire the phenotype of these macrophages [62]. CD11c+ cells that are replenished independently of CCR2 and are CSF1R independent are detected in the intestinal and colonic lamina propria [55,59,61] and are likely equivalent to the preDC-derived classical DCs [37]. Monocytes may also replenish myeloid populations in the skin. Continued replacement of CD11c+ dermal populations by Ly6C+ monocytes is strongly supported by the elevated levels of non-host chimerism observed in both populations from Ccr2 / mice in long-term parabiotic union with Ccr2+/+ animals [16,46] and repopulation studies following lethal irradiation or macrophage depletion suggest monocytes can differentiate into both CD11c+ CD64+ dermal ‘macrophages’ and a population of CD11c+ CD64 ‘DCs’ [54]. Adoptively transferred Ly6C+ monocytes can also be traced to CD11b+ MHCII+ CD11c+ cells in the liver and kidney [53], and peritoneal cavity [63]. In the detailed study using a lysM–cre-driven conditional reporter gene [36], where classical DCs in the spleen and lymph nodes show little evidence of recombination, the CD11b+ ‘MHCII+ CD11c+ ‘DCs’ in lung, dermis and intestine all resembled monocytes in the extent of lysM–cre-mediated recombination. Many tissues therefore contain at least one population of short-lived phenotypically distinct APCs (macrophages or DCs) that are continually replenished from blood monocytes. These cells are most likely derived from the Ly6C+ monocyte subset, raising the possibility that the mechanisms used to replenish monocyte-derived macrophages during the steady state resemble those that occur during inflammation. Origins of monocytes and macrophages during development and their relation in homeostasis If we leave aside the question of whether myeloid APCs in some sites are labelled as macrophages or DCs, all tissues contain large populations of cells that all would agree are definitive macrophages (Figure 2) for which it has been suggested high levels of F4/80 provide a marker [64]. Many 362
Trends in Immunology August 2014, Vol. 35, No. 8
of these resident populations are believed to exist independently of monocytes and may even form a distinct embryonic lineage from the haematopoietic system. If these cells do not arise directly from monocytes, from where do they come? During embryonic development, phagocytic cells appear first in the yolk sac. They enter the head and disperse through the circulation, before the appearance of haematopoietic islands and the onset of definitive haematopoiesis in the liver (reviewed by [65]). Csf1r-positive macrophages that appear in the yolk sac independently of the haematopoietic blood islands do not appear to go through a monocyte-like stage [66], and are readily visualised with a csf1r–EGFP transgene [67]. Several recent papers have suggested that the yolk-sac-derived cells are the exclusive progenitors of microglia, the resident macrophages of the brain [68,69], and that they contribute significantly to macrophage populations resident in many adult tissues [64]. Consistent with their derivation outside the blood islands, Schultz et al. [64] found that yolk-sac-derived macrophages appeared to develop and colonise tissues normally in embryos lacking myb, a key transcription factor in haematopoietic stem cell renewal, but were absent in mice that lacked the macrophagespecific transcription factor, PU.1. This observation may be difficult to interpret, because myb is a repressor of Csf1r transcription in mice [70], and knockdown of myb alone, in a human myeloid line, is enough to drive Csf1r expression and monocytic differentiation [71]. So, the myb KO could potentially drive yolk sac macrophage differentiation to occupy embryonic macrophage niches, while at the same time blocking definitive haematopoiesis. Furthermore, an absence of infiltrating foetal monocytes in myb-deficient mice could lead to the build up of local growth factors that allows abnormal survival of yolk sac macrophages in the tissues (Box 1). Caution must also be taken when interpreting data generated from inbred mouse strains, because yolk-sacderived macrophages develop normally in PU.1 / outbred mice in contrast to their dependence on PU.1 on the C57Bl/6 background [66]. Indeed, comparative profiling of C57Bl/6 and BALB/c macrophages identifies many strain-specific differences, and confirms that the former lack expression of cathepsin E, a key enzyme in antigen processing [72]. Accordingly, the effect of the myb KO on C57Bl/6 background is not necessarily generalisable. Additional evidence for an exclusive yolk sac origin for microglia has been provided using inducible lineage tracers in inbred mice (in which tamoxifen is administered to the mother) and by fate-mapping foetal monocytes adoptively transferred in utero or post-birth [68,69]. This directly contrasts with the interpretation of earlier studies that monocytes infiltrate in the perinatal and postnatal period to remove dying neurons and to form microglia in the brain and retina [73,74]. The monocyte transfers performed by Hoeffel et al. were done well before the wave of neuronal cell death that occurs later in gestation, and it remains possible that infiltration by monocytes occurs at this later stage. The use of tamoxifen to induce cre-mediated recombination in lineage tracing experiments may also be compromised by the fact that CSF1 is itself induced by oestrogen (and presumably tamoxifen) [75], and macrophage/DC differentiation may be directly regulated by the oestrogen receptor [76]. A subsequent study characterized
Opinion
Trends in Immunology August 2014, Vol. 35, No. 8
Box 1. A model of MPS homeostasis based upon CSF1 availability The MPS has several levels of intrinsic homeostasis. CSF1R signalling is required for the maintenance of most tissue macrophages, and exogenous CSF1 can promote their proliferation. CSF1 is produced constitutively by most mesenchymal cells, as a membrane-bound form, a secreted glycoprotein, and a proteoglycan [105]. The cell surface and secreted proteoglycan forms appear to have unique roles in homeostasis [106], and induced proliferation during inflammation [107], implying that local, rather than systemic, CSF1 production is important in homeostasis. If CSF1 is produced by tissue stroma at a fairly constant rate, local levels will depend upon the rate of removal by the macrophages themselves, and CSF1 will rise locally in response to the death or emigration of individual cells (see Figure 3A,B in main text). In solid tissues, this would be predicted to lead to proliferation of cells that neighbour a recently vacated space, which would explain the tendency of each tissue macrophage to occupy a specific niche or territory, not overlapping with its neighbours [1]. Elevated proliferation of resident cells that occurs following the macrophage disappearance reaction during inflammation is also dependent upon CSF1 [95,98] and appears to reflect a homeostatic process as the population is restored to the original density [88,95] without elevation in local CSF1 production [98] (see Figure 3C in main text). Elevated CSF1 production can also drive increased proliferation of resident macrophages [20,103], resulting in higher macrophage density within the tissues (see Figure 3D in main text). However, elevated CSF1 production also leads to increased recruitment of monocytes [20,103] via macrophage production of
CCR2 ligands [103] (see Figure 3E in main text), suggesting elevated CSF1 output acts to fill the tissue macrophage niche by all means possible. CSF1 levels can also be increased to promote monocyte recruitment by blocking CSF1 degradation; the CSF1R is itself a target of proteolytic cleavage by tumour necrosis factor (TNF)-a converting enzyme (TACE, ADAM17) in stimulated macrophages [108] and human monocytes cultured with CSF2 and IL-4 [109]. IL-34 provides a second CSF1R ligand and is conserved through evolution from birds to mammals [110]. IL-34 is expressed by keratinocytes in the epidermis and neurons in the brain; both sites in which monocyte recruitment does not generally occur. IL-34 allows self-renewal of Langerhans cells and microglia [22], although it is unclear if it stimulates production of CCR2 ligands by macrophages in the same way as CSF1 [103]. Thus, IL-34 may be produced specifically in the epidermis and brain in order to avoid steady-state recruitment of monocytes. The cerebellum and brain stem are the only areas of the brain in which microglia appear independent of IL-34 [22], but the reasons are unknown. The T helper (Th)2 cytokine IL-4 can also stimulate resident macrophages to proliferate beyond homeostatic levels in order to elevate tissue density during nematode infection [20] (see Figure 3F in main text). IL-4 appears to stimulate proliferation in a CSF1-independent manner although the expanding cells remain dependent upon CSF1 for survival [20]. Unlike CSF1, IL-4 does not apparently also drive monocyte recruitment, which may be part of a process ensuring the absence of monocyte-derived macrophages in certain pathologies.
be seeded predominantly from foetal liver-derived monocytes late in embryonic development [69,79,80]. Indeed, a recent study [15] focused mainly on cardiac macrophage self-renewal, concluded on the basis of a Flt3–cre-dependent reporter, that the large majority of macrophages in adult tissues do not, in fact, derive from the yolk sac. Whatever their origin during development and in the postnatal period, in the adult mouse brain microglia appear to be maintained without significant input from blood monocytes ([81,82] and references therein). A foetal monocyte origin for some adult tissue macrophages clearly implies self-renewal of the population [83]. Genetically or experimentally monocytopenic mice and humans do not have a deficiency of tissue macrophages [84–87], so there must be a capacity for tissue macrophages to replace themselves without a monocyte precursor. This
a c-kit-expressing erythromyeloid progenitor that depended upon both PU.1 and interferon-regulatory factor (IRF)8 for subsequent differentiation to become microglia when explanted onto hippocampal slices [77]. However, because yolk-sac-derived macrophages do not derive from the blood islands and haematopoietic progenitors [66], this finding suggests that definitive progenitors can, at least under defined conditions, produce microglia. There is also evidence that progenitors of definitive haematopoiesis arise in the yolk sac [78]. For all these reasons, the argument that microglia derive solely from yolk-sac progenitors based upon lineage tracing is not yet fully compelling. In our view, it remains likely that they share origins with epidermal Langerhans cells and alveolar lung macrophages in adult murine tissues, which similar inducible lineage tracing and adoptive transfer experiments show to
Blood monocytes
CSF-1 per cell
(A)
(B)
(C)
(D)
(E) CCR
Recruitment
a 2 lig
nds
(F)
Proliferaon
Local CSF1 producon Steady-state density
Steady-state death/migraon
Mφ disappearance during inflammaon
Heightened CSF1 secreon
Excess CSF1 secreon
IL-4/IL-13 secreon TRENDS in Immunology
Figure 3. Model of how CSF1R controls tissue macrophage density. CSF1R signalling controls proliferation of tissue macrophages [20,98]. Constant secretion of CSF1 by tissue stroma and consumption by macrophages maintains population density just below the CSF1R signalling threshold required for proliferation (A). Steady state death/ migration (B) or disappearance during inflammation (C) increases available CSF1 without necessarily changing production of CSF1, thereby allowing proliferation to restore normal density. Elevated CSF1 secretion stimulates cells to proliferate to higher tissue density than normal (D), but can also stimulate monocyte recruitment via macrophage chemokine production (E). Recruitment likely requires a higher threshold of CSF1R signalling than proliferation but this remains to be established. Others factors, such as IL-4, can allow macrophages to proliferate independently of CSF1 thereby increasing macrophage numbers when CSF1 is limiting and without concurrent increase in monocyte recruitment (F). Abbreviations: CCR, chemokine CC receptor; CSF, colony-stimulating factor; IL, interleukin; MF, macrophage.
363
Opinion possibility was never excluded from the definition of the MPS [2]. Measurement of Ki67 expression, DNA content, or phosphorylated histone H3 has shown that Langerhans cells [80] and peritoneal macrophages [88] undergo an initial burst of proliferation during neonatal development. Similar snapshots of proliferation suggest the majority (
Homeostasis in the mononuclear phagocyte system Stephen J. Jenkins1 and David A. Hume2 1
University of Edinburgh, Centre for Inflammation Research, Edinburgh EH16 4TJ, UK University of Edinburgh, The Roslin Institute and Royal (Dick) School of Veterinary Studies, Easter Bush Campus, Midlothian, EH25 9RG, UK
2
The mononuclear phagocyte system (MPS) is a family of functionally related cells including bone marrow precursors, blood monocytes, and tissue macrophages. We review the evidence that macrophages and dendritic cells (DCs) are separate lineages and functional entities, and examine whether the traditional view that monocytes are the immediate precursors of tissue macrophages needs to be refined based upon evidence that macrophages can extensively self-renew and can be seeded from yolk sac/foetal liver progenitors with little input from monocytes thereafter. We review the role of the growth factor colony-stimulating factor (CSF)1, and present a model consistent with the concept of the MPS in which local proliferation and monocyte recruitment are connected to ensure macrophages occupy their welldefined niche in most tissues. Introduction The MPS was originally defined as a family of cells derived from a pluripotent progenitor in the bone marrow and includes committed bone marrow progenitors, blood monocytes, and tissue macrophages; the latter comprising around 10% of total cells in every organ in the body [1,2]. MPS cells share many features, notably their phagocytic activity, but are also extremely plastic in their patterns of gene expression, defying identification based upon surface markers [3]. The proliferation and differentiation of MPS cells is controlled by macrophage CSF1 and interleukin (IL)-34, acting through a common receptor, CSF1R [4]. The traditional view of the MPS is that the major proliferative compartment is within the progenitors in the marrow, whereas blood monocytes provide the immediate precursors to replace tissue macrophages. This view has been challenged by several recent studies that, in effect, divide the MPS into separate cell lineages arising at different stages of embryonic development. The relationship of DCs with the MPS continues to be controversial; overlap of function and marker expression with monocyte-derived cells has made these cells difficult to delineate from macrophages, particularly in nonlymphoid tissues. Recent advances have provided evidence of a Corresponding author: Hume, D.A. ([email protected]). Keywords: macrophage; dendritic cell; proliferation; hematopoiesis; colony-stimulating factor 1. 1471-4906/ ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.it.2014.06.006
358
Trends in Immunology, August 2014, Vol. 35, No. 8
distinct precursor for DCs that would allow their definition as a separate haematopoietic lineage. It is therefore timely to ask whether these findings necessitate a complete revision of the MPS concept. The complex relation between DCs and macrophages One challenge to the concept of the MPS was the identification of the antigen-presenting DCs. Should these cells be included in the MPS, and if not, how can they be distinguished from macrophages? Efforts to define DCs have been based on function, molecular markers, and dependence upon growth and transcription factors. Each of the criteria are discussed below. The definition of a macrophage, or a mononuclear phagocyte, as a professional phagocyte is relatively straightforward and owes its origins to Metchnikoff. When first described, DCs were obviously distinct, nonphagocytic by definition, and corresponded to the interdigitating cells in T cell areas in lymphoid organs and migratory cells in afferent lymph. We refer to these cells defined by Steinman and Cohn as ‘classical’ DCs, while recognising that even this strict definition may in some circumstances combine several cell types of different developmental origin. With time, the definition of classical DCs in the mouse expanded to include phagocytes, and was subdivided into numerous functional subsets, including crosspresenting, migratory, myeloid, lymphoid, tolerogenic, and inflammatory or TipDCs, that differ in expression of surface markers including CD11c, CD8, CD103, F4/80, and CD11b, and may also differ between tissues [5]. DCs are believed to have a unique ability to present antigen to naı¨ve T cells; this has in turn led to the implicit view that any cell with antigen-presenting cell (APC) activity is a DC by definition. The alternative view is that subsets of macrophages can also present antigen to naı¨ve T cells [6]. This remains a divisive issue in immunology. As noted by Randolph [7], the question effectively divides the community into the macrophage and DC believers, who do not communicate because they disagree on basic principles. The view that classical DCs are unique as APCs requires a definition of a classical DC. That in turn depends upon the identification of markers to distinguish clearly all classical DCs from all macrophages. As reviewed previously, it is not clear that any surface marker fulfills this criterion, partly because no one marker is perfectly correlated with any other [3,6,7]. Recent systematic analysis of large gene expression arrays by the ImmGen Consortium identified molecular markers that distinguish DCs from a
Opinion set of prototypical macrophages, and similarly identified macrophage-specific markers, notably the Fc receptor CD64 and the signalling molecule Mer tyrosine kinase receptor (MerTK) [8,9]. Re-analysis of the same data, and a large meta-analysis of many datasets in the public domain for both mice and humans, using a network clustering approach, did not support the identification of these definitive DC or macrophage markers [10–12]. Instead, that analysis divided the myeloid APCs into two major clusters based upon selective expression of genes associated with endocytosis and lysosomal degradation. On the basis of this analysis, rather than distinguishing DCs from macrophages, the authors suggested a dichotomy between phagocytic APCs (antigen-presenting macrophages) and nonphagocytic APCs (classical DC). The phagocytic APCs included cells derived from the growth of bone marrow in CSF2 [granulocyte–macrophage (GM)-CSF], which have been termed bone-marrow-derived DCs (BMDCs). Many of the transcription factors associated with classical DCs, such as helix–loop–helix (HLH) transcription factor inhibitor of DNA binding or differentiation 2 (Id2), zinc finger and BTB domain containing 46 (zbtb46), nuclear factor, interleukin 3 regulated (Nfil3), and basic leucine zipper transcription factor, ATF-like 3 (BATF3), may act primarily as repressors of phagocytic differentiation (reviewed in [13]), and have been found to be functionally redundant for generation of CD8+ DCs in bone marrow reconstitution experiments [14]. So, one might take the view that a classical DC is the default option in MPS differentiation when the phagocyte gene expression cluster is repressed. Zbtb46 has been suggested as a definitive classical DC marker in several studies (e.g., [15]), but it does not fall within a DC-specific gene expression cluster. The proposed functional divide between macrophages and DCs as APCs has been further compromised with the recent recognition that blood monocytes can enter tissues and capture and transport antigen to lymph nodes without acquiring characteristics of classical DCs [16]. A separate question from the functional distinction is whether classical DCs, if they can be strictly defined, constitute a distinct cell lineage. Macrophages and granulocytes are clearly both phagocytes, and share a committed GM progenitor, but are commonly regarded as separate lineages, even though they may be interconverted [17]. Indeed, a recent paper claimed that granulocytes can also differentiate into DCs (or perhaps more accurately, active APCs) [18]. Classical DCs certainly share growth factors and committed progenitors with monocytes and macrophages. The differentiation of precursors of both macrophages and classical DCs is regulated by stem cell factor (SCF/c-kit), macrophage CSF (CSF1), IL-34, GM-CSF (CSF2), granulocyte CSF (CSF3), IL-3, and fms-related tyrosine kinase ligand. Flt3 ligand administration to mice produces a selective increase in the numbers of classical DCs in lymphoid and peripheral tissues [19], whereas CSF1 can expand tissue macrophage populations [20]. As discussed below, antibodies against the receptors for these factors are commonly used to identify progenitors, but there is considerable redundancy in their actions and the factors interact in complex ways. Mutations in CSF1 or IL-34, or their shared receptor (csf1r, CD115) cause a
Trends in Immunology August 2014, Vol. 35, No. 8
substantial reduction in particular tissue macrophage populations, including bone-resorbing osteoclasts, microglia, and Langerhans cells [21–23]. A CSF2, or CSF2R knockout (KO) had a selective effect on lung macrophage populations, corresponding to the human disease, alveolar proteinosis [24–26]. CSF1 also contributes to lung macrophage populations, as evident from the impact of the double KO with CSF2 [24]. Mutations of SCF/c-kit or Flt3L had severe effects on haematopoiesis, reflecting expression of their receptors on multipotent progenitors. Mutation of Flt3L produces a selective loss of both lymphoid and nonlymphoid tissue classical DCs [27,28], whereas CSF2R deletion produces a loss in many tissues of a subset of classical DCs defined by the CD103 (Itgae) surface marker. There was less effect of CSF2R deletion on DCs that express CD11b [25]. In mouse marrow, a high proliferative potential macrophage colony-forming unit requires costimulation with CSF2 or IL-3 plus CSF1 [29]. There are likely to be other factors that substitute for CSF2 or IL-3, because mice that lack the common receptors for these two factors still produce macrophage and classical DC progenitors [25]. Similarly, although CSF1R is expressed at low levels on progenitors, whereupon it can influence fate decision [30], it is not essential for monocyte generation [31]. Taken together, the data do not provide a definitive view of the separation of macrophage and DC lineages based upon growth-factor dependence, but do suggest a broad distinction between cells that depend upon Flt3L and/or GM-CSF versus those that depend upon CSF1 (or IL-34). Accordingly, the growth factor receptors, c-kit (CD117), Flt3 (CD135) and Csf1r (CD115) have been used as markers to purify bone marrow progenitor cells [32]. The DC/ macrophage lineages were unified through the identification of a shared clonogenic progenitor, the macrophage and DC precursor (MDP) [33] (Figure 1A). These cells, purified based upon coexpression of Csf1r and Flt3, gave rise to classical DCs, plasmacytoid DCs, monocytes, and macrophages following adoptive transfer [34]. The concept of ontogenically distinct precursors of DCs and macrophages arose upon identification of a committed dendritic cell progenitor, the common dendritic cell progenitor (CDP), in the bone marrow, that did not give rise to monocytes upon intravenous adoptive transfer, but was proposed to derive from the MDP [35] (Figure 1A). A circulating precursor that seeded classical DCs and not macrophages/ monocytes, termed the pre-DC, was subsequently identified following a similar adoptive transfer strategy [5], and indeed, fate-mapping studies indicate that classical DCs are not derived from mature monocytes but from this phenotypically distinct pre-DC (Figure 1A). Jakubzick et al. [36] first used a lysM-cre dependent reporter strain to demonstrate that resident classical DCs in lymphoid tissues are unlikely to derive from monocytes. More recently, Schraml et al. [37] have reported that the C-type lectin domain family 9, member A (CLEC9A) is expressed in CDP and pre-DC but not MDP or several conventional DC populations. Subsequent fate-mapping of CLEC9A-expressing cells using Clec9a-cre mice showed progeny within classical lymphoid and nonlymphoid tissue DC subsets, but not in blood monocytes, plasmacytoid DCs, nor in 359
Opinion
Trends in Immunology August 2014, Vol. 35, No. 8
Bone marrow (A)
Blood
Current model
GMP
LMPP
HSC
CDP
preDC
preDC
DC
MoP
Mono
Mono
MΦ
MDP
Model proposed here
(B)
preDC HSC
Tissue
LMPP
GMP
CDP
preDC
DC
Mono
MΦ
MDP MoP
Mono
(C) Possible alternave Key:
Flt3 CSF1R
LMPP
preDC
preDC
DC
LMPP
Mono
Mono
MΦ
TRENDS in Immunology
Figure 1. Models of differentiation of circulating DC and macrophage precursors from pluripotent progenitors in the bone marrow. (A) The CDP and MoPs are the direct ancestors of the pre-DCs and monocytes (Mono) in the bone marrow, respectively, and share a common origin in the MDP. (B) MDP is derived from CDP and gives rise to both pre-DCs and MoPs. (C) Fate decision is imprinted much earlier during haematopoiesis in individual LMPP. Abbreviations: CDP, common DC progenitor; CSF1R, colony stimulating factor 1 receptor; DC, dendritic cell; Flt3, fms-related tyrosine kinase; GMP, granulocyte-macrophage progenitor; HSC, haematopoietic stem cell; LMPP, lymphoidprimed multipotent progenitors; MDP, macrophage and DC precursor; Mono, monocyte; MoP, monocyte progenitor; MF, macrophage; pre-DC, precursor dendritic cell.
inflammatory monocyte-derived CD11c+ cells (i.e., classically activated macrophages) or tissue-resident CD11c+ macrophages such as alveolar and lamina propria macrophages. However, these findings do not necessarily imply that classical DCs and macrophages have distinct progenitors prior to circulating monocytes and pre-DCs, because the likelihood of recombination would increase in pre-DCs with level and duration of expression of the cre recombinase. Exactly how the precursors of pre-DCs and monocytes in the bone marrow are related and regulated is unclear [34], because the markers used to separate the putative CDP from the MDP populations overlap. It probably involves differential regulation of the key receptors, Csf1r and Flt3. Although CSF1 can apparently instruct fate in purified haematopoietic stem cells (HSCs), suggesting these cells probably express low level of Csf1r [30], Csf1r mRNA is low in mixed populations of kit-positive progenitors [38,39]. Indeed, the transcriptional events that lead to Csf1r induction in these progenitors have been analysed in detail [39]. Conversely, Flt3 is highly expressed at the mRNA level in the same c-kit-positive progenitor populations, and downregulated in common myeloid progenitors (www.biogps.org; [8]). This observation was confirmed in both mice and humans in the recent large promoter-based gene expression analysis by the FANTOM5 consortium [38]. One alternative interpretation that remains compatible with the available 360
data is that the CDP is actually a precursor of the MDP in the bone marrow and its fate is determined by whether it encounters Flt3L before it encounters Csf1 (Figure 1B). When this cell either enters the blood, or is adoptively transferred, it would presumably traffic to the sites of highest expression of Flt3L in lymphoid organs and tissues [38] (www.biogps.org). If this is the case, one key event in fate divergence in the bone marrow is the induction of Csf1r, and formation of an intermediate cell (the MDP, probably a subset of common myeloid progenitors) that expresses both receptors. The subsequent loss of Flt3 would then give rise to a committed monocyte progenitor (MoP) described recently by Hettinger et al. [40]. Conversely, retention of Flt3 and stimulation with Flt3L would direct formation of circulating precursors of DC (pre-DCs) and mature classical Flt3expressing DCs [5,41]; some of which nevertheless retain expression of Csf1r [42]. An alternative to the hierarchical commitment view of myeloid haematopoiesis derives from cellular barcoding of individual purified lymphoid-primed multipotent progenitors (LMPPs) [43]; progenitors thought to occur before the divergence of lymphoid and myeloid potency [44]. Adoptive transfer of tagged LMPPs followed by quantitative analysis of the frequency of progeny in B cells, lymphoid-tissue DCs, and monocytes has indicated that fate decisions are already imprinted in some individual progenitors. Output patterns
Opinion
Trends in Immunology August 2014, Vol. 35, No. 8
from tagged sibling LMPPs generated in vitro showed the same fate when adoptively transferred into separate mice, thereby discounting stochastic effect or effects of microenvironment niches after cell transfer. The high frequency of progenitors that gave rise to monocytes or classical DCs alone argues against an obligate linear relation between the MDP and CDP (Figure 1C). The conclusion of this discussion does not produce a consensus. A definition of DCs based upon function (DC=APC) cannot exist side-by-side with definitions based upon markers, lineage, or growth-factor dependence that clearly divide APCs into multiple classes. We take the view that the large majority of DCs are part of the MPS and the term DC should be strictly reserved for cells demonstrably derived from an Flt3+ pre-DC. More specifically, a monocyte-derived cell is a macrophage by definition, whether or not it acquires APC activity. Differentiation and turnover of monocytes and their recruitment into tissues Another challenge to the MPS concept is evidence that monocytes are not the immediate precursors of tissue macrophages. This concept is discussed in detail below but clearly warrants an appraisal of our current understanding of monocyte function. Monocytes in the peripheral blood of mice can be divided into two subsets of roughly equivalent numbers based upon the expression of lymphocyte antigen 6C (Ly6C) and chemokine CC receptor (CCR)2 and chemokine CX3C receptor (CX3CR)1 [45,46]. These subpopulations represent a differentiation series controlled by CSF1 with Ly6C+ cells emerging from the bone marrow via a CCR2-dependent mechanism [47] and Bone marrow
Blood
subsequently differentiating into Ly6C cells [31,48,49] (Figure 2). Ly6C monocytes were originally considered the immediate precursors of resident macrophages [46], whereas Ly6C+ cells were recruited under conditions of inflammation via CCR2 ligation, but the alternative proposal is that Ly6C cells represent a functional endpoint [49], the blood macrophage [50] that among other functions, patrol and monitor vascular endothelium [51] (Figure 2). Ly6C+ monocytes have an estimated half-life of 18 h compared to 5–7 days for the Ly6C subset [48,49], therefore, the majority of Ly6C+ cells must be lost to other (functional) endpoints, and thus, are more likely to be the direct precursors of tissue macrophages described in the original MPS model (Figure 2). The generation of the Ly6C monocyte population requires the early response gene, NR4a1 (Nur77) [50], but deletion of this gene does not alter the numbers of macrophages within tissues [52], further arguing that the Ly6C monocyte is an endpoint. A refinement of the MPS concept would be to propose that monocytes replenish some, but not all, tissue macrophages in the adult. This again returns to the question of nomenclature. Many mouse nonlymphoid tissues, such as lung interstitium, dermis, heart, liver, and kidney, contain populations of CD11b+ MHCII+ CD11c+ cells with short (7– 15 days) half-lives, sometimes referred to as ‘DCs’ [53]. Subsets of these cells express F4/80, Fc receptor (CD64) and MerTK [8,16,54,55] markers that have been identified as macrophage associated [8,11]. In contrast to the CD64negative populations, in all peripheral tissues examined except the kidney, these subsets are not tagged with the Clec9A–cre fate-mapping system described above, strongly suggesting they do not derive from pre-DCs [37]. In the gut Tissues
DC preDC
preDC
MΦ
MoP
Ly6C+
Ly6C+
Short-lived classical DC Gut MΦ Dermal MΦ Liver MΦ Kidney MΦ Peritoneal MΦ
Short-lived non replicang MΦ (oen CD11c+) Undifferenated monocytes migrang to dLN
Ly6C+
Ly6C–
Ly6C–
Bone marrow Mφ
Lung MΦ Liver Kupffer cells Peritoneal MΦ Langerhans cells
Long-lived selfrenewing MΦ (oen F4/80High)
TRENDS in Immunology
Figure 2. Peripheral tissues contain multiple populations of ontogenically distinct myeloid cells: a model for their relation to circulating precursors. Pre-DCs and Ly6C+ monocytes derive from the bone marrow, and circulate via the blood into the tissues. Pre-DCs continually replenish all tissue DC populations, with the exception of plasmacytoid DCs that exit from the bone marrow fully formed [37]. A minor proportion of Ly6C+ blood monocytes differentiate into blood Ly6C monocytes that function as vasculature macrophages [49,50]. Remaining Ly6C+ blood monocytes enter tissues and become either macrophages, for example in the gut and dermis [16,55,61] or remain undifferentiated, acquire antigen, and migrate to the draining lymph nodes [16]. Some tissue macrophages, including those in the bone marrow, are mainly maintained by self-renewal and longevity [16,49,64,93,95], even during inflammation [88,92,95,100]. In several tissues both self-renewing and continually replenished macrophages co-exist suggesting divergent functions. Abbreviations: DC, dendritic cell; MoP, monocyte progenitor; MF, macrophage; pre-DC, precursor dendritic cell.
361
Opinion where CD11b+ MHCII+ CD11c+ cells are more numerous, and possess a vacuolar cytoplasm resembling that of macrophages [56], they are uniformly positive for a Csf1r–EGFP transgene, completely depleted using anti-CSF1R antibody treatment [31], and appear predominantly to be continually replaced from the Ly6C+ monocyte population during homeostasis (reviewed by [57]). There is really no consensus among different authors as to whether these are called macrophages or DCs [37,58,59]. However, Cerovic et al. [60] have clearly separated them from the less numerous classical DCs, and demonstrated the contrasting expression of Csf1r and Flt3. The strongest evidence for their monocyte origin is provided by adoptive transfer of control Ly6C+ monocytes into Ccr2 / mice. Although engraftment of donor monocytes into control recipient animals rarely results in detectable populations in the tissues [56], the low frequency of endogenous circulating monocytes in Ccr2 / animals provides engrafted cells with a competitive advantage, following which they can be clearly tracked differentiating into the CD11c+ MHCII+ CD11b+ gut population [55,61]. Furthermore, Ccr2 / mice have fewer macrophages in the gastrointestinal tract [61], and monocytes injected directly into the colonic lamina propria acquire the phenotype of these macrophages [62]. CD11c+ cells that are replenished independently of CCR2 and are CSF1R independent are detected in the intestinal and colonic lamina propria [55,59,61] and are likely equivalent to the preDC-derived classical DCs [37]. Monocytes may also replenish myeloid populations in the skin. Continued replacement of CD11c+ dermal populations by Ly6C+ monocytes is strongly supported by the elevated levels of non-host chimerism observed in both populations from Ccr2 / mice in long-term parabiotic union with Ccr2+/+ animals [16,46] and repopulation studies following lethal irradiation or macrophage depletion suggest monocytes can differentiate into both CD11c+ CD64+ dermal ‘macrophages’ and a population of CD11c+ CD64 ‘DCs’ [54]. Adoptively transferred Ly6C+ monocytes can also be traced to CD11b+ MHCII+ CD11c+ cells in the liver and kidney [53], and peritoneal cavity [63]. In the detailed study using a lysM–cre-driven conditional reporter gene [36], where classical DCs in the spleen and lymph nodes show little evidence of recombination, the CD11b+ ‘MHCII+ CD11c+ ‘DCs’ in lung, dermis and intestine all resembled monocytes in the extent of lysM–cre-mediated recombination. Many tissues therefore contain at least one population of short-lived phenotypically distinct APCs (macrophages or DCs) that are continually replenished from blood monocytes. These cells are most likely derived from the Ly6C+ monocyte subset, raising the possibility that the mechanisms used to replenish monocyte-derived macrophages during the steady state resemble those that occur during inflammation. Origins of monocytes and macrophages during development and their relation in homeostasis If we leave aside the question of whether myeloid APCs in some sites are labelled as macrophages or DCs, all tissues contain large populations of cells that all would agree are definitive macrophages (Figure 2) for which it has been suggested high levels of F4/80 provide a marker [64]. Many 362
Trends in Immunology August 2014, Vol. 35, No. 8
of these resident populations are believed to exist independently of monocytes and may even form a distinct embryonic lineage from the haematopoietic system. If these cells do not arise directly from monocytes, from where do they come? During embryonic development, phagocytic cells appear first in the yolk sac. They enter the head and disperse through the circulation, before the appearance of haematopoietic islands and the onset of definitive haematopoiesis in the liver (reviewed by [65]). Csf1r-positive macrophages that appear in the yolk sac independently of the haematopoietic blood islands do not appear to go through a monocyte-like stage [66], and are readily visualised with a csf1r–EGFP transgene [67]. Several recent papers have suggested that the yolk-sac-derived cells are the exclusive progenitors of microglia, the resident macrophages of the brain [68,69], and that they contribute significantly to macrophage populations resident in many adult tissues [64]. Consistent with their derivation outside the blood islands, Schultz et al. [64] found that yolk-sac-derived macrophages appeared to develop and colonise tissues normally in embryos lacking myb, a key transcription factor in haematopoietic stem cell renewal, but were absent in mice that lacked the macrophagespecific transcription factor, PU.1. This observation may be difficult to interpret, because myb is a repressor of Csf1r transcription in mice [70], and knockdown of myb alone, in a human myeloid line, is enough to drive Csf1r expression and monocytic differentiation [71]. So, the myb KO could potentially drive yolk sac macrophage differentiation to occupy embryonic macrophage niches, while at the same time blocking definitive haematopoiesis. Furthermore, an absence of infiltrating foetal monocytes in myb-deficient mice could lead to the build up of local growth factors that allows abnormal survival of yolk sac macrophages in the tissues (Box 1). Caution must also be taken when interpreting data generated from inbred mouse strains, because yolk-sacderived macrophages develop normally in PU.1 / outbred mice in contrast to their dependence on PU.1 on the C57Bl/6 background [66]. Indeed, comparative profiling of C57Bl/6 and BALB/c macrophages identifies many strain-specific differences, and confirms that the former lack expression of cathepsin E, a key enzyme in antigen processing [72]. Accordingly, the effect of the myb KO on C57Bl/6 background is not necessarily generalisable. Additional evidence for an exclusive yolk sac origin for microglia has been provided using inducible lineage tracers in inbred mice (in which tamoxifen is administered to the mother) and by fate-mapping foetal monocytes adoptively transferred in utero or post-birth [68,69]. This directly contrasts with the interpretation of earlier studies that monocytes infiltrate in the perinatal and postnatal period to remove dying neurons and to form microglia in the brain and retina [73,74]. The monocyte transfers performed by Hoeffel et al. were done well before the wave of neuronal cell death that occurs later in gestation, and it remains possible that infiltration by monocytes occurs at this later stage. The use of tamoxifen to induce cre-mediated recombination in lineage tracing experiments may also be compromised by the fact that CSF1 is itself induced by oestrogen (and presumably tamoxifen) [75], and macrophage/DC differentiation may be directly regulated by the oestrogen receptor [76]. A subsequent study characterized
Opinion
Trends in Immunology August 2014, Vol. 35, No. 8
Box 1. A model of MPS homeostasis based upon CSF1 availability The MPS has several levels of intrinsic homeostasis. CSF1R signalling is required for the maintenance of most tissue macrophages, and exogenous CSF1 can promote their proliferation. CSF1 is produced constitutively by most mesenchymal cells, as a membrane-bound form, a secreted glycoprotein, and a proteoglycan [105]. The cell surface and secreted proteoglycan forms appear to have unique roles in homeostasis [106], and induced proliferation during inflammation [107], implying that local, rather than systemic, CSF1 production is important in homeostasis. If CSF1 is produced by tissue stroma at a fairly constant rate, local levels will depend upon the rate of removal by the macrophages themselves, and CSF1 will rise locally in response to the death or emigration of individual cells (see Figure 3A,B in main text). In solid tissues, this would be predicted to lead to proliferation of cells that neighbour a recently vacated space, which would explain the tendency of each tissue macrophage to occupy a specific niche or territory, not overlapping with its neighbours [1]. Elevated proliferation of resident cells that occurs following the macrophage disappearance reaction during inflammation is also dependent upon CSF1 [95,98] and appears to reflect a homeostatic process as the population is restored to the original density [88,95] without elevation in local CSF1 production [98] (see Figure 3C in main text). Elevated CSF1 production can also drive increased proliferation of resident macrophages [20,103], resulting in higher macrophage density within the tissues (see Figure 3D in main text). However, elevated CSF1 production also leads to increased recruitment of monocytes [20,103] via macrophage production of
CCR2 ligands [103] (see Figure 3E in main text), suggesting elevated CSF1 output acts to fill the tissue macrophage niche by all means possible. CSF1 levels can also be increased to promote monocyte recruitment by blocking CSF1 degradation; the CSF1R is itself a target of proteolytic cleavage by tumour necrosis factor (TNF)-a converting enzyme (TACE, ADAM17) in stimulated macrophages [108] and human monocytes cultured with CSF2 and IL-4 [109]. IL-34 provides a second CSF1R ligand and is conserved through evolution from birds to mammals [110]. IL-34 is expressed by keratinocytes in the epidermis and neurons in the brain; both sites in which monocyte recruitment does not generally occur. IL-34 allows self-renewal of Langerhans cells and microglia [22], although it is unclear if it stimulates production of CCR2 ligands by macrophages in the same way as CSF1 [103]. Thus, IL-34 may be produced specifically in the epidermis and brain in order to avoid steady-state recruitment of monocytes. The cerebellum and brain stem are the only areas of the brain in which microglia appear independent of IL-34 [22], but the reasons are unknown. The T helper (Th)2 cytokine IL-4 can also stimulate resident macrophages to proliferate beyond homeostatic levels in order to elevate tissue density during nematode infection [20] (see Figure 3F in main text). IL-4 appears to stimulate proliferation in a CSF1-independent manner although the expanding cells remain dependent upon CSF1 for survival [20]. Unlike CSF1, IL-4 does not apparently also drive monocyte recruitment, which may be part of a process ensuring the absence of monocyte-derived macrophages in certain pathologies.
be seeded predominantly from foetal liver-derived monocytes late in embryonic development [69,79,80]. Indeed, a recent study [15] focused mainly on cardiac macrophage self-renewal, concluded on the basis of a Flt3–cre-dependent reporter, that the large majority of macrophages in adult tissues do not, in fact, derive from the yolk sac. Whatever their origin during development and in the postnatal period, in the adult mouse brain microglia appear to be maintained without significant input from blood monocytes ([81,82] and references therein). A foetal monocyte origin for some adult tissue macrophages clearly implies self-renewal of the population [83]. Genetically or experimentally monocytopenic mice and humans do not have a deficiency of tissue macrophages [84–87], so there must be a capacity for tissue macrophages to replace themselves without a monocyte precursor. This
a c-kit-expressing erythromyeloid progenitor that depended upon both PU.1 and interferon-regulatory factor (IRF)8 for subsequent differentiation to become microglia when explanted onto hippocampal slices [77]. However, because yolk-sac-derived macrophages do not derive from the blood islands and haematopoietic progenitors [66], this finding suggests that definitive progenitors can, at least under defined conditions, produce microglia. There is also evidence that progenitors of definitive haematopoiesis arise in the yolk sac [78]. For all these reasons, the argument that microglia derive solely from yolk-sac progenitors based upon lineage tracing is not yet fully compelling. In our view, it remains likely that they share origins with epidermal Langerhans cells and alveolar lung macrophages in adult murine tissues, which similar inducible lineage tracing and adoptive transfer experiments show to
Blood monocytes
CSF-1 per cell
(A)
(B)
(C)
(D)
(E) CCR
Recruitment
a 2 lig
nds
(F)
Proliferaon
Local CSF1 producon Steady-state density
Steady-state death/migraon
Mφ disappearance during inflammaon
Heightened CSF1 secreon
Excess CSF1 secreon
IL-4/IL-13 secreon TRENDS in Immunology
Figure 3. Model of how CSF1R controls tissue macrophage density. CSF1R signalling controls proliferation of tissue macrophages [20,98]. Constant secretion of CSF1 by tissue stroma and consumption by macrophages maintains population density just below the CSF1R signalling threshold required for proliferation (A). Steady state death/ migration (B) or disappearance during inflammation (C) increases available CSF1 without necessarily changing production of CSF1, thereby allowing proliferation to restore normal density. Elevated CSF1 secretion stimulates cells to proliferate to higher tissue density than normal (D), but can also stimulate monocyte recruitment via macrophage chemokine production (E). Recruitment likely requires a higher threshold of CSF1R signalling than proliferation but this remains to be established. Others factors, such as IL-4, can allow macrophages to proliferate independently of CSF1 thereby increasing macrophage numbers when CSF1 is limiting and without concurrent increase in monocyte recruitment (F). Abbreviations: CCR, chemokine CC receptor; CSF, colony-stimulating factor; IL, interleukin; MF, macrophage.
363
Opinion possibility was never excluded from the definition of the MPS [2]. Measurement of Ki67 expression, DNA content, or phosphorylated histone H3 has shown that Langerhans cells [80] and peritoneal macrophages [88] undergo an initial burst of proliferation during neonatal development. Similar snapshots of proliferation suggest the majority (
