Macrophage heterogeneity in tissues: phenotypic diversity and functions

Macrophages are remarkably versatile in their ability to recognize and respond to a wide range of stimuli, expressing a variety of surface and intracellular receptors, multiple signal transduction pathways and complex, adaptable arrays of gene expression. They are long-lived, biosynthetically active cells with potent endocytic, phagocytic, and secretory functions, able to modulate their properties upon contact with different cell types as well as extracellular matrix. Their intrinsic heterogeneity during differentiation is compounded by reciprocal interactions with neighboring cells, including macrophages themselves, diverse microorganisms, sterile particulates and soluble mediators. We summarize general features of the mononuclear phagocyte system, before dealing with diversity of macrophages in specific locations. We emphasize early studies with membrane antigen markers, mainly in the mouse, to provide a foundation for contemporary analysis of global gene expression analysis, in both human and mouse.

Although we have learned a great deal about macrophages, our understanding is still fragmentary, and an overall view of tissue-specific functions, selective markers, or even accurate numbers of macrophage subpopulations in health and disease is lacking, especially in humans. Examples and fragmentary evidence are becoming available, but tools need to be applied in a consistent manner to gather a general perspective of the macrophage system and its functional heterogeneity. Genomics, transcriptomics and proteomics have advanced our understanding of the complexity of macrophage gene signatures and their phenotypic modulation during physiologic and immune responses. In vitro and ex vivo studies have shown that macrophages respond with consistent gene programs that behave in a modular fashion 154 – 156 ; these modules can be regulated by many extrinsic tissue environmental factors including cytokines, cell to cell contact, and the extracellular matrix, but also usefully as well as detrimentally by drugs. Although at single gene level there is little specificity, this can be achieved increasingly by studying gene signatures as a whole. Importantly, these signatures are not fixed, but change in response to the same stimulus over time, as well as tissue localization. As we aim to understand the genome-transcriptome-proteome-interactome-phenome relationship for macrophages in every tissue and in different diseases, integration of all these analyses poses a formidable challenge. Detailed studies of tissue macrophage gene expression and their heterogeneity within and among many tissues are still limited, especially in humans. Gautier, Randolph et al . reported in 2012 157 , as part of the Immgen project, on the identity and differences in mouse tissue macrophages; four populations were assessed as a whole, peritoneal, red pulp splenic, lung macrophages, and microglia. Recently, Okabe and Medzhitov 97 assessed macrophages from peritoneal cavity, lung, liver, spleen, intestine, and adipose tissue. Candidate tissue-specific overexpressed markers from these studies included F5, FN1, and CD102 for peritoneal macrophages, Ear2, Ear3, Chi3L4, and chi3l3 in alveolar macrophages, CD207, GPR182, CD26, and clec4g for liver macrophages, CD121b, MMP2, and MMP13 for gut macrophages and CD82, SPIC, and CCR3 for spleen. When comparing the gene lists for specific tissue macrophages reported by both studies, e.g. peritoneal macrophages or spleen macrophages, we find similarities and differences that need further investigation. This is almost a hallmark of microarray studies so far; isolation procedures, array platforms, inter-individual, strain and housing differences, among other factors, are all able to affect gene signatures. In humans, similar variability has been shown for monocytes, for example 61 . Proteomics and phenomics assessment have not usually been carried out at this level of analysis. An important ultimate goal is to understand the modulation and interpretation of macrophage signatures in a broad range of disease states. Alveolar macrophages provide one of the best-studied populations in both species; we focus in the following paragraph on lung because findings in this system carry lessons that can be applied to macrophages in other locations. The lung is an exceptional organ for phenomic studies because of the highly developed tools to assess organ performance. In lung, Shaykhiev, Crystal et al . 158 defined gene expression signatures in smokers versus non-smokers and chronic obstructive pulmonary disease (COPD) patients to detect a large number of expressed mRNAs and signs of repression of interferon signatures and exacerbation of trophic signatures in smoking and more COPD patients. This study shows a relation between disease cause and consequence based on gene signatures. Furthermore, it reveals that the genes that we find to be important for activation in vitro appear as modules in vivo , but with unexpected tendencies; for example macrophages from healthy individuals have basally more Interferon genes expressed than smokers or COPD patients. It does not necessarily mean that alveolar macrophages are constitutively activated. Interpretation of in vitro findings in relation to ex vivo data can be difficult, because strong markers of IL-4 and steroid signaling such as MRC1 and CD163 are almost ubiquitously expressed in all tissue macrophages, requiring further investigation (FME, personal observation). An elegant demonstration of how the cellular environment affects the alveolar macrophage response was published by Snelgrove, Hussell et al . 159 . They showed that airway macrophages have high expression of the negative regulator CD200 receptor (CD200R) and their inflammatory potential was restrained by CD200 expressed on airway epithelium. Mice lacking CD200 had more macrophage activity and enhanced sensitivity to influenza infection and consequent mortality. How specific cellular, matrix, and soluble signals influence macrophage responses need to be investigated in detail. Infection is also a strong modulator of tissue macrophage activation and development of specialized phenotypes. An interesting study addressed inflammatory signatures in laser micro-dissected macrophages in caseous human pulmonary tuberculosis (TB) granulomas 160 . Microarray and histology confirmed overrepresentation of lipid sequestration and metabolism signatures in granuloma macrophages. Phenotypically, this could be relevant to pathogenesis since the caseum has an abundance of cholesterol ester, cholesterol, triacylglycerol, and lactosylceramide, known to influence both cholesterol sequestration and cell death. The study also shows that not only does the infection provide a strong stimulus, but that the tissue environment shapes macrophage gene expression significantly. Genomic studies of host responses to tuberculosis by Thuong, Hawn et al . 161 , show that macrophages from individuals with different clinical manifestations of Mycobacterium tuberculosis have distinct gene expression profiles and polymorphisms linked to susceptibility. The authors identified CCL1 as biomarker to distinguish the clinical groups and carriage of 6 single nucleotide polymorphisms correlated with infection in a case–control genetic association study with different clinical outcomes 161 . Other transcriptome studies dedicated to microglia, placental and Hofbauer macrophages, adipose tissue and Kupffer cells are summarized in Table  2 . Selection of whole genome studies in tissue macrophages Macrophage gene expression at whole genome level has been carried out for different tissues and species. However, given the heterogeneity in platforms and isolation methods it is difficult to compare results. Nevertheless, here we wanted to point to interesting studies available in Geo Dataset, which give an idea of the diversity and responsiveness of tissue macrophages. Although high throughput techniques have begun to shed light on different aspects of macrophage heterogeneity, further integration of gene expression in all its stages to function is indispensable. In particular, proteomics, lipidomics, and phenomic in situ analysis need to be combined with microarrays to clarify the role of genes, mutations, expression levels, splicing, expression and ultimately gene function in these cells.

Most of our knowledge regarding tissue macrophages is derived from mouse studies 1 , 2 . Mouse tissue macrophages originate from precursors in hematopoietic organs, e.g. yolk sac and fetal liver in the embryo, and bone marrow and other hematopoietic sites such as spleen, postnatally 3 – 6 . The nature of the cellular receptors and chemotactic signals that guide their constitutive migration within tissues remains obscure. Within tissues, macrophages have been traditionally described as sessile, or ‘fixed’; however, their further migration can be enhanced by inflammatory stimuli, especially chemokines generated locally by humoral and cellular mechanisms. In response to inflammation and infection, bone marrow-derived blood-borne monocytes give rise to cells that remain within the circulation or enter tissues, giving rise to both macrophages and dendritic cells (DCs) 7 ; ‘elicited’ macrophages often show enhanced turnover at local sites 8 . Recruited monocytes and sinus-lining tissue macrophages interact with blood, lymph, and extracellular fluid, as well as every cell type, including lympho-hematopoietic cells, endothelium, epithelium, mesenchymal, and neuronal cells. They express receptors and ligands for the products of resting and activated immune and non-immune cells, interacting through cell–cell contact, secretion, and endocytosis. The resident and newly recruited populations of macrophages at particular locations result from a dynamic process of production and entry, local proliferation, senescence, death, and emigration; recent studies have rediscovered the importance of local proliferation and self-renewal, dependent on growth factors such as M-CSF, GM-CSF and cytokines including IL-4/IL-13 and IL-6 9 , 10 . The macrophage phenotypic changes that accompany residence in different tissues are still poorly understood. The effects of heterogeneous macrophage populations on their local tissue environment include trophic as well as cytocidal interactions with neighboring cells, the remodeling of matrix, as well as phagocytic clearance of dying cells and other homeostatic and defense functions. These involve non-phagocytic intercellular adhesion, release and capture of secretory products, microvesicles and exosomes, nanotube formation, and cell fusion. Macrophage-derived secretory molecules include enzymes such as lysozyme, neutral proteinases, pro-and anti-inflammatory cytokines, chemokines and growth factors, arachidonate, oxygen, and nitrogen-derived metabolites. Macrophages are metabolically active, influencing the functions of local and systemic target tissues; they are responsive to purinergic ligands, hormones, and vitamins, and are sensitive to oxygen and ionic changes in their immediate environment. Their role in heme and iron metabolism and recycling are considered below. Tissue macrophages are embedded in extracellular matrix, express receptors for collagens, elastin, proteoglycans, and reticular/fibrillar constituents, including bound cytokines, and in turn generate potent extracellular proteinases and anti-proteinase inhibitors to modulate matrix, cellular, and plasma protein functions. Macrophages express a range of opsonic and non-opsonic membrane receptors for uptake and detection of microbial, sterile, and altered self-ligands; these include FcR, Complement R, TLR, lectins, and Scavenger receptors 11 . Cytoplasmic sensors include Nod-like and RIG-I-like helicases and cyclic dinucleotides. The effects of macrophage activation on plasma membrane, endocytic and cytosolic recognition and sensing mechanisms, signal transduction, and altered gene expression have been extensively documented 12 , 13 . Aside from differentiation, determined by lineage-determining transcription factors such as PU-1, chromatin organization and epigenetic mechanisms, extrinsic stimuli such as infection elicit dynamic changes in gene transcription, loosely known as cell activation 14 , 15 . Based on a few signals, these have been classified as ‘innate’ (TLR),’classical' (Th1/M1-type), and ‘alternative’ (Th2/M2-type) activation 14 , but are now thought to constitute a more complex multipolar spectrum of gene expression signatures 16 , 17 . A more realistic paradigm, extending it to the complexity of ligands present in different tissues and inflammatory conditions, is urgently needed. For the present, the questions arise, how is the phenotype of macrophages influenced by different tissue environments, and what are the effects of macrophage activation on the particular tissue in which they reside? A further issue is whether organ-specific differences persist after macrophage activation by inflammation, infection, and malignancy. We consider these questions with particular reference to selected resident and recruited macrophage populations and in their response to granuloma formation.

Since the early studies by light and intravital microscopy of invertebrates as well as mammals by developmental biologists/experimental pathologists such as Metchnikoff, it was apparent that macrophage-like cells were dispersed in many tissues, performing diverse homeostatic and defense functions 18 . The concept of a reticulo-endothelial system of phagocytic cells (RES) associated with Aschoff, emphasized the clearance functions of sinus-lining cells in liver, spleen, as well as in selected endocrine organs. Regional macrophages and related cells were also described in liver (Kupffer cells), lung (alveolar macrophages), and brain (microglia). Transmission and scanning electron microscopy of lympho-hematopoietic tissues emphasized the morphologic heterogeneity of monocytes and macrophages in different organs. From work on the spleen, it became apparent that macrophages could occupy different anatomical niches and perform specialized functions even within the same organ. In 1968, a group of investigators proposed the term ‘mononuclear phagocyte system’ to unify diverse populations of macrophage-like cells; however, there was still considerable confusion about which properties could be used to define the cells belonging to this extended family, as well as their origins, relationships, activation, and functional diversity 19 . Subsequent development of monoclonal antibodies directed against membrane antigens made an enormous impact on the field, followed by cellular biology and molecular genetic techniques to manipulate and establish precursor-product and lineage relationships. The group of Alan Williams and Neil Barclay, at Oxford 20 , utilized a systematic approach to produce and characterize a panel of anti-lymphocyte monoclonal antibodies of which several also reacted with myeloid cells. These have been particularly useful in characterizing rat lymphoid subsets, their localization and functions. In our own laboratory, we initially sought macrophage-specific reagents for mouse macrophages. The rat anti-mouse reagent F4/80, subsequently shown to recognize an adhesion GPCR EMR1, was particularly useful for immunocytochemical definition of macrophages during development, in the normal and challenged adult mouse, and drew our attention to tissue macrophage diversity in antigen expression 21 , 22 . The F4/80 epitope was selected to be stable to perfusion fixation and is mainly expressed on the plasma membrane, thus proving ideal to establish macrophage interactions with neighboring cells 23 , 24 . From these studies it became possible to delineate F4/80+ populations, which were associated with epithelia, endothelia and connective tissue, as well as free cells in the peritoneal and other serosal cavities. Definition of macrophages in bone marrow and spleen, suggested novel cell adhesive interactions and new monoclonal antibodies were subsequently isolated to discover some of the surface molecules involved; reagents for antigens such as CD169 25 exemplified and clarified the striking heterogeneity between marginal metallophils, for example, and red pulp macrophages in mouse spleen. Other monoclonal antibodies such as anti-CD68 provided pan-macrophage markers shared by macrophages 11 . A similar strategy was used to investigate macrophage adhesion, providing reagents to probe cell recruitment (CD11b) and scavenger receptor A (SR-A)-mediated endocytosis. Specific monoclonal antibodies were generated for previously described lectin-like receptors, such as the Mannose receptor, CD206. Table  1 summarizes the properties and expression of a range of these and related reagents. Application of this panel of reagents revealed striking heterogeneity between different tissue macrophages, e.g. in liver and mouse spleen ( Figs  1 and 2 ), but also within individual organs such as the brain ( Fig.  3 ). Selected antigens expressed on murine macrophages Ab, antibody; Ag, antigen; CNS, central nervous system; DC, dendritic cells; ICAM, intercellular adhesion molecule; Ig, immunoglobulin; LAMP, lysosome-associated membrane protein; LPS, lipopolysaccharide; LTA, lipoteichoic acid; Mϕ, macrophages; NK, natural killer cells; PMN, Sn, sialoadhesin; SR-A, type A scavenger receptor. For further information and references consult the text and Taylor et al . 11 . Immunocytochemistry illustrates the differential staining of macrophages in mouse liver and spleen. Kupffer cells (A) and splenic red pulp macrophages (B) are strongly F4/80 + , unlike F4/80 − marginal metallophils (C), which express CD169 strongly, and red pulp macrophages, which are CD169 dim or negative. Images (A) and (B) courtesy of D. A. Hume, image (C) courtesy of P. R. Crocker. A subpopulation of CD169 + metallophils binds the cysteine-rich domain of the mannose receptor (CyR-Fc) (A), and associates with IgD + B lymphocytes (B). Images courtesy of L. Martinez-Pomares. Reference 74 should be consulted for further details. Heterogeneity of F4/80 + cells in adult mouse brain. F4/80 + microglia are present in large numbers in all major divisions of the brain but are not uniformly distributed. There is a more than fivefold variation in the density of immunostained microglial processes between different regions. More microglia are found in gray matter than white. Microglia vary in morphology depending on their location. Compact cells are rounded, sometimes with one or two short thick limbs, bearing short processes. They resemble Kupffer cells of the liver and are found exclusively in sites lacking a blood–brain barrier. Longitudinally branched cells are found in fiber tracts and possess several long processes which are usually aligned parallel to the longitudinal axis of the nerve fibers. Radially branched cells are found throughout the neuropil. They can be extremely elaborate and there is wide variation in the length and complexity of branching of the processes. The systematic variation in microglial morphology provides evidence that these cells are sensitive to their microenvironment. Drawings to illustrate the morphological heterogeneity of macrophage populations of the central nervous system: A,C,D show microglia of the brain parenchyma (A, cortex; B, white matter; C, ventral pallidum). Macrophages with a simpler morphology are also present in the circumventricular organs (B), the meninges (E), and choroid plexus (F). Based on 106 which should be consulted for further details. Image courtesy of L. J. Lawson and V. H. Perry. Further reagents were developed by cDNA expression cloning and library subtraction. The function of Dectin-1as a β-glucan receptor was discovered by a screen using zymosan particles 26 . It was then straightforward to isolate monoclonal antibodies to study tissue expression of Dectin-1 and related antigens, their regulation and signaling. In some cases, we failed to identify ligands for cloned ‘orphan’ lectin-like antigens such as MICL 27 , recently achieved by others 28 . We have exploited monoclonal antibodies such as F4/80 and CD169 to study macrophage heterogeneity in wildtype and mutant mice. For example, osteopetrotic (op/op) naturally mutant mice lacking the macrophage-specific growth factor CSF-1 exhibit striking differences in F4/80+ expression: monocytes and resident peritoneal cells are markedly deficient in number as well as staining, whereas thymic macrophages are preserved; there is no expression of CD169 in op/op spleen, due to developmental absence of the marginal metallophilic population 29 . Panels of well characterized monoclonal antibodies provide a powerful tool to analyze genetically engineered mice and human samples; however, their full potential to determine the mechanisms and significance of macrophage phenotypic diversity has not yet been realized. In this review, we summarize applications of several monoclonal antibodies produced in our laboratory over the past 3 decades ( Table  1 and Figs  4 ), which helped to establish a basis for our knowledge of in situ heterogeneity of mouse macrophages in mouse tissues. BCG-induced granulomata. BCG-induced granulomata in mouse liver express F4/80 antigen (A). These recruited macrophages are distinct in distribution from Kupffer cells and monocytes in sinusoids and express lysozyme mRNA strongly and uniformly (B). Images courtesy of D. A. Hume and S. Keshav. See Martinez-Pomares and Gordon 146 for further details.

Macrophages play a key role in organ development 30 , 31 . F4/80 negative progenitors and precursors of macrophages can be detected by adoptive transfer and colony-forming assays; F4/80 staining first appears in mouse embryo at 8–9 days in the yolk sac, then in fetal liver from day 10, peaking at day 12–14. Subsequent hematopoiesis appears in bone marrow before birth. Definitive erythropoiesis is associated with the appearance of stromal-type macrophages in fetal liver, forming clusters with developing erythroblasts through non-phagocytic adhesion receptors, as discussed below. Fetal liver progenitors have been used to develop macrophage and DC lines in vitro 32 , 33 . In humans, placental cord blood is also a source of progenitors. DC and macrophage differentiation diverges from earlier common precursors, through distinct transcription factors 34 , 35 . Although they share many properties with macrophages, DCs constitute a functionally specialized, heterogeneous family of cells which are not discussed in detail in this review. One of the key functions of yolk sac and fetal liver macrophages is to seed developing organs; recent studies have highlighted the ultimate source of the yolk sac for microglia, Langerhans cells, and many other resident tissue macrophage populations that persist throughout adult life, depending on local replenishment and continuous low turnover 4 , 36 , 37 . Embryonic macrophages have been associated with phagocytosis of apoptotic cells in tissue remodeling and interaction with the extracellular matrix, important for general organogenesis and vasculogenesis 38 . Although infection is rarely surmountable in early stages of development, embryonic macrophages seem equipped with the gene repertoire to deal with pathogens 39 . We know very little of the differences among different embryonic macrophage populations or how they differ from macrophages in the adult; clues may arise from studies of the changes in extracellular matrix during development, known to alter the morphology and localization of cells, as well as colonization of the gut and skin by bacteria, shaping macrophage phenotypes and interactions with other cells, and finally, increased death or proliferation of cells needed in embryogenesis for tissue remodeling. In the adult, embryonic seeded resident tissue macrophages still constitute an abundant population; postnatally, it is convenient to distinguish resident macrophages, arising during development from embryonic sources, from elicited macrophages, recruited mainly from bone marrow in response to inflammatory stimuli.

We have reviewed some of the properties that underlie resident and recruited monocyte/macrophage heterogeneity in different mouse organs and tissue environments. Although it is apparent that differential gene expression is crucial during macrophage differentiation and activation, we have hitherto not fully exploited the power of genome and proteome-wide analytic tools to catalog common as well as distinctive tissue macrophage genes, proteins, and functions in different environments. This is partly due to difficulties in macrophage extraction from solid tissues, and limited in situ single cell analysis of multiple candidate gene products. Furthermore, we have very little understanding of macrophage-specific-transcriptional regulators, except where particular tissue functions can be implicated, e.g. following the clearance of senescent erythrocytes and turnover of hemoglobin in splenic red pulp, bone marrow, and liver. The macrophage plasma membrane receptor profile within different tissues, which we have emphasized in this review, is likely to determine their distinct gene expression profile, responding to information from other cells and ligands in each particular environment. Above all, we are ignorant of stromal-macrophage interactions, especially the nature and possible regional variation in stromal cells and extracellular matrix within different organs. This may benefit from novel culture models, and also from the experience of cancer cell biologists in studying tumor–host interactions in vivo and in vitro 162 . Improved understanding of phenotypic heterogeneity will guide our attempts to manipulate and target macrophage subpopulations selectively, an unmet need in disease prevention and treatment. The increasing promise of monoclonal antibody therapy depends on systematic study in mouse models and extensive validation in humans. This will require better monoclonal antibodies for immunocytochemistry of human macrophages in normal tissue and disease and improved methods to analyze their complexity.

One of the questions raised earlier relates to the impact of different tissue environments on monocytes and macrophages recruited as a result of inflammation and infection. Even in highly specialized tissues, for example brain, it can be difficult to distinguish these activated cells from resident macrophages such as microglia, considered above. Granulomata provide useful models to explore this issue. Fig.  4 A illustrates F4/80 expression in BCG-induced granulomata in mouse liver, which provides a relatively uniform background of Kupffer cell staining; Fig.  4 B demonstrates lysozyme mRNA expression by the bulk population of activated macrophages, detected by in situ hybridization 147 . Expression of TNFα message, by contrast, is restricted to a subpopulation of activated macrophages (data not shown). Granuloma formation involves dynamic 148 , organized collections of recruited monocytes/macrophages, together with lymphoid and other myeloid cells, especially eosinophils and neutrophils, with connective tissue cells, particularly fibroblasts and vascular/lymphatic elements. These are often associated with chronic inflammation and giant cell formation but include foreign body granulomata. It is customary to distinguish Th1 type, e.g. in tuberculosis, and Th2 type granulomata, e.g. surrounding Schistosome eggs 149 . Granuloma formation is exceptional in viral infection but can accompany autoimmune/idiopathic inflammatory disorders such as sarcoidosis, Crohn's disease, and Wegener's granulomatosis. Granulomata can be relatively benign or tissue destructive, e.g. in bone, but are not malignant, monoclonal disorders except where mimicked by lymphoid malignant diseases such as Hodgkin's disease. The foreign body type granuloma tends to be relatively inert, without lymphoid cell activation; immune granulomata cells turn over more actively, and cytokines such as TNF and IL-4/13 play an important role in their formation 149 , 150 . The macrophage phenotype depends on the nature of the eliciting agent, e.g. persistent infection by poorly degradable mycobacteria or parasite eggs, as well as the cytokine environment. In tuberculosis, the mycobacterial lipids contribute to giant cell formation and macrophage fusion. The lesions contain recently recruited CR3 + monocytes, so-called epithelioid cells 151 , unfused macrophages, and polykaryons (Langhans giant cells) 152 , 153 . Differences among these macrophage populations have been described, with regard to antigen expression, endocytosis/phagocytic and secretory activity, but need further validation in vivo . Th2 type multinucleated giant cells, dependent on fusogenic IL-4/IL-13, lack bone degradation capacity, demonstrated by osteoclasts, with which they have much in common (Helming, Milde, unpublished observations). It is not clear whether particulates, T cells/macrophages or stromal cells/extracellular matrix are the primary determinants of the macrophage phenotype in granuloma formation and fibrosis. It would be interesting to compare the phenotypes of granuloma macrophages with those of foam cells in atherosclerosis and with tumor-associated macrophages in malignant diseases. Such studies would provide insights into the importance of environment in macrophage heterogeneity during granuloma formation, the possible advantage to host or pathogen of sequestration, the relevance to restricted cellular migration or retention in tissues, and metastasis.