The Mononuclear Phagocyte System of the Rat

This review focuses on advances in understanding the biology of innate immune cells in the rat. This focus immediately raises the question, why rats? The laboratory rat (Rattus norvegicus) was first used in physiological research in the early 19th century, and defined strains were generated early in the 20th century (1). Rats are still recognized as informative and tractable models for many human diseases, including cardiovascular, neurologic, cancer, diabetes, respiratory, and inflammatory disease (2). They are often chosen as a model species because of their greater similarity to humans (compared with mice) in multiple biological aspects, along with their larger size enabling easier surgical procedures and high-resolution imaging (3, 4). In terms of neurologic research specifically, rats are favored relative to mice because of their greater intelligence and behavioral repertoire. They are less stressed by human handling compared with mice, which in turn reduces compounding effects in physiological studies (3, 5, 6). One review even asked the rhetorical question, “Are rats more human than mice?” (7). Some of these advantages have driven the development of rat models with humanized immune systems (8, 9). The disadvantage of rat models is that there are fewer available inbred/defined and mutant strains, and they require more space and are more expensive to breed and maintain than mice. Consequently, there is a smaller research community, and the current set of available rat-specific diagnostic tools, such as monoclonal Abs, is relatively limited compared with mouse and human.

It is almost 20 years since a draft rat genome became available (10). Until quite recently, the targeted modification of the rat genome was considered difficult (11). However, the genetic toolbox for rats evolved rapidly with the introduction of N-ethyl-N-nitrosourea mutagenesis (12), zinc-finger nucleases (13), and homologous recombination (14). With the more recent advent of CRISPR–Cas9 technology for genome editing, it has become feasible to simply and reliably create rat strains with multiple different gene-targeted knockouts in one step by delivery of specific guide RNAs, Cas9 mRNA, and donor sequences into fertilized eggs (15–17). Szpirer (18) cataloged more than 350 rat genes in which rat lines with natural or introduced variants provide models for human disease. The Cre recombinase system has been widely used with mice to generate conditional, tissue-specific, and inducible knockouts and to perform lineage tracing studies (19, 20). This system has also now been extended to rats, and several studies have described lineage-specific Cre alleles and Cre-dependent reporters in this species (21–26). The generation of rat embryonic stem cell lines has also enabled analysis of the impact of mutations on differentiation processes in vitro (23) and the generation of macrophages from embryoid bodies (27).

High throughput genome sequencing has also changed the rat research landscape. The availability of whole-genome sequences of multiple rat strains with well-characterized genetic disease susceptibility [see https://rgd.mcw.edu/ (2)] revealed evidence of selective sweeps associated with breeding for the disease trait, in many cases overlapping human disease susceptibility loci to a much greater extent than similar trait loci in inbred mice strains (28). One such susceptibility locus revealed a novel role for inducible zinc transporter ZIP12, encoded by Slc39a12, in hypoxia-associated pulmonary hypertension (29). Because of its larger size, the rat is especially relevant to cardiovascular research—particularly for hypertension and stroke (3). Cardiovascular disease is associated with atherosclerosis, which is caused by the disproportion of blood lipids (30). APOE, which is highly expressed by macrophages, contributes to the transport of cholesterol and other lipids (30). The generation of an Apoe knockout rat provided a major advance. The same group also produced a targeted disruption of the low-density lipoprotein receptor (Ldlr) gene and compared the pathologies of vascular lesions between mouse and rat models (31). The initial stage of atherosclerosis in the Apoe knockout rat had adventitial pathology that closely resembled human lesions (32). Specifically, adventitial immune infiltrate comprising macrophages and T cells was detected prior to intimal thickening (31).

Baud et al. (33, 34) reported the analysis of 195 phenotypic traits related to metabolism, osteoporosis, hypertension, anxiety, immune status, and hematology in a population of heterogeneous outbred rats descended from eight rat inbred progenitor lines. They also characterized susceptibility of individual animals in a number of disease models, including wound healing and experimental autoimmune encephalitis. The genomes of the progenitor inbred lines were sequenced, and the phenotyped animals were genotyped using an 800K single-nucleotide polymorphism array. A total of 35 putative causal variants were implicated in 31 different phenotypes, but in most cases, the extent of variation at individual quantitative trait loci prevented connection between a single sequence variant and the phenotypic impact.

All these technologies have driven what has been called a comeback or renaissance for rat models in biomedical research after 20–30 y of mouse dominance (4, 5). A current PubMed search on rat AND macrophage AND disease AND model (March 2021) identified 6900 references (still 4-fold less than the same search with mouse). In part driven by skepticism about the value of inbred mice as a model (35), our group has invested in the rat as an alternative in which to study homeostasis and the functions of macrophages in development and disease. In this article, we review what is known about the rat mononuclear phagocyte system (MPS), the available and missing resources, and the future opportunities.

The MPS is a family of cells with related function and gene expression profile that includes progenitors in bone marrow, monocytes, dendritic cells (DC), and resident macrophages in every organ of the body. Cells of the MPS are central to innate immunity and development of acquired immunity but also contribute to pathology of infection, chronic inflammation, malignancy, and obesity and to normal development, tissue homeostasis, and wound repair (36–40). The original concept of the MPS recognized that resident tissue macrophages can be replaced by circulating monocytic progenitors but have a relatively long half-life and are capable of self-renewal (41). The concept has been misrepresented in more recent literature (42) as a dogma that proposed that blood monocytes continuously replace resident macrophages in the steady state. Macrophages arise first in the yolk sac and subsequently in the fetal liver during embryonic development in mice (43, 44). Some of the earliest detailed studies of mammalian fetal macrophage differentiation were carried out in the rat. Takahashi and colleagues (45, 46) described the appearance of macrophage-like cells within blood islands in the rat embryonic yolk sac around fetal day 9, and the subsequent rapid maturation of active phagocytes that continued to proliferate. These embryonic phagocytes did not transition through a monocyte-like intermediate. When the vitelline vessels merged with the embryonic cardiovascular system, the yolk sac–derived macrophages spread throughout the rat tissues.

Based upon lineage trace studies in inbred mice, it has been suggested that macrophages in some organs, notably brain, develop from an erythro-myeloid progenitor that arises in the yolk sac and/or fetal liver and remain exclusively of fetal origin throughout life. Macrophages in other organs such as gut, skin, heart, and kidney are apparently replaced progressively by monocytes derived from classical hematopoietic stem cells (HSC) (40, 42, 47, 48). The evidence supporting the current view of MPS ontogeny in the mouse is not unequivocal and has been critically reviewed elsewhere (35). Whether the emerging alternative “fetal origin” dogma is true in any system other than a male C57BL/6 mouse in a specific pathogen–free facility remains to be determined (35, 49). Comparable lineage trace studies have not yet been feasible in rats. What is clearly the case is that in mice, cells derived from HSC via a blood monocyte intermediate are able to re-establish the macrophage populations in most organs and adopt local tissue-specific phenotypic adaptations if the original residents are experimentally depleted (40). As discussed below, adult bone marrow progenitors are also able to repopulate all tissue macrophage populations in macrophage-deficient rats (S. Keshvari, M. Caruso, L. Batoon, A. Sehgal, N. Teakle, O. L. Patkar, C. E. Snell, C. Chen, A. Stevenson, F. M. Davis, et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859).

There are countless well-defined rat anti-mouse monoclonal Abs that bind to cell surface or intracellular proteins expressed by mouse mononuclear phagocytes, examples include LY6C, CD14, CD11B, CD11C, CD115 (CSF1R), CD32, CD43, CD64, CD163, CD169, CLEC4F, SIGLECF, MERTK, TIMD4, TMEM119, P2RY12, CD206 (MRC1), LYVE1, and F4/80 (ADGRE1). Many of these surface markers have been used to define subpopulations of macrophages in specific locations within tissues of the mouse (50, 51). Although mRNA encoding each of these markers is highly expressed by at least some populations of rat macrophages (27, 50, 52) and mouse anti-rat leukocyte Abs were among the first made using mAb technology (53), the set of available Abs recognizing rat mononuclear phagocytes has not expanded greatly since it was reviewed 20 y ago (54). Table I summarizes available monoclonal Abs that have been used to define macrophages in rat tissues or by flow cytometry. The large majority of published studies that localize macrophages in tissues in rat disease models still use CD68 (detected by mAb ED1) and CD163 (ED2) as markers, even though it is clear that they label only a subset of tissue macrophages. CD169 (ED3), as in mice, labels specific subpopulations of tissue macrophages, including those of the marginal zone of spleen, the subcapsular sinus, and paracortical macrophages of lymph node (55) and alveolar macrophages (56) but has also been considered a marker of inflammatory macrophages (57, 58). In mice, CD169 in bone is a marker for both hematopoietic island macrophages and bone-associated osteal macrophages (osteomacs) that regulate osteoblast function (59), whereas in rat marrow, CD169 was undetectable (54), and erythroblastic island macrophages express CD163 (60). Rabbit anti-CX3CR1 has also been used in some studies as a marker of inflammatory macrophages (57). The MRC OX-42 Ab, which binds CD11B/C (61), is commonly used to distinguish microglia in the rat brain (62) but is also detected on the majority of resident macrophages in the rat (61). MRC OX-41, which also labeled most resident macrophages/DC in the same study (61), was later shown to bind SIRPA (CD172A). Abs raised against several members of the Ag-presenting lectin-like gene complex [CLEC4D (MCL), CLEC4E (MINCLE), and DCAR1 (63–65)] have been shown to label many tissue macrophage and subpopulations. One marker that has not been widely employed in rat studies is IBA1 (Aif1 gene). IBA1 is a highly expressed cytoplasmic calcium-binding protein involved in myeloid cell motility. Although it is most commonly applied to staining microglia in the brain, it is expressed by tissue macrophages throughout the body in rats and mice (Ref. 66 and S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859). A commercial Ab raised against rat CD32 (FCGR2) is used to block nonspecific binding to Fc receptors on flow cytometry (67) but has not itself been widely used as a marker.

There are cross-reactive commercial recombinant rabbit Abs or polyclonal Abs against several key markers that would be useful markers for analysis of rat MPS progenitors and subpopulations (e.g., CD206, CD14, CD64, KIT, P2RY12, P2RX7, FLT3, LYVE1, TIM4, TMEM119, and CSF1R) but few publications demonstrating their utility or specificity. As discussed below, we have used some of these reagents to localize macrophage subpopulations in tissues. One obvious missing reagent is a mouse anti-rat ADGRE1 (F4/80). F4/80 is widely employed as a macrophage marker in mice (35, 50); F4/80^hi resident macrophages were proposed to be exclusively of embryonic origin (68). A rabbit anti-rat F4/80 heteroantiserum was used to localize rat macrophages and microglia (69), but a monoclonal is not yet available.

The DC as a distinct entity within the MPS was originally defined in lymphoid tissues based upon its arborized morphology (70) and later based on a proposed unique ability to present Ag to naive T cells (71, 72). The latter view has evolved into a circular logic; any cell with APC capacity is often considered a DC by definition, and DC are described as the sentinels of the immune system (73). However, the distinction between macrophages and DC in mice based upon APC activity (74) or surface markers (51, 75) remains tenuous and somewhat ephemeral. For example, active efferocytic resident macrophages in T cell areas of lymph nodes were previously classified as DC (76). The confusion based upon markers and function in turn has led to efforts to classify the MPS based upon ontogeny, segregating monocyte-derived “DC” from a separate classical DC lineage derived from a committed DC progenitor (77). As in mice (78), cultivation of bone marrow cells in CSF2 (GM-CSF) compared with CSF1 (M-CSF) leads to selective expansion of class II MHC (MHCII)⁺ APC (79). The key growth factor for the proposed classical DC (cDC) lineage is FLT3 ligand. Based upon this paradigm, several studies (80, 81) have compared surface marker and function in rat bone marrow–derived DC generated in CSF2 and FLT3 ligand.

There are no definitive markers to segregate rat DC from macrophages. CD11C has continued to be employed as a DC marker in mice despite abundant evidence of expression by definitive macrophages (51). In rats, an Ab specifically recognizing CD11C raised originally with rat alveolar macrophages as the immunogen bound uniformly to blood monocytes and resident splenic macrophages (82) and to bone marrow–derived DC and macrophages (79). OX-62 (CD103) has been referred to as a DC marker in rats (83–85), whereas in mice, it would be considered a specific marker for the cDC1 subset of DC (77). Also, by contrast to mice, in which many tissue macrophages are MHCII^low/− (86), the large majority of tissue MPS cells in the rat expressed abundant MHCII recognized by the OX-6 Ab (87, 88). At the time of these studies, the MHCII^hi cells in rat tissues were referred to as DC, based purely on morphology and inferred APC activity. Park et al. (89) produced an Ab (HD83) against human LY75 (also known as DEC-205) that cross-reacted with rat and labeled rat splenic putative DC (defined by morphology and MHCII expression). The early studies on rat DC in vivo were focused on migrating cells isolated from afferent lymph and employed CD274, SIRPA (OX41), MHCII, CD80/86, CD103, and CD4 as markers to distinguish subpopulations (90–92). Indeed, many rat macrophages coexpress both CD4 and CD8 (7, 82). In a rather elegant study using a rat mesenteric lymphadenectomy model, Yrlid et al. (93) showed in adoptive transfer studies that monocytes could enter intestinal mucosa and give rise to migratory DC in afferent lymphatics. This key finding calls into question the view from the mouse studies that classical DC and monocyte-derived APC represent separate lineages (77).

The proliferation and differentiation of most resident tissue cells of the MPS is controlled primarily by signals from the M-CSF receptor (CSF1R), which responds to two ligands, M-CSF (CSF1) and IL-34 (35, 94, 95). Consistent with the core function of CSF1R in macrophage development and homeostasis, Csf1r mRNA is expressed by the earliest phagocytes in the developing mouse yolk sac and in progenitors, blood monocytes, and all tissue-resident cells of the MPS (43, 51, 96). On this basis, we have studied the transcriptional regulation of the mouse Csf1r locus as a model for understanding MPS differentiation (97). Various tools have been developed based upon the promoter region, including Csf1r–EGFP and mApple reporter genes (98–100). Recently, the development of a knock-in FusionRed reporter within the mouse Csf1r locus provided conclusive evidence that the CSF1R protein is expressed only in MPS lineage cells (101). To address the need for an equivalent reporter in rats, Irvine et al. (94) generated a Csf1r–mApple transgenic line in which the reporter is robustly expressed in MPS lineage cells. As in the mouse, direct visualization of Csf1r–mApple in tissues highlights the abundance, the shared morphology, and the remarkably regular distribution of resident macrophages in every tissue of the body, including locations such as smooth and skeletal muscle, where the resident population has not been widely recognized. Irvine et al. (94) presented a wide diversity of tissues imaged using spinning disc microscopy. Fig. 1 illustrates the point further with whole-mount direct imaging of populations of macrophages in a novel set of additional locations captured by confocal microscopy on unfixed tissues. The abundance of resident macrophages in situ contrasts with low yields of isolated cells in mouse and rat following standard tissue disaggregation procedures.

As noted above, it is not clear whether the details of mononuclear phagocyte development and homeostasis based on studies of lineage trace models and mutations in one inbred mouse strain can be extended to other species or even to other mouse strains. One key difference between mice and rats lies in the transcriptional regulation of Csf1. Macrophages generated in vitro from either monocyte or bone marrow progenitors by cultivation in CSF1 have been employed as models for the study of macrophage biology in multiple species (102). CSF1 was also used in procedures developed to generate macrophages from mouse embryonic stem cells, a method applicable with modifications to the rat (27). Macrophages themselves internalize and degrade their own growth factors via CSF1R-mediated endocytosis, providing one homeostatic mechanism that links the entire MPS through the local and circulating CSF1 concentration (35, 40, 75). Mouse bone marrow–derived macrophages (BMDM) require continued stimulation with exogenous CSF1, and without that stimulation, they undergo apoptosis (103). Csf1 mRNA is undetectable in mouse BMDM and very low in isolated resident macrophages, although inflammatory macrophages can express Csf1 and CSF1 target genes (104). In rats and humans, in a range of other mammalian species examined (102), and even in chickens (105), Csf1 mRNA is expressed at high levels in BMDM or monocyte-derived macrophages (MDM), and exogenous CSF1 is not required for macrophage survival in vitro. The Csf1 gene encodes multiple protein isoforms by alternative splicing, including secreted glycoprotein and proteoglycan forms and a plasma membrane form (106). In rat macrophages, the predominant mRNA includes all of exon 5 and encodes the membrane-anchored protein (data in Ref. 102). This form of the protein is cleaved from the cell surface by the enzyme ADAM17 (107). The differences in CSF1 expression between mice and other species may contribute to the differences in effects of mutations discussed below.

By comparison with CSF1, less is known of the biology of IL-34, and evidence for specific functions in disease is emerging (reviewed in Ref. 108). Il34 mRNA in mice is enriched in expression in brain and epidermis, and loss-of-function mutations in mice are associated with selective loss of microglia in the brain and Langerhans cells in the skin. The expression of IL-34 in the mouse brain is region specific, and knockout mutation in mice is associated with selective loss of microglial populations. A recent study used anti-CSF1 and anti–IL-34 Ab treatment to infer specific requirements for microglial maintenance in white and gray matter regions, respectively (109). By contrast, IL34 mRNA is equally abundant in skin and in a wide range of human brain regions (Ref. 110; see also http://biogps.org) but was also expressed by many other tissues and cell types. This pattern is also evident in a detailed analysis of brain regions of juvenile rats (111) in which we found no evidence of region-specific expression, and Csf1 and Il34 were expressed at similar levels.

A frame-shift mutation in the Csf1 gene in mice was recognized as the cause of osteopetrosis in the op/op mouse strain. Many subsequent analyses have revealed selective loss of tissue macrophage populations and pleiotropic impacts of the Csf1^op/op mutation on somatic growth, fertility, and organ development (reviewed in Ref. 112). On the original outbred background, the osteopetrosis (osteoclast deficiency) resolves with age, whereas on the inbred C57BL/6 background, the Csf1^op/op mutation (112) or a targeted Csf1 knockout (113) has severely compromised postnatal survival. Mutation of Csf1r on a mixed genetic background mimicked many effects of Csf1 mutation (114). The Csf1r mutation is also much more severe in C57BL/6 mice, with few homozygotes surviving to weaning (115).

Loss-of-function mutation in the rat Csf1 gene was demonstrated in the toothless (tl) rat line (116, 117), confirming previous indications of the underlying cause of osteopetrosis based upon partial phenotypic reversal by CSF1 administration (118). The published studies of Csf1^tl/tl rats have focused on a distinct skeletal developmental phenotype (119–123). Unlike Csf1^op/op mice, which have some residual osteoclasts and improve with age, Csf1^tl/tl rats were devoid of osteoclasts. The bone phenotype is more severe than in mice and does not resolve with age (117).

We knocked out the Csf1r gene in rats using homologous recombination in embryonic stem cells (62). Unlike the inbred Csf1rko mouse, the Csf1rko rats even on an inbred background are viable to 7–10 wk despite the complete loss of macrophages in the embryo and of all resident myeloid cells (macrophages and DC) recognized by IBA1 staining in most organs (Refs. 62, 111, and S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859). There are severe impacts of the loss of tissue macrophages on postnatal somatic growth, musculoskeletal development, maturation of all major organs, and fertility that are partly attributable to defective postnatal growth and functional maturation of the liver and associated with almost complete lack of circulating IGF1 (S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859). The Csf1^tl/tl rat was also reported to be deficient in circulating IGF1 (124). The skeletal phenotype of Csf1rko rats closely resembled the impacts of biallelic recessive mutation in humans (48). In direct contradiction of the reported impact of heterozygous Csf1r mutation in mice (125), there was no detectable phenotype associated with heterozygosity (haploinsufficiency) in rats, even though the mutation is not dosage compensated at the mRNA or protein level in individual MPS cells (62, 111). Deficient macrophage populations and the phenotypic impacts of the Csf1rko in rats were reversed by i.p. transfer of wild-type bone marrow without conditioning at weaning (S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859), providing a novel model in which to dissect the precise mechanisms of action of macrophages in control of postnatal development.

Loss-of-function mutation in rat Il34 has not yet been analyzed. However, the phenotype of the Csf1rko appears to be much more severe than the CSF1-deficient Csf1^tl/tl rats. Csf1^tl/tl rats have unremitting osteopetrosis, but they achieve adult body weights of 250–300 g (compared with 50–100 g in the Csf1rko) and have normal longevity (126). In mice, Csf1rko and Csf1^op/op mutations on a common mixed genetic background had indistinguishable effects of skeletal development and postnatal weight gain (114). Furthermore, unlike Csf1^op/op mice (127), Csf1^tl/tl male rats were fertile and were used in breeding experimental cohorts (117, 126). Hence, we can infer that IL-34 in rats likely has nonredundant functions in postnatal development that extend beyond those inferred from the Il34ko in mice.

The reciprocal to the loss-of-function studies can be achieved following the development of rCSF1. Injection of recombinant human CSF1 in mice caused a substantial increase in blood monocytes and tissue macrophages (128) and, as noted above, partly rescued the Csf1^tl/tl rat bone phenotype (118). In wild-type rats, CSF1 administration aggravated pathology in an experimental arthritis model (129). However, studies of CSF1 action and evaluation of therapeutic potential were compromised by the short circulating half-life that necessitated infusion in early clinical trials (reviewed in Ref. 130). This constraint was resolved by the generation of a CSF1–Fc fusion protein with an increased circulating half-life (131). Administration of this protein to neonatal rats promoted the growth and maturation of the liver (132), the reciprocal of the hepatic developmental failure in the Csf1rko. In adult rats, CSF1–Fc promoted a monocytosis, expansion of all tissue macrophage populations, and further growth of the liver and spleen (94). Hence, CSF1–CSF1R signaling controls and integrates development and function of the rat MPS, and the rat may provide an alternative model to the mouse for evaluation of therapeutic interventions targeting this axis.

Blood monocytes develop from committed progenitors in the marrow through a highly proliferative monoblast intermediate (133). Their release into the circulation and their subsequent extravasation into tissues in the steady state and in response to inflammation is controlled by chemokines signaling through chemokine receptors, notably CCR2 and CX3CR1 (134). Within the blood, monocytes in mice, humans, and rats have been classified into two subpopulations, referred to as classical and nonclassical (135). The markers used to distinguish the subpopulations differ between species, as do the relative proportions. In the rat, the nonclassical subpopulation, defined by expression of CD43, is predominant (93, 135, 136), whereas in mice, the split is around 50:50, and in humans, the majority of monocytes are classical (defined as CD14^hi, CD16^lo). The subdivision is artificial; it is clear that monocytes in all species are a differentiation series dependent upon CSF1R signaling (137). An intermediate human monocyte population has a transcriptomic profile that is intermediate between the extremes (138). Accordingly, in the Csf1rko rat, there is a selective loss of CD43^hi monocytes, leading to an apparently more severe monocytopenia than in the mouse (S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859). The differentiation of CD43^hi monocytes from CD43^lo classical monocytes and the associated reduced expression of the chemokine receptor CCR2 and the adhesion molecule L-selectin (CD62L) and increased expression of CX3CR1 were confirmed in the rat by Yrlid et al. (93). The classical monocytes have a relatively short half-life, whereas nonclassical monocytes are long lived and may have specific roles in blood vessel homeostasis (139). The relatively low proportion of classical monocytes in rats compared with mice likely reflects differences in turnover/extravasation of short-lived classical monocytes: rats appear to have approximately half the total monocyte count of mice (140). There are few recent studies of monocyte turnover in rats, but older studies (141) indicate a half-life of only 12–13 h, based upon thymidine labeling, compared with 22–24 h for mouse monocytes estimated using similar approaches (41) and later confirmed using alternative assays (139). Differences in monocyte turnover most likely also contribute to distinctive patterns in other species: the very low monocyte count in sheep (142) and the lack of a definitive nonclassical population in pigs (143).

Blood monocytes are the most accessible MPS population in humans, and they are commonly differentiated in vitro in CSF1 to generate MDM or in CSF2 (GM-CSF) to generate so-called monocyte-derived DC. These populations in humans were expression profiled by the FANTOM consortium (144). Monocyte culture is not practical in mice, but in rats, it is feasible because of their 10× greater blood volume. We recently showed that MDM and BMDM from rats, grown in CSF1, have very similar gene expression profiles (27) as previously shown in pigs (145).

Although they have similar morphology and location in tissues and share expression of Csf1r and genes associated with endocytic functions, resident tissue macrophage populations adapt to their specific tissue niche and environment with different gene expression patterns that enable specific functions per tissue (39, 42, 51, 146). In mice, tissue-specific macrophage adaptation is associated with unique transcriptional profiles and expression of specific markers that become evident late in embryonic development (96, 147). Local adaptation is, in turn, driven by unique transcription factors: SALL1 in microglia, GATA6 in the peritoneum, NR1H3 in the marginal zone of spleen and liver, SPIC in the splenic red pulp, PPARG in alveolar macrophages, AHR in Langerhans cells, and BATF3 in classical DC (39, 42, 51, 146). In the past 5–10 y, there has been a deluge of RNA sequencing (RNA-seq) data (including single cell RNA-seq) on mouse MPS cells isolated by tissue disaggregation, describing macrophage heterogeneity between and within tissues (reviewed in Ref. 51).

There have been few comparable analyses of rat macrophage populations, in part because of the relative lack of markers to support separation of cells from disaggregated tissues. The Csf1r–mApple rat transgenic line described above provides one solution (94). To date, expression profiling has only been carried out on relatively accessible populations from the lung and peritoneal cavity and on brain microglia (27). Microarray analysis of rat alveolar macrophages revealed shared enriched transcriptional regulator genes between mouse and rat include Pparg, Cebpa, Runx2, Nr1h3, and Lmo4. Among the most strongly enriched genes in rat peritoneal macrophages (relative to all the other populations) were Serpine1, encoding plasminogen activator inhibitor-1 (PAI-1), and Serpinb2, encoding PAI-2, also specific to this population in mice (148). Similar to mice, rat peritoneal macrophages also overexpress multiple serine protease inhibitors, Slpi, Serpinb6, Serpinb9, Serpinb10, and Serping1, perhaps reflecting the large number of trypsin-binding proteins observed in peritoneal macrophage lysates (149). MHCII-expressing peritoneal macrophages in mice depend upon the transcription factor IRF4 (86), which is also strongly expressed in rat peritoneal macrophages. In addition to Gata6, the rat peritoneal macrophages selectively expressed genes encoding multiple transcription factors at least 2-fold higher than in alveolar macrophages, notably Ahr, Mitf, Tfec, Batf3, Batf2, Stat1, Creb5, Mef2c, Id1, Etv1, and FoxP1.

Microarrays have largely been supplanted by RNA-seq, but compared with the extensive mouse resources (51), there have been fewer RNA-seq datasets generated for isolated rat macrophages. One approach to identifying macrophage-associated transcriptional signatures is to identify sets of transcripts that are correlated with each other in large datasets derived from diverse tissues and cell types. We have used this approach to derive coexpression signatures associated with macrophages in human, mouse, and multiple livestock species using our own and public domain data (150–156). As part of a larger rat transcriptional atlas project, we have downloaded and renormalized available public rat RNA-seq data for major lymphoid organs and isolated macrophages using the same approach described for mouse, pig, and chicken (51, 151, 155). Fig. 2 shows a sample-to-sample and gene-to-gene correlation matrix for these data generated using the freely available network analysis tool Biolayout as also used in the previous analyses. Supplemental Table I contains the complete dataset and the genes found within coexpressed gene clusters. This is an idiosyncratic dataset including tissues from an ageing profile and various experimental treatments. For coexpression analysis, such diversity is an asset because it enables identification of the most robust clusters of transcripts that share transcriptional regulation. Consistent with that view, network analysis enables the identification of a number of robust rat macrophage-enriched coexpression clusters that are, to some extent, tissue specific. The largest, cluster 2, which contains Csf1r, is most highly expressed in microglia and contains many of the known microglial markers (including Aif1, Cx3cr1, P2ry12, Trem2, Tmem119, and others highlighted in Supplemental Table I), the majority of which were depleted in the total mRNA analysis of the brains of Csf1rko rats (62, 111). Cluster 8 provides a signature of genes associated with endosomes (Cd68 and Gpnmb) and lysosomal hydrolases that were also enriched as expected in macrophages in all species studied above. Expression of transcripts in cluster 14 is highest in peritoneal exudate cells and includes the peritoneal macrophage transcription factor, Gata6. Adgre1, which encodes the macrophage-expressed F4/80 marker in mice, is part of a smaller cluster (cluster 38); as in mice, it is expressed at lower levels in monocytes and in lung. Similarly, Mrc1 (encoding CD206), which defines subpopulations of resident tissue macrophages in many tissues in mice (51), forms part of a small cluster (cluster 40) that was most highly expressed in a population of perivascular macrophages isolated from the brain. Clusters 17, 18, and 30 contain immediate early genes, Il6, Tnf, and many known IFN-responsive and proinflammatory genes. The three clusters differ in the temporal profile of response to LPS (see below). This network analysis also identifies distinct clusters containing markers associated specifically with B cells (Cd19, cluster 3) T cells (Cd3, cluster 5), granulocytes (Elane and Mpo, cluster 24 and cluster 33), NK cells (Ncr1, cluster 20), and cellular processes including cell cycle (cluster 1), mitochondria/oxidative phosphorylation (cluster 23), and protein/RNA synthesis (cluster 11). Interestingly, cluster 83, which contains the transcription factor Ciita, Cd74, and MHCII genes, also contains the DC-associated Flt3 transcript. Clearly, this is a preliminary analysis that will become more powerful as larger datasets become available for the rat.

The analysis of isolated resident macrophages in mice is compromised by two artifacts: the extensive contamination with other cells and activation during the isolation process (51). One alternative is to identify the sets of genes that are absent or greatly reduced in the tissues of Csf1rko rats. Analysis of liver and spleen of Csf1rko rats (Ref. 62 and S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859) revealed the selective loss of transcripts that were previously shown to be Kupffer cell–specific in mice (e.g., Clec4f, Cd5l, and Vsig4) and those associated with the marginal zone macrophages in spleen (e.g., Siglec1 and Cd209).

Microglia, the resident macrophages of the CNS, have been among the most studied MPS populations in rats because of the links to dementia and neurodegeneration and the utility of the rat in behavioral studies. There are >10,000 published reports on PubMed (microglia AND rat). Microglia are believed to be essential for innate immunity, normal CNS development and function, and modulation of neuroinflammation and homeostasis (49, 157–159). Against this background, the characterization of a mouse hypomorphic Csf1r enhancer mutation (Csf1r^{ΔFIRE/ΔFIRE}) that lacks microglia but is otherwise healthy and normal was surprising (160).

The Csf1rko rat is also entirely microglia deficient and, unlike the Csf1r^{ΔFIRE/ΔFIRE} mice, also lacks brain-associated and many peripheral macrophage populations. Nevertheless, the brain develops relatively normally apart from ventricular enlargement, a phenotype shared with biallelic CSF1R mutations in humans (62, 110). We performed total RNA-seq expression profiling of multiple brain regions of juvenile male and female Csf1rko rats and identified a set of 105 Csf1r-dependent transcripts that was consistent with published microglia-specific expression profiles (62, 111). These include genes encoding surface receptors such as P2RY12 and TMEM119, which have been considered definitive markers of microglia. Fig. 3 illustrates the strict colocalization of these markers with IBA1 in rat brain, also emphasizing the extreme ramification of microglial processes. The microglial signature, detected readily in total RNA-seq, provides a surrogate indication of microglial abundance and phenotype. The relative abundance of the set of microglia-specific transcripts detected by total RNA-seq was not dependent upon sex and did not differ among brain regions. This conclusion contrasts with an emerging literature in mice that emphasizes microglial heterogeneity but that depends largely upon analysis of cells isolated from the brain by disaggregation (49, 157–159), a process that introduces many potential artifacts (51). As in the mouse, a dense network of IBA1⁺ macrophages is established in the rat embryo by mid gestation and is almost entirely depleted in the Csf1rko rat (S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859). Nevertheless, somatic growth, the microglial population of the brain as well as peripheral macrophage populations, and the gross phenotype can be rescued by i.p. transfer of Csf1r^+/+ bone marrow cells at weaning (S. Keshvari et al., manuscript posted on bioRxiv, DOI: 10.1101/2020.11.29.402859). Surprisingly, the wild-type bone marrow cells do not restore CSF1-responsive bone marrow progenitors or blood monocytes. Alongside data on bone marrow transfer in the chick (161) and in the mouse Csf1rko (162), these data indicate that there are cells within bone marrow that are distinct from HSC that are able to sustainably repopulate the entire MPS.

Macrophage numbers in tissues increase in response to inflammatory stimulation through a combination of monocyte recruitment and local proliferation. They respond to signals in the inflammatory lesion with changes in gene expression and function that are directed toward removal, resolution, and repair. The gene expression profiles of macrophages change depending upon the nature of the stimulus and time following stimulation. The literature related to macrophage activation, mainly based upon mouse studies, was influenced by the distinction between “classically activated” macrophages stimulated with IFN-γ (IFNG), the major product of Th1 lymphocytes and “alternatively activated” macrophages stimulated with IL-4, the Th2 lymphokine. This led, in turn, to the concept of M1 and M2 polarization (163). In fact, mouse and human macrophages can respond to hundreds of different stimuli and are exposed to complex combinations of stimuli in vivo, which change during disease progression and resolution (164, 165). The combination of IFNG with bacterial LPS is commonly used as a model of mouse M1 polarization in vitro. Signaling pathways initiated by IFNG and LPS interact, but each of these agonists also acts independently, and in vitro models do not correlate well with in vivo macrophage gene expression profiles (166). Based on the M1/M2 paradigm, Guo et al. (167) presented RNA-seq data comparing unstimulated rat BMDM and cells stimulated with IFNG/LPS and IL-4 for 24 h. The ramification of the M1/M2 concept is the identification of markers of each state (168). The rationale behind the use of markers is that their expression on the cell surface is correlated with underlying functions. Unfortunately, this is not the case, especially in the case of M2 markers. Meta-analyses of large gene expression datasets in mouse and human demonstrated that 1) all proposed M2 markers are expressed by subpopulations of tissue-resident macrophages and cannot distinguish them from alternatively activated macrophages, 2) they each have idiosyncratic transcriptional regulation and do not correlate with each other, and 3) there is almost no overlap between proposed M2 markers in mice and humans (164, 165). Nevertheless, there remain numerous studies in rat disease models that refer to M2 markers, notably CD163 and MRC1, largely based upon mouse studies (e.g., Refs. 169, 170), and attempt to infer associated functions.

Not surprisingly, given the selection pressure exerted by pathogens on the innate immune system, inducible gene expression varies substantially among mammalian species. We have published comparative analysis of large animal and rodent BMDM responses to the TLR4 agonist LPS (102). Although there is a shared core set of inducible cytokines, there are also very large differences. Rats share with mice the inducible expression of genes involved in arginine uptake and metabolism to produce NO. In mice, arginase (Arg1) is induced by IL-4 and is also considered an M2 marker (171), but in rat macrophages, Arg1 was massively induced by LPS (102). By contrast, macrophages from humans and most large animals do not take up arginine and do not produce detectable NO; they instead metabolize tryptophan (172). The variation among species involves the extensive gain and loss of cis-acting promoter elements (173).

Lam et al. (174) reported a direct comparison between microglia from rats and mice and regulation of selected genes in response to various stimuli based upon the M1/M2 paradigm. The differences observed were mainly quantitative rather than qualitative and based upon a single time point (24 h). Another group (175) compared the regulations of a focused set of 50 genes in rat and mouse microglia exposed to TGF-β1 in vitro. TGF-β1 is believed to be crucial for the development and maintenance of the unique microglia phenotype (25, 176, 177). Overall, the patterns were similar, but selected mediators, immune receptors, and modulators showed differences between the mouse and rat cells. However, in both cases, the comparison was between outbred rats and the inbred mouse strain C57BL/6, which also differs substantially in macrophage gene expression compared with other mouse strains (178). Indeed, Buscher et al. (179) analyzed the extensive diversity of mouse macrophage responses to LPS using the hybrid mouse diversity panel of 83 inbred strains as a surrogate for human immune variation. There is also extensive diversity at the individual gene level in the human monocyte response to LPS (180). It is very likely that similar variation exists among inbred and outbred rat strains.

Macrophages are a major target for the anti-inflammatory actions of glucocorticoids. Mice also differ from humans in the set of genes induced in macrophages by glucocorticoids, which includes feedback repressors of the response to LPS (181). It is unclear to what extent rats are mouselike, but they are used in preclinical anti-inflammatory drug development. For example, Graversen et al. (182) tested a novel approach in rats to target dexamethasone specifically to macrophages by conjugating it to an anti-CD163 Ab.

Despite the advantages of rats as models, it remains the case that the rat community is small and operates to some extent in parallel with the mouse-focused biomedical research mainstream. The rapid development of genomic technologies is changing the landscape. Multiple companies (e.g., Genoway, Cyagen, Polygene, and Charles River Laboratories) provide practical and cost-effective generation of targeted mutations in the rat germline and already stock tools required for conditional deletions relevant to macrophage biologists (e.g., Cx3cr1-Cre). Over the last 2–3 y, multiple studies have reported macrophage-related and/or inflammatory analysis of targeted mutations. Examples in addition to those already mentioned include Tlr4 (113), Cp (183), Dusp5 (184), Dpp4 (185), Mospd2 (186), and Rnaset2 (187). Many of these studies explore disease models that are more accessible or informative in rats than mice. It is perhaps timely that 2020 was the Chinese year of the rat. Nevertheless, our overview also highlights the many resources (e.g., Abs and models) and knowledge gaps that need to be addressed to the fully use the rat as a model of MPS biology.