Since the discovery of microglia by del Rio-Hortega and extending into the first half of the 20th century, the origin of tissue phagocytes, including microglia, have long been debated. In 1968, van Furth and Cohn proposed a model known as “the mononuclear phagocyte system” (MPS) asserting that bone marrow (BM) progenitors give rise to circulating monocytes, which then differentiate into macrophages upon extravasation into the body tissues (32–34). Although this theory became dogmatic in immunology textbooks, some experimental evidence challenged its universal validity (35). For example, hematopoietic progenitors generating macrophages were found in the yolk sac before the onset of definitive hematopoiesis in the embryo proper (36–38). Furthermore, yolk sac progenitors exhibited mostly myeloid/erythroid potential compared to the definitive hematopoietic stem cells (HSCs) that emerged in the paraaortic splanchnopleura of the dorsal aorta before migrating into the fetal liver (39–44), suggesting that the two hematopoietic waves are temporally, spatially and ontogenically unrelated (45–47). Nevertheless, BM transplantation studies using irradiation or chemotherapeutic regimen observed infiltration of monocyte-derived macrophages in the host brain parenchyma, generating a long-lasting controversy on the possible BM origin of microglia (48, 49). However, later studies unambiguously demonstrated that BM transplantation strategies induce blood-brain-barrier damage thus eliciting artificial monocytes recruitment (50).

The advent of genetic fate-mapping tools provided definitive proof that microglia originate from erythro-myeloid progenitors (EMPs) emerging in the yolk sac between E7.5-E8.5 ( Figure 1 ), with no significant contribution from fetal liver or BM hematopoiesis (26, 51–54). Unlike HSCs, yolk sac EMPs, a di-potent progenitor also capable of generating primitive erythrocytes, produce macrophages in a MYB-independent manner and without a monocytic intermediate stage (52, 55, 56). Instead, EMPs differentiate into committed myeloid progenitors, macrophage precursors and eventually macrophages upon activation of a myeloid transcriptional program that relies on the transcription factors RUNX1, PU.1 and IRF8 (57, 58). These yolk sac-derived macrophages infiltrate the brain rudiment around E9.5-10.5, after embryonic circulation is established, and serve as the primary source of microglia (26, 59) ( Figure 1 ). Indeed, Ncx1-deficient embryos, which have impaired circulation, lacked brain microglia while preserving yolk sac myelopoiesis (26). Lineage tracing approaches also confirmed that under homeostatic conditions microglia are maintained by slow self-renewal throughout the mouse lifespan with negligible input from BM hematopoiesis; a property shared with other tissue-resident macrophages of embryonic origin, such as Kupffer cells of the liver, and Langerhans cells in the epidermis (30, 50, 51, 60–67).

Microglia survival critically depends on Colony Stimulating Factor 1 Receptor (CSF1R) signaling. Genetic deletion of Csf1r (26, 68, 69) or CSF1R blockade using antagonist antibodies or pharmacological inhibitors (70, 71) leads to complete microglia depletion. Similarly, knock-out mice for the fms-intronic regulatory element (FIRE), a super-enhancer in the second intron of the Csf1r locus, are completely devoid of microglia (72), while preserving most of the peripheral macrophage populations. The importance of CSF1R signaling is further underscored by its role in human neurological diseases. Individuals carrying CSF1R loss-of-function mutations develop a series of neurological conditions such as adult-onset leukoencephalopathy with axonal spheroids and pigmented glia (ALSP), or hereditary diffuse leukoencephalopathy with spheroids (HDLS), brain abnormalities, neurodegeneration, and dysosteosclerosis (BANDDOS) (73–77). Aged FIRE mice also developed ALSP-related features, such as myelin degeneration and disruption of cholesterol metabolism (78). These findings highlight the importance of CSF1R signaling in microglia to maintain CNS homeostasis.

Interestingly, CSF1-deficient (Csf1op/op) mice only exhibit about a 30% reduction in microglia (26), suggesting that other CSF1R ligands may play compensatory roles for microglia survival. Indeed, the identification of the cytokine IL34, the second CSF1R ligand, was a fundamental discovery for the microglia field (79). IL34 and CSF1 are analogous cytokines that share a similar 3D structure despite having little sequence homology, suggesting evolutionary convergence (80, 81). IL34 is highly expressed in neurons, especially in gray matter where CSF1 production is scarce. Indeed, Il34 knock-out mice display a significant microglia deficiency in forebrain regions such as the cortex, hippocampus, and striatum (69, 82–84). Conversely, IL34 is not expressed in the cerebellum, where CSF1 is abundant. Thus, Csf1 deletion in neuroectodermal cells results in microglia depletion from cerebellar gray matter, while forebrain microglia remain unaffected (84–86). Consistently, the inhibition of either CSF1 or IL34 with blocking antibodies depleted microglia in distinct brain regions. Specifically, anti-CSF1 selectively depleted microglia in white matter regions, such as the corpus callosum, hippocampal fimbria, and striatal white matter, whereas anti-IL34 primarily affected gray matter microglia in the cortex and striatum (87). Overall, CSF1 appears critical for microglia survival in white matter, while IL34 mostly maintains microglia in gray matter.

In addition to their spatial specificity, IL34 and CSF1 also differ in their temporal expression patterns. IL34 levels in the brain peak during the perinatal period and decline during adulthood (82). Consistently, the brains of IL34-deficient mice appear normal in E10.5–18.5 embryos (82), while microglia loss starts being observed from postnatal day 2 (83), suggesting that microglia become IL34-dependent after birth only (87). In summary, the regional and temporal specificity of IL34 and CSF1 underscores the non-redundant roles of these cytokines in supporting distinct subpopulations of microglia throughout the brain (69). These findings illustrate how different CNS microenvironments orchestrate heterogeneous communication pathways with microglia.

After reaching the boundaries of the CNS rudiment around E10.5, yolk sac-derived macrophages express high levels of CD206 (Mrc1), while microglia genes are not yet detectable. Differentiation into microglia commences immediately upon their infiltration into the developing brain (88, 89). Once in the parenchyma, embryonic microglia undergo a specification program sculpting their homeostatic phenotype throughout the embryonic and early post-natal period. Along this process, microglia downregulate genes like Mrc1, Lyz2 and Apoe that are highly expressed in extra-parenchymal macrophages and gradually upregulate microglia signature genes like Tmem119, P2ry12, Hexb, Slc2a5, Sparc and Selplg, reaching complete maturity around the third post-natal week (90–94).

Microglia specification is orchestrated by a repertoire of transcription factors like Mafb, Irf8, Smad4 and Sall1 (91, 95–100), as well as by stimuli from the brain parenchymal environment, such as the transforming growth factor beta (TGFβ) signaling (101). Multiple studies have demonstrated that TGFβ crucially instructs the microglia developmental program in macrophages infiltrating the brain parenchyma. Deletion of critical components of the TGFβ signaling cascade, such as the cytokine Tgfb1 (101), the latent-TGFβ binding-proteins Lrrc33/Nrros (102, 103), the integrin activating latent-TGFβ (104), the TGFβ receptors (54, 95, 105, 106), or the downstream transcription factor Smad4 (98, 99) consistently impaired microglia development, holding microglia in an undifferentiated state. More specifically, it has been shown that TGFBR2- or SMAD4-deficient microglia failed to upregulate their homeostatic signature genes, while maintaining expression of genes enriched in CD206+ microglia precursors (Mrc1, Lyz2, Apoe, Pf4, Ms4a7) that accumulate at the borders of the embryonic brain (54, 98). Such developmental failure is phenocopied by the deletion of the microglia-specific Sall1 super-enhancer activated by SMAD4 (99). Transcriptomic and epigenetic analyses demonstrated that SMAD4 and SALL1 cooperate to activate the expression of microglia genes, while repressing genes of extra-parenchymal macrophages (99). Interestingly, microglia are the main source of TGFβ1 in the brain (78), suggesting that their maturation is sustained through an autocrine feedforward loop. Indeed, Tgfβ1 deletion in microglia recapitulated the developmental failure observed after conditional deletion of Tgfbr2 or Smad4 (107, 108). Importantly, postnatal deletion of TGFβ1, TGFβR2 or SMAD4 caused microglia dedifferentiation and reestablished their immature phenotype (99, 107, 108). This data suggests that TGFβ signaling in microglia is not only required to activate the microglia specification program in the embryonic brain, but also to maintain their mature phenotype after birth.

Intriguingly, a similar microglia phenotype has been described in mice with Integrin-β8 (Itgβ8) deletion in neuroectodermal brain cells using Nestin-Cre or Emx1-Cre (104, 107), targeting neuronal stem cells (or radial glia) during brain development. ITGβ8 forms a heterodimer with Integrin-αV (commonly αVβ8) binding the latent-TGFβ complex produced by microglia themselves. This interaction unleashes active TGFβ which engages TGFβ-receptor on microglia in an autocrine manner (109). Thus, radial glia lining the ventricle walls in the embryonic CNS activate the TGFβ signaling in microglia precursors while infiltrating the brain parenchyma during development (107).

Despite being largely consistent, these studies occasionally reported distinct behavioral phenotypes in mice with abrogation of TGFβ signaling in microglia, ranging from neurodevelopmental alterations and neuromotor dysfunctions (104, 107), white matter disease (78, 104) or learning and memory deficits (98, 108). Different behavioral phenotypes may stem from the target gene being deleted. TGFβ or TGFβ-receptors deletion simultaneously abrogates canonical (SMAD4-dependent) and non-canonical (SMAD4-independent) TGFβ pathways. Conversely, deletion of SMAD4 impairs the canonical pathway only, while preserving non-canonical pathways (110). Further studies are required to better understand the exact implications of the different TGFβ pathways in microglia.

In summary, microglia precursors produce latent-TGFβ1, which is activated by radial glia during infiltration into the brain parenchyma. Active TGFβ1 engages TGFβ-receptors on microglia precursors thus promoting a self-sustained SMAD4/SALL1 transcriptional program leading their stepwise maturation throughout the embryonic and early post-natal period.

Like microglia, virtually all BAM subsets originate from yolk-sac hematopoiesis, with minor contributions from fetal liver or BM monocytes (54, 120). Although microglia and BAMs disseminate the developing CNS at the same time, they undertake completely distinct developmental paths (54, 88). While TGFβ signaling shapes the microglia homeostatic signature throughout the embryonic and early postnatal period, BAMs develop independently from TGFβ (54, 98) and acquire a different repertoire of marker genes (Mrc1, Lyz2, Apoe, Pf4, F13a1, Ms4a7, Cd163, Lyve1, Folr2, Cd209f) (98, 112). Divergent specifications programs in microglia and BAMs are likely controlled by different transcription factors. For example, microglia show stronger transcriptional signatures associated with of IRF8, MAFB, SMAD2/3 and SALL1 transcriptional networks, while BAMs are more enriched for STAT1/2, IRF7, SPIC, CEBP and cMAF (98, 100, 112, 121, 122). Interestingly, both SMAD4 and MAFB are not required for the development of BAMs (98, 112, 121, 122), suggesting that these transcription factors are specifically induced by the parenchymal compartment. The mechanism through which the extra-parenchymal CNS niches instruct the transcriptional trajectory in BAM is still unclear.

The earliest populations colonizing the CNS border regions are MnMΦ and CpMΦ between E10.5-12.5 (88, 116). The perivascular spaces begin to open alongside the large-medium caliber blood vessels of the cerebral cortex early postnatally. Between the first and the second postnatal weeks, MnMΦ migrate along the pia mater penetrating the upper cortical layers, thus generating PvMΦ, which locally expand by cell division (88). Migration of MnMΦ in the perivascular compartment is controlled by integrins and is facilitated by the presence of vascular smooth muscle cells (88); however, the exact mechanism remains to be fully elucidated. In the postnatal brain, PvMΦ are largely MHCII low/negative, expresses high levels of CD206, FOLR2, LYVE1 and CD38, and retain embryonic origin and self-renewal capacity (98, 118, 123). In the leptomeninges, however, at least two distinct populations of MnMΦ can be observed (98, 118, 123). One subset expressing CD206, FOLR2, LYVE1 and CD38 is phenotypically and ontogenically equivalent to PvMΦ. Conversely, a second subset dim or negative for these markers but positive for MHCII is mostly HSCs-derived and is established postnatally (98, 112, 118). A similar phenotypic heterogeneity has been observed in DmMΦ (112, 124). Like other BAM populations, DmMΦ initially originate from yolk sac progenitors, however, they are partially replaced by monocyte-derived macrophages afterbirth. Nevertheless, some DmMΦ of embryonic origin persist throughout adulthood (124). Similar to other CNS border regions, DmMΦ comprise two main populations: one subset expressing high levels of CD206, FOLR and CD301a, and another subset expressing MHCII and low levels of CD206. While CD206hi DmMΦ are initially yolk sac-derived, MHCII+ DmMΦ expand significantly during the first postnatal weeks. Monocyte-derived macrophages contribute to both subsets during development and adulthood, with a stronger input to the MHCII+ population. Indeed, the MHCII+ DmMΦ appear reduced in Ccr2-/- mice, where monocytes are retained in the BM (112, 124).

CpMΦ encompass highly heterogeneous macrophage populations with different phenotypes, origin and spatial distribution (112). The choroid plexus is a ventricular organ that produces cerebrospinal fluid (CSF). It consists of a monolayer of ciliated epithelial cells contiguous with the ventricular ependyma, supported by a stroma containing mesenchymal cells, connective tissue and fenestrated blood vessels. The choroid plexus begins to form around E10.5, to become fully developed at birth (115, 125). Macrophages are highly abundant in this region and generally fall into two main categories: stromal CpMΦ and epiplexus Kolmer cells located on the epithelial layer facing the ventricle’s lumen. Both populations of CpMΦ are established early during embryonic development; however, stromal CpMΦ are gradually replaced by monocyte-derived macrophages during aging (112, 116). Like MnMΦ, stromal CpMΦ can be MHCII+ or MHCII−, though it remains to be determined whether these two subsets represent developmentally and biologically distinct populations rather than transitory activations states (112). Stromal CpMΦ share several marker genes with PvMΦ/MnMΦ (SDMs or PBMs), however there are notable differences (112). PvMΦ/MnMΦ express high levels of Lyve1 and Cd209f, which are not detectable in CpMΦ. Conversely, CpMΦ express the inhibitory receptor Lilra5. These differences may reflect highly specialized functions tailored to the local environment. For example, PvMΦ/MnMΦ might be more prone to interact with vascular components such as hyaluronan (LYVE1 ligand), while CpMΦ are more frequently exposed to immunological challenges and may play a role in the production of CSF molecules.

Epiplexus Kolmer cells are an elusive subset of CpMΦ due to the lack of effective staining techniques for histology or flow cytometry. Consequentially, this CpMΦ population remains highly understudied. Nevertheless, one study reported that Kolmer cells co-express BAM genes (such as Lyz2, Apoe, but low levels of Mrc1), along with several microglia signature genes (Trem2, Hexb, Tmem119, P2ry12, Siglech, Slc2a5, Sparc and Sall1). Like microglia, Kolmer cells appeared long-lived and did not receive significant input from blood monocytes (112). The optimization of staining, sorting and genetic targeting strategies for Kolmer cells is warranted to study this population in greater detail.

Microglia are known to play several essential functions to maintain CNS homeostasis. For example, microglia regulate the brain’s physiology by shaping neuronal connectivity (71, 84, 126, 127), remove cellular debris, apoptotic corpses and toxic protein aggregates during brain development, aging and disease (128–131), promote homeostatic myelination and myelin repair after injury (78, 132, 133), and activate healing processes after tissue damage (134–136).

In comparison to microglia, our understanding of the BAMs’ immunobiology remains limited. However, recent seminal studies provided some insight into the functional roles of BAMs under steady state or disease. For example, PvMΦ have been shown to regulate CSF flow through the meninges by modulating extracellular matrix deposition along the CNS endothelial cells (119). Additionally, PvMΦ have been identified as key contributors to neurovascular dysfunction and impaired cerebral blood flow in mice expressing human APOE4 (137). These findings suggest that PvMΦ play key roles in regulating the physiology of brain blood vessels.

During immune responses, BAMs represent the first line of defense against meningeal infections caused by viruses (138–140), bacteria (141, 142), and parasites (143). These studies have demonstrated that BAMs support the recruitment of monocytes and effector T cells, facilitating pathogen clearance. Due to their strategic location at the brain borders, BAMs are able to present antigens to cognate CD4 T cells during CNS autoimmunity or neurodegeneration (144–147). Nevertheless, several studies showed that infiltration and activation of meningeal T cells occurs independently of MHCII expression in BAMs or microglia (111, 124, 139, 144). Instead, BAMs appear to play a major role in cytokines and chemokines secretion through type-I interferon signaling (111, 139). Additionally, DmMΦ have been shown to perform efferocytosis of apoptotic neutrophils in the dura mater during neuroinflammation (124). These findings suggest that BAMs may play a role both in the onset and resolution of inflammatory dynamics within the CNS. Consistently, a study on bacterial meningitis showed that CpMΦ contribute to epithelial barrier repair following neuroinflammation, raising the hypothesis that other BAM subsets, such as MnMΦ and PvMΦ, might also possess similar tissue-healing properties (142).

Taken together, BAMs are emerging as key regulators of vascular dynamics, immune surveillance, and tissue repair within the CNS. Nevertheless, our understanding of these mechanisms remains limited. Further investigations are warranted to explore BAMs as potential therapeutic targets for vascular and neuroinflammatory conditions.

Several studies demonstrated that the CNS environment crucially instructs the microglial phenotype by shaping the repertoire of tissue-specific enhancers and transcription factors (148–150). To study these mechanisms in vivo, strategies of microglia depletion/repopulation have been employed. Head irradiation induces blood-brain-barrier damage and, to some extent, monocyte infiltration into the brain (50). However, a persistent microglia ablation elicits massive monocyte infiltration into the CNS, even in absence of blood-brain-barrier preconditioning. Multiple groups have utilized genetic models of microglia depletion to promote monocyte infiltration into the brain, thus replacing resident microglia with monocyte-derived macrophages (151–156). Consistently across these studies, monocyte-derived macrophages acquire a phenotype and epigenetic profile largely overlapping with that of endogenous microglia, suggesting that local tissue factors within the brain parenchyma induce the differentiation of monocytes into microglia-like cells. Nonetheless, monocyte-derived microglia-like cells still retain transcriptional and epigenetic traits reminiscent of their BM origin. For example, genes like Apoe, Lyz2, Clec12a or the surface marker F4/80 were consistently upregulated in microglia-like cells compared to endogenous microglia, while Sall1 expression was barely detectable (152, 154, 155). Furthermore, monocyte-derived microglia-like cells mounted more pronounced inflammatory responses during immunological challenges than did endogenous microglia (153, 155).

The importance of the CNS environment in maintaining the microglial phenotype has been widely demonstrated through transplantation experiments. Primary microglia lose their homeostatic signature within a few hours under cell culture conditions. Genes such as Tmem119, P2ry12, Cx3cr1, Sparc and Sall1 are rapidly downregulated when microglia are exposed to the cell culture environment (157, 158). Conversely, in vitro cultured microglia from both mouse and human regain their homeostatic phenotype once transplanted into the mouse brain in vivo or partially regain them when cultured within brain organoids (157, 159–165). These findings highlight the crucial roles of CNS-derived factors in sustaining the microglial transcriptional program. The nature and cellular sources of these factors are currently under investigation.

Our understanding of whether and how the extra-parenchymal environment shapes the phenotype and functions of BAMs remains limited. Different groups showed that pharmacological depletion of BAMs elicits the recruitment of monocyte-derived macrophages into the brain border niches (112, 166). Similarly, xenotransplantation experiments of human embryonic HSCs into a host mouse brain revealed that these cells have the potential to locally differentiate into BAM-like cells (160, 161). However, it remains unclear whether these BAM surrogates are phenotypically and functionally diverse compared to endogenous BAMs. Nevertheless, identification of macrophage populations phenotypically similar to BAMs in peripheral organs such as lungs, hear, liver, kidney and adipose tissue (121, 123, 167) suggest a conserved transcriptional program presumably instructed by similar environmental cues. Deeper transcriptional and epigenetic analyses are warranted to determine how ontogeny and environmental niches shape the phenotype and epigenetic landscape of BAMs across compartments and conditions.

Epigenetic analyses of BAMs have been hampered by the large number of cells required for the analysis of chromatin accessibility, histone modifications and long-range chromatin interactions. However, bulk low-input and single-cell epigenomic techniques are now becoming increasingly reliable and accessible (168–172). Moreover, techniques to generate iPSC-derived BAMs in vitro or within brain organoids require further optimization and benchmarking. The ability to maintain and genetically manipulate BAMs in culture will offer the opportunity to perform high-throughput single-cell analyses such as CRISPR screens (173). We optimistically expect that in the near future these technologies will be applied to BAMs as well.

Multiple studies have identified a broad phenotypic overlap between human and mouse microglia under homeostatic conditions (174). For example, human microglia express high levels of CX3CR1, P2RY12, TMEM119, TREM2, MERTK, CST3, NAV3, and SELPLG (94, 175–178). This similarity extends to transcription factor repertoire, as both human and mouse microglia express PU.1, IRF8, MAFB, MEF2C, and SALL1 (158, 175), indicating conserved transcriptional programs across species. Transcriptional profiling of microglia from evolutionarily distant species (ranging from zebrafish, chicken, rodents, monkeys, and human) revealed conserved core signature genes, including Csf1r, P2ry12, Spi1, Irf8, and Tgfbr2 (179), suggesting that such gene expression profile has been positively selected throughout evolution. Nevertheless, human microglia are enriched for several immune-related genes sets which are normally not expressed in mouse microglia under steady state. Genes highly expressed in human microglia but not in mouse include Toll-like receptors, antigen presentation machinery (CD58, HLA-DR, CD74, ERAP2, CIITA), interferon-induced genes (IFI16), chemokines (CCL4), antimicrobial genes (GNLY), complement components (C3), extracellular matrix adhesion molecules (SPP1) (94, 158, 163, 175, 179). Thus, human microglia seem to be prone to immune surveillance and antigen presentation compared to their mouse counterparts. Like microglia, the BAM signature appears highly conserved across species and several marker genes, such as MRC1, F13A1, LYVE1, STAB1, CD163, CD36, MS4A7, CYBB, TGFBI, RBPJ, COLEC12 are shared between mouse and human (178, 180).

Similar to mice, human microglia and BAMs allegedly originate from yolk sac myeloid progenitors. Recent studies have assessed the transcriptional trajectories of human hematopoiesis throughout embryonic and fetal development (181–184), providing evidence that yolk sac-derived macrophages differentiate into various macrophage lineages across multiple embryonic tissues, including the brain (182). Yolk sac-derived myeloid precursors enter the human embryonic brain through the leptomeninges and the ventricular lumen starting from the 4th gestational week (GW4) (185). Upon infiltrating of the CNS parenchyma, these cells differentiate into microglia, which proliferate and migrate throughout the entire brain. The first peak of microglia expansion occurs around GW9 during the transition from embryonic to fetal stage, followed by a contraction due to apoptosis. A second surge in microglial proliferation occurs during the first postnatal year, reaching a homeostatic microglia density during childhood (185).

Upon seeding the developing human brain, the transcriptional and epigenetic profiles of fetal microglia undergo stepwise maturation until the early postnatal period (186). In the early fetal brain (GW8-9) microglia are enriched for phagocytic pathways and hypoxia-related genes, while immune-related genes (i.e. HLA-DR, C3, ITGAX) are upregulated at later fetal stages (GW17-18) or postnatally (163, 178, 186). This phenotypic maturation is underpinned by significant changes in the transcription factor activity. Transcription factors such as MiTF/TFE, MAF-family, MEF2A, ELF2, E2F2, SOX and SP1 are primarily active in fetal microglia and regulate chemotaxis, cytokine production, adhesion to extracellular matrix, cell cycle and chromatin remodeling. Activity of IRFs, PRDM1, STAT2, KLFs, ETS, MAFB and SMADs increases in postnatal microglia, thus shaping the enrichment for innate immune pathways (186, 187). Likewise, human BAMs undergo a substantial phenotypic maturation across the fetal period gaining expression of HLA, APOE, CD44 and CD163 (178). To date, the transcription factor network orchestrating the development of human BAMs remains poorly investigated.

As described in mice (65–67), human microglia are sustained by a continuous process of apoptosis and cell division, with a daily turnover rate of approximately 0.08%, accounting for about 28% microglia renewal every year, and a medial lifespan of 4.2 years for each microglial cell (188). Assessing the contribution of blood monocytes to the pool of brain macrophages is technically challenging. One study analyzed the frequency of somatic mutations in monocytes and microglia in elderly subjects with clonal hematopoiesis, a condition causing a benign expansion of clonal monocytes carrying mutations in transcriptional and epigenetic regulators. In principle, mutant microglia allegedly derive from monocytes and not from yolk sac progenitors. This study found that 30% to 95% of microglia carried somatic mutations found in monocytes, suggesting BM origin of these cells. Additionally, the clonal size of blood cells strongly correlated with the frequency of mutant microglia, suggesting that monocytes could be a main source of brain macrophages in the aged brain (189). It should however be noted that subjects enrolled in this study were around 80 years old, and age-related medical conditions may alter the balance between yolk sac and BM-derived microglia in the brain. Furthermore, it cannot be excluded that monocytes-derived macrophages carrying somatic mutations may locally undergo clonal expansion at the expense of non-mutant microglia. Another recent study measured the engraftment of monocyte-derived macrophages in the brain of female patients receiving sex-mismatched BM transplant and used the Y chromosome for lineage tracing of BM-derived cells (178). This analysis revealed that about 20% of microglia are BM-derived. Additionally, significant variability in monocyte infiltration was found across different CNS compartments, with a much faster engraftment rate in the choroid plexus as compared to the brain parenchyma, perivascular space and leptomeninges. At the moment, the effect of chemotherapeutic regimen on the turnover of brain macrophages is unknown. The results emerging from these studies suggest that BM contributes to the population of brain macrophages more in humans than it does in mice. Future studies may perform lineage tracing of human brain macrophages across ages and disease conditions using single-cell whole-genome sequencing and random somatic mutations as fate-mapping barcodes (190, 191).

A recent study, currently in preprint, mapped somatic genetic variants at single-cell resolution in microglia and blood to trace monocyte-derived microglia-like cells in the brains of healthy donors and Alzheimer’s patients (192). This study determined that somatic mutations differ substantially between blood and brain, indicating a minor contribution (<5%) of circulating monocytes to the pool of resident microglia. Although still in preprint, this study demonstrates that genome sequencing at the single-cell level is a feasible method to assess the fate of microglia and monocyte-derived macrophages across ages and disease settings.

In summary, the core signature of microglia and BAMs exhibits remarkable similarities in mouse and human, reflecting conserved ontogeny and developmental programs between species.