The central nervous system (CNS) is considered immune-privileged as the blood–brain barrier (BBB) prevents its overexposure to external pathogens [1]. The CNS homeostasis is also maintained by immune cells, such as tissue-resident macrophages neutralizing noxious factors that are potentially harmful due to neurons’ limited capacity for renewal ^[2][3][2,3]. For a long time, microglia were referred to as the main and only innate immune cells of the CNS [4]. Recently, another population of tissue-resident macrophages was distinguished from the parenchymal macrophages of the CNS [5]. These cells are called CNS border-associated macrophages (BAMs) or CNS-associated macrophages (CAMs), first described as perivascular phagocytes in rats, and are located in the non-parenchymal region of CNS ^[6][7][6,7]. Specifically, the meninges, choroid plexus and perivascular spaces of the brain are populated by BAMs [8].

The single-cell analysis and fate-mapping in transgenic animal models facilitate the detection of microglia and BAMs, shedding light on the cells’ molecular identity without the need for irradiation in the experimental protocols ^{[5][9][10][11][12][13]}[5,9,10,11,12,13]. Indeed, BAMs were clustered in different groups depending on their anatomical positions, revealing each subset’s heterogeneity ^[14][15][14,15]. Specifically, ΒAMs are divided into subdural/leptomeningeal macrophages (sdΜΦ), dural macrophages (dmΜΦ), stromal choroid plexus macrophages (cpΜΦ), choroid epiplexus macrophages (cp^epiΜΦ), also known as Kolmer’s cells, and perivascular macrophages (pvΜΦ) [14]. The subpopulations of BAMs generally vary in morphology, motility, and function. Although BAMs display differences in their transcriptional profiles and dynamics, their functional diversity in each anatomical location has yet to be fully elucidated [16].

The overview of the origin of CNS macrophages has radically evolved during the last decades. In the past, all tissue-specific macrophages were considered to be derived from bone marrow progenitors ^[17][18][19][17,18,19]. However, according to recent data from studies that have utilized new genetic tools, the early perception that tissue-resident macrophages derive solely from adult blood circulating monocytes is no longer prevalent ^{[8][11][20][21][22][23][24]}[8,11,20,21,22,23,24]. Although several studies have investigated the role of microglia in neuroinflammation and neurodegenerative diseases, such as Alzheimer’s disease, Parkinson’s disease and multiple sclerosis, little is known about the contribution of BAMs in these pathophysiological patterns ^{[25][26][27][28][29][30]}[25,26,27,28,29,30]. There are apparent limitations to elucidating the contribution of BAMs in neurological disease; the exact pathophysiological role of BAMs and other CNS-innate immune cells cannot be interpreted by the majority of available tools; therefore, the distinction of BAMs from microglia under disease remains challenging ^[12][31][12,31]. Nevertheless, data obtained via contemporary technologies indicate their potential involvement and miscellaneous phenotypes in neurodegenerative and neuroinflammatory CNS diseases ^{[5][12][32][33][34][35][36]}[5,12,32,33,34,35,36].

2. Origin of BAMs during Embryogenesis and Adulthood

The BAM embryonic origin was first investigated in rodents using bone marrow chimeras and whole-body irradiation, proposing that BAMs are bone marrow-derived ^[37][38][37,38]. In 2016, Goldmann et al., performing fate mapping analysis, observed that BAMs originate from the mouse yolk sac’s early erythro-myeloid progenitors (EMPs) during embryogenesis [8]. A tamoxifen-inducible Runx1^CreERR26^YFPfate-mapping mouse model confirmed that BAMs originate from early EMPs in the yolk sac, which gave rise to two different macrophage populations, namely, CD206⁺ (BAM progenitors) and CD206⁻ (microglial progenitors) without the contribution of fetal liver or definitive hematopoiesis [20]. Interestingly, the mannose receptor C-type 1 (MRC1 or CD206) is a unique marker for BAMs ^[8][12][14][8,12,14]. The expression of Mrc1 is upregulated from E8.5 when the primitive macrophages, originating from EMPs, prepare to invade the embryonic tissues [39]. Recently, Masuda et al. investigated the progenitors of BAMs utilizing single-cell RNA sequencing and fate mapping analysis in the Mrc1^CreERT2 mouse model. Although flow cytometry confirmed the presence of a CD206⁺ subpopulation within the A2 cells (CD45⁺ c-kit⁻ CX₃CR1⁺ cells), meningeal macrophages and microglia were found to originate from common CD206⁺ A2 progenitors in contrast with previous results ^[20][40][20,40]. The pvΜΦ were generated postnatally from sdΜΦ, requiring integrin-signaling and vascular smooth muscle cells (VSMCs) [40]. Regarding the repopulation pattern of BAMs in adulthood, there is a great heterogeneity between BAM clusters; specifically, the sdΜΦ, pvMΦ and cp^epiΜΦ exhibit similar longevity with microglial cells as being self-maintained in the CNS independently from blood monocytes’ contribution ^[8][14][8,14]. The cp^epiΜΦ were solely derived from local SALL1⁺ macrophages [14]. In Ccr2-deficient mice, the number of cpMΦ decreased, revealing their replenishment from Ly6Chi monocytes and shorter turnover [8]. In accordance with these results, Van Hove et al., combining single-cell RNA sequencing with complementary approaches in mice, suggested that dmΜΦ and cpΜΦ were gradually replenished by bone marrow-derived monocytes [14]. As dura mater and choroid plexus stroma are more accessible brain regions than (i) subdural space, (ii) the apical surface of the choroid plexuses, and (iii) brain parenchyma, the tissue permeability may be considered a crucial factor for brain macrophage ontogeny. However, the ablation of BAMs through CSF1R blockade led to the replenishment of cpΜΦ and dmΜΦ via local expansion, indicating their self-renewal capacity, while sdΜΦ presented difficulties in their repopulation [14]. By utilizing the Cx3cr1^CreER:R26^tdTomato fate mapping system in an experimental autoimmune encephalomyelitis (EAE) mouse model, Jordão et al. proposed that BAMs remained stable and locally self-renewed in addition to the recruitment of bone marrow-derived progenitors [12]. In Cx3cr1^gfpCcr2^rfp bone marrow chimeric mice, CD169⁺ BAMs proliferated after ischemia, while a small proportion of BAMs was bone marrow-derived, populating the perivascular and ischemic regions [36]. Both in homeostasis and disease, skull and vertebrae bone marrow constitute a pool of myeloid cells that can invade non-parenchymal and parenchymal CNS regions, transforming into tissue-resident macrophages [41]. A fate-mapping analysis in a mouse model of Alzheimer’s disease (AD) revealed that BAMs are a stable cell population with an unaffected turnover rate and a minimal replenishment from bone marrow-derived cells during this neurodegenerative disease [42]. Summarizing, the origin of BAMs has been extensively studied in the last few years using new genetic tools, e.g., fate mapping analysis. It has been proposed that BAMs originate from early EMPs in the yolk sac during embryogenesis. Although specific BAMs are replenished by peripherally-derived monocytes postnatally, some remain solely derived from the local pool. BAMs have been shown to remain stable and locally self-renewed in both homeostasis and disease. Further investigation is needed to (i) confirm BAM origin, (ii) detect the precise embryonic progenitors of BAMs, especially of the dura mater and choroid plexus macrophages, (iii) determine the timing of each BAM subpopulation’s generation, and (iv) delineate their repopulation pattern.

3. Molecular Drives Orchestrating BAM Development

The transcription factor PU.1 (or SFPI) could be essential for the BAM generation during embryonic development since research has showed that in mice with deletion of the Sfpi1 gene, pvMΦ, sdMΦ, and cpMΦ were ablated [8]. Progenitors of BAMs express the runt-related transcription factor 1 (RUNX1), which regulates the expression of PU.1 during embryogenesis ^[20][43][20,43]. The impairment of PU.1 factor in mice results in a reduced number of A1 (CD45⁺ c-kit^lo CX₃CR1⁻ immature cells) and A2 (CD45⁺ c-kit⁻ CX₃CR1⁺ cells) progenitor cells of the yolk sac, from which both microglial cells and BAMs originate. In contrast, the lack of interferon regulatory factor 8 (IRF8) exclusively decreased the number of A2 cells [44]. Furthermore, the colony-stimulating factor 1 receptor (CSF1R) signaling could be essential for BAM development ^[5][11][14][5,11,14]. In a zebrafish model carrying the panther mutation, a loss-of-function mutation in the fms gene orthologue which encodes CSF1R, primitive macrophages of the yolk sac could not colonize the embryonic tissues [45]. After progenitors’ migration and invasion in the CNS, BAM generation is initiated (Figure 1). The BAMs may be developed independently of transforming growth factor beta receptor (TGF-βR) signaling. In Tgfbr2-deficient mice, no alteration in cell numbers of BAMs occurred, while transforming growth factor beta (TGF-β) is required for the generation of microglial cells ^[20][46][20,46]. Three main brain border regions are filled with BAMs, namely, meninges, choroid plexus, and perivascular spaces. The postnatal expansion of sdMΦ was influenced by IRF8 and MAFB [40]. Indeed, in Irf8-deficient mice, a reduction of sdMΦ was observed [8]. The lack of integrin subunit beta 1 (ITGB1) in mice resulted only in a minor change in the numbers of sdΜΦ [40]. Similarly, the absence of insulin-like growth factor 1 (IGF1R) induces transcriptomic changes in BAMs via its implication in RNA processing, growth, migration and intracellular signaling [47]. The MYB, BATF3, and NR4A1 transcription factors were not necessary for BAM development [8]. Specific molecular cues may also regulate the development of choroid plexus macrophages. Colony stimulating factor 1 (CSF1 or M-CSF), produced by stromal and epithelial cells, is crucial for macrophages’ ontogeny, orchestrating their proliferation and differentiation [48]. CSF1 binds to its receptor, namely, CSF1R, a homodimeric type III receptor tyrosine kinase [49]. Fms-intronic regulatory element (FIRE) is a highly conserved enhancer found in the second intron of the Csf1r gene [50]. In mutant mice with deletion of FIRE, the production and maintenance of cpMΦ were partially impaired [51]. On the contrary, Rojo et al. demonstrated that in FIRE-deficient mice, microglial cells were absent from the brain, whereas BAMs were retained [52]. Interestingly, cpMΦ remained unaltered in a study with Irf8-deficient mice [8], while other research considered IRF8 as a regulator of cpMΦ maturation since the gene ablation suppressed the transcriptional programme of cpMΦ ^[14][53][14,53]. The transcription factor c-MAF, a member of Maf family transcription factors, could be crucial for regulating the pvMΦ transcriptional programme as the deletion of c-Maf in macrophage lineages resulted in the ablation of pvMΦ in the mouse brain [54]. The postnatal expansion of pvΜΦ was also influenced by IRF8 and MAFB [40]. Moreover, VSMCs have a potential role in the distribution of the pvMΦ during development. In Notch3-deficient mice, VSMCs are reduced similarly to the pvMΦ, while the number of sdΜΦ was maintained ^[40][55][40,55]. The distribution of pvMΦ is also controlled by integrin signaling. Talin 1 (Tln1) is an integrin-related gene which encodes a cytoskeletal protein. In Tln1^−/− mice, a significantly lower number of pvMΦ was observed, while microglia and sdΜΦ were not affected in the developing brain, underscoring the impaired vascularization as the cause of pvMΦ reduction [40]. Nevertheless, the absence of integrin subunit beta 1 (ITGB1) in mice resulted only in a minor change in the numbers of pvΜΦ [40]. To recapitulate, the emergence of BAMs is considered a complex process tightly regulated by multiple molecular cues in a similar pattern to oligodendrogenesis and microgliogenesis ^[24][56][24,56]. Although some molecular drivers orchestrating BAM generation have been recently discovered, it remains a largely uncharted territory.

4. BAMs vs. Microglia

Although microglia and BAMs are immune-competent cells of the CNS with common progenitors, their different localization may contribute to variations in their biological roles. The microglial populations’ functions have been reviewed in detail ^[57][58][59][65,66,67]. Concisely, microglial cells are involved in developmental processes, including cell positioning, survival, myelinogenesis, synaptic patterning, and axonal dynamics ^[60][68]. In adult CNS, microglia, as the regulators of acute and chronic immune responses, are implicated in removing pathogens and noxious particles, scavenging cellular debris and synapses, protecting neural tissue, and mediating neurogenesis in CNS injury ^[24][58][61][24,66,69]. The unique localization of BAMs between brain parenchyma and peripheral tissues pinpoint their pivotal role in the immune surveillance of pathological antigens [16]. Their antigen-presenting capacity is attributed to MHC II molecules on some BAM surfaces ^{[12][32][62][63]}[12,32,70,71]. Furthermore, the pvΜΦ and dmΜΦ mainly phagocytose intruding pathogens and any foreign molecule or substance that can be detected in the bloodstream and cerebrospinal fluid ^[64][64][72,72]. The pvΜΦ also appear to regulate the accessibility of brain parenchyma to circulating cells and molecules by increasing the contractility of regional vessels and capillaries or diminishing the BBB permeability ^{[65][66][67][68]}[73,74,75,76]. Interestingly, the latest approaches demonstrate the involvement of BAMs in ensuring a well-balanced metabolic environment for neurons, especially in the course of systemic perturbations ^[69][70][77,78]. Data regarding morphology, motility, and molecular identity of microglia and BAMs are summarized in Table 1. 

Cell Type	Cell Type	Morphology	Morphology	Motility	Motility	Cell-Specific Markers	Cell-Specific Markers
Microglia	Microglia	Ramified in homeostasis; Amoeboid in inflammation	Ramified in homeostasis; Amoeboid in inflammation	Cell bodies with limited-motility but highly dynamic processes in homeostasis; Highly phagocytic in inflammation	Cell bodies with limited-motility but highly dynamic processes in homeostasis; Highly phagocytic in inflammation	SIGLEC-H+, P2RY12+, HEXB+, TMEM119+, ANXA3+, SALL1+	SIGLEC-H+, P2RY12+, HEXB+, TMEM119+, ANXA3+, SALL1+
pvΜΦ	pvΜΦ	Slightly elongated cell bodies	Slightly elongated cell bodies	Non-motile cell bodies with extending and retracting projections through the blood vessel wall in homeostasis; Dendritic-like processes in inflammation	Non-motile cell bodies with extending and retracting projections through the blood vessel wall in homeostasis; Dendritic-like processes in inflammation	CD206+, CD38+, LYVE1+, CD36+, CD163+, CD169+	CD206+, CD38+, LYVE1+, CD36+, CD163+, CD169+
dmΜΦ	dmΜΦ	Elongated; Spindle-shaped cells; Few thick membrane projections; Dendriform	Elongated; Spindle-shaped cells; Few thick membrane projections; Dendriform	Limited motility and highly dynamic protrusions in homeostasis; Extending projections in inflammation	Limited motility and highly dynamic protrusions in homeostasis; Extending projections in inflammation
sdΜΦ	sdΜΦ	Elongated; Amoeboid; Spindle-shaped cells; Few thick membrane projections	Elongated; Amoeboid; Spindle-shaped cells; Few thick membrane projections	Limited motility and highly dynamic protrusions in homeostasis; Extending projections in inflammation	Limited motility and highly dynamic protrusions in homeostasis; Extending projections in inflammation
cpΜΦ	cpΜΦ	Star-like shape	Star-like shape	Unknown	Unknown
cpepiΜΦ	cpepiΜΦ	Round; Bipolar; Stellate	Round; Bipolar; Stellate	Unknown	Unknown