Brain border-associated macrophages: common denominators in infection, aging, and Alzheimer’s disease?

Abstract

Mammalian brain border-associated macrophages (BAMs) are strategically positioned to support vital properties and processes, e.g., the composition of the brain’s perivascular extracellular matrix and cerebrospinal fluid flow via the glymphatic pathway. BAMs also effectively restrict the spread of infectious microbes into the brain. However, while fighting infections, BAMs sustain long-term transcriptomic changes and can be replaced by inflammatory monocytes, potentially leading to a gradual loss of their beneficial homeostatic functions. We hypothesize that by expediting the deterioration of BAMs, multiple infection episodes might be associated with accelerated brain aging and the putative development of neurodegenerative diseases. Our viewpoint is supported by recent studies suggesting that rejuvenating aged BAMs, and counterbalancing their detrimental inflammatory signatures during infections, might hold promise in treating aging-related neurological disorders including Alzheimer’s disease.

BAMs: newly identified players in neuroimmunology

Holistically, the immune landscape of the mammalian central nervous system (CNS) and its border tissues comprises virtually all cell types in any ordinary “non-immune privileged” peripheral organ. A vast and increasingly convincing body of evidence favors the idea that well-tuned CNS innate immunity is vital to foster an environment that nurtures neural function and restrains the emergence of deleterious immune responses that, in extreme circumstances, can lead to blood-brain barrier disruption, edema, and neurodegeneration (reviewed in [1–4]). Collectively, microglia and border-associated macrophages (BAMs) are the main immune cell types in the homeostatic CNS in mice and humans, and constantly survey the brain parenchyma and its border tissues for endogenous and exogenous antigens ([5]; reviewed in [6–9]). Antigens can arise from, but are not exclusive to, infectious agents, neural cell molecular content and debris, lipid species from highly myelinated white matter regions, or toxic misfolded proteins such as amyloid beta (Aβ), tau, and alpha synuclein (αSyn) (reviewed in [8,10–13]). The brain BAMs are found in mice and humans, and encompass perivascular macrophages (PVMs), as well as macrophages embedded in the choroid plexus (intra-stroma and apical side of the epithelium) and meningeal layers ([5]; reviewed in [8,9]). These professional phagocytes can acquire different activation states, secrete inflammatory cytokines and chemokines and their strategic positioning at border tissues (in many cases devoid of blood-brain barrier) facilitates the crosstalk between the periphery and the brain, for instance during infections [14–16]. Overall, BAMs represent newly identified players in the field of neuroimmunology; a deeper comprehension of these cells and their altered states could lead to a better understanding of the pathophysiology of neurodegenerative diseases, ideally offer exciting new therapeutic opportunities. Here, we discuss the ontogeny and properties of brain BAMs at steady state, their unique functions, and alterations upon infection. We also examine recent experimental data on the potential role of BAMs in neurodegenerative diseases. We hypothesize that several rounds of infection/inflammation could fundamentally alter the properties of BAMs, setting a platform for pathological aging and the development of Alzheimer’s disease (AD).

Macrophages inhabit the murine and human brain border niches at homeostasis

Origins and diversity of BAMs

In adult mammals, including mice and humans, a specific type of BAM called perivascular macrophage ( PVM) can be found in the perivascular spaces of the brain and spinal cord, surrounding the basement membrane of blood vessels (mainly arteries and arterioles) [17]. The remaining BAMs inhabit the choroid plexus and the three anatomical layers that surround the mammalian CNS parenchyma, known as the meninges. The meninges consist of the dura mater, which is located beneath the skull and contains lymphatic vessels and fenestrated blood vessels, followed by the arachnoid mater, and finally the pia mater, a thinner tissue that adheres to the CNS parenchyma surface [18]. Meningeal BAMs are largely distributed between the dura and pia, two meningeal layers that are quite distinct when it comes to resident immune cell compositions [5,7,19,20].

In mice, macrophages initially enter the embryo proper from day E9.5, following the stepwise establishment of embryonic circulation; brain PVMs and pial BAMs initially originate from yolk sac-generated c-Myb-independent, c-KIT⁺ erythromyeloid progenitors during embryogenesis [21–24]. In the dura and choroid plexus, BAMs derived from yolk-sac progenitors undergo partial replacement by monocyte-derived macrophages in adulthood, which preferentially differentiate into major histocompatibility complex type-II (MHC-II)⁺ BAMs [25,26]. In humans, a cluster with the transcriptomic signature of monocyte-derived macrophages expressing cluster of differentiation 206 (CD206) is detected in the dura mater and choroid plexus, aligning with a constant input of myeloid cells into those compartments [5]. Surprisingly, labeling of skull bone marrow (SBM) cells (e.g. using skull grafts from adult fluorescent reporter mice) showed that some myeloid cells in the dura originated from hematopoietic progenitor cell reservoirs in the adjacent calvaria, rather than from classical circulatory monocytes [14,27–30] (see Box 1). This suggests that BAMs with a higher turnover, like dural BAMs, might be replenished by skull-derived myeloid cells and have different properties when compared to blood monocyte-derived BAMs [31].

Box 1. Replenishment of the meningeal myeloid cell pool by skull bone marrow (SBM) precursors in mice.

Immune cells can be mobilized to the meninges not only from blood vessels but also from the adjacent SBM. Indeed, direct channels connecting the SBM to the meninges were discovered recently in mice and humans, facilitating the communication between the brain, meninges, and skull [28,29,90]. Using mouse models and several techniques to distinguish blood-from skull-derived leukocytes, such as parabiosis and whole skull grafts from fluorescent reporter mouse lines, the SBM was shown to serve as a source of meningeal and brain immune cells (e.g., monocytes and neutrophils) at steady-state [30]. This was also concluded from the similar transcriptomic profiles of myeloid cells residing in the SBM and the meningeal dura, suggesting a shared origin [31]. In mice, increased cell migration from the SBM into the meninges has been observed in models of stroke [29], experimental autoimmune encephalomyelitis [30] and certain CNS cancer models [91]. Importantly, this communication between the SBM and meninges is bidirectional. Following intracerebroventricular injection of Streptococcus pneumoniae in mice, the bacteria were detected in the meninges and reached the SBM through vascular channels, subsequently boosting cranial hematopoiesis [27]. Similarly, following spinal cord injury in mice, molecules secreted by brain cells into the CSF diffused from the subarachnoid space into the dura via arachnoid cuffs, and from there, into the SBM [28,92]. This CNS-to-SBM signaling culminates in a the mobilization of bone marrow myeloid cells, as evidenced by the exacerbated recruitment of dural monocytes after CSF injection from injured mice into the cisterna magna of healthy recipient mice [28]. Overall, based on the bidirectional communication between the brain and the skull, it seems plausible that some BAM populations, once thought to be replenished solely from blood monocytes [93], are instead replaced by monocytes coming from the SBM. However, the exact contributions of blood or SBM myeloid precursors to the pool of brain BAMs might be context-dependent and needs to be further elucidated. This is relevant, especially in the context of aging-associated neurodegenerative diseases, because BAM precursors from these two different sources might not have the same transcriptomic profiles, migration kinetics, and properties; thus, blood- or SBM-derived myeloid precursors might not ensure the homeostatic functions of BAMs upon differentiation and engraftment into the brain border tissues [31]. We hypothesize that, in AD, brain-derived inflammatory signals might reach the SBM via the CSF, and modify the balance between BAM replacement by SBM or blood myeloid precursors, leading to an accelerated BAM dysfunction [27–31,90,91]. Modulating this new source of brain myeloid cells could affect the properties of BAMs in health and disease.

In summary, in mice and humans alike, BAMs are heterogeneous, which is likely to impact their function, and several aspects of brain physiology and pathology. Even though the impact of this heterogeneity is not yet fully understood, recent studies have begun to elucidate some of the essential functions of BAMs.

Homeostatic roles of BAMs

Several studies have now tried to understand the roles of BAMs in modulating brain homeostasis. For instance, in mice with reduced meningeal T cells due to treatment with FTY720 (inhibiting immune cell exit from lymph nodes), meningeal dural BAMs exhibited a pro-inflammatory phenotype characterized by increased TNF production, correlating with impaired cognitive performance relative to untreated mice [32]. Similar results were observed in severe combined immunodeficiency (SCID) mice reconstituted with IL-4-deficient T cells, but not with T cells expressing IL-4, suggesting a role for T cell-derived IL-4 in shaping brain BAM phenotypes [32] (Figure 1). In addition, mice lacking macrophages, such as those deficient for the colony-stimulating factor-1 receptor (Csf1r) gene, exhibited enlarged brain ventricles and hydrocephaly compared with wild type mice, suggesting a role for CNS macrophages in controlling the production and/or drainage of cerebrospinal fluid (CSF) [33]. Also, PVMs can regulate CSF flow dynamics, since their depletion in mice through either pharmacological (intra-cisterna magna injection of clodronate) or genetic (Lyve1-Cre:Csf1r^fl/fl mice) methods resulted in accumulated extracellular matrix proteins, which obstructed the access of CSF to perivascular spaces [34]. This led to impaired CNS perfusion and cleansing via the glymphatic system [34]. Of note, clodronate is not necessarily specific for BAMs and there might be potential off-target effects on other CNS-associated myeloid subsets that should be further investigated. As a final example, due to the close proximity of BAMs to the CSF-filled perivascular and ventricular spaces [8,15,16,27,34,35], we speculate that BAMs might also modulate brain fluid composition and renewal with active secretion of signaling molecules (e.g., cytokines). The secretory function of different brain BAMs and their impact on fluid homeostasis is a hot topic that warrants further investigation.

Figure 1. Schematic view of BAMs and their presumed roles before, during, and after different infections in mice.

The three boxes represent border regions of the brain (e.g., dural meninges) at 3 timepoints (before, during, after infection). At steady-state (left box), BAMs are polarized through IL-4 and maintain brain function [32,34]. A positive impact on the brain is noted by the “red cross” sign. In an infectious context (middle box), BAMs play a key role in fighting pathogens and protecting the brain [14–16,36]. Red lines represent the defense provided by BAMs in each infectious context: upon S. agalactiae infection, BAMs reduce bacterial load but this is dampened by pain signals coming from CGRP⁺ nociceptors activated by the bacteria; upon LCMV acute neurotropic infection, BAMs strongly reduce viral load and require IFN-I for their activation; upon T. brucei infection, BAMs reduce parasite load but cannot block its dissemination into the brain. A thicker red line represents a stronger BAM defense role. After pathogen clearance (right box), inflamed BAMs maintain long-lasting transcriptomic and phenotypic alterations [16]. In addition, monocytes engrafting the meninges may affect brain function in the long run [26]. Monocytes can arrive via blood or hypothetically via the skull bone marrow (indicated as a question mark). Impact of post-infectious BAMs on brain function is still uncertain and indicated as a question mark. BAMs, border-associated macrophages; CGRP, calcitonin gene-related peptide; IFN-I, interferon type I; IL-4, interleukin-4; LCMV, lymphocytic choriomeningitis virus; S. agalactiae, Streptococcus agalactiae; SBM, skull bone marrow; T. brucei, Trypanosoma brucei. Figure created with Biorender.com.

Overall, even though largely speculative at this point, recent data suggest that deviations from BAMs’ physiological properties (e.g., the acquisition of a persistent inflammatory signature following infection) could lead to defects in CSF composition and glymphatic flow, accumulation of toxic proteins in the CSF and parenchyma, and impaired cognitive function, all of which might contribute to the development of neurodegenerative diseases.

BAMs play a crucial role during infection but retain an inflammatory scar

Brain BAMs are highly responsive to infectious agents. Following peripheral infection of mice with lymphocytic choriomeningitis virus (LCMV, Armstrong acute neurotropic strain), BAMs blocked viral spread into the CNS and prevented fatal meningitis [15,36] (Figure 1). Specifically, one study showed that rendering BAMs incapable of responding to interferon type I (IFN-I) signaling (using Cd163-Cre:Ifnar^fl/fl mice) culminated in fatal meningitis [15], suggesting that BAMs’ activation through IFN-I was required for their antiviral function. It is important to mention that the referred intervention was likely affecting other body macrophages expressing CD163 and were not limited to brain BAMs. In addition, following intracranial injection of the same LCMV strain, dural BAMs were infected and their numbers reduced. Consequently, monocyte-derived macrophages (tracked using Cx3cr1-CreERT2:R26-Yfp reporter mice) engrafted the meningeal dural tissue and partially replaced resident BAMs [26]. Compared to the original dural BAMs, the engrafted macrophages displayed a higher inflammatory signature (assessed via sustained inflammatory-stimulated gene 15 (Isg15) mRNA production following LPS exposure), even in the presence of the anti-inflammatory molecule acetylcholine [26]. This suggested a long-term dysregulation of BAMs after viral infection, which could have implications for brain function and future neuroinflammatory responses [26].

Recent studies have also investigated the response of murine BAMs to peripheral infection by Trypanosoma brucei, the parasitic agent responsible for sleeping sickness [16]. Following infection with Trypanosoma brucei, Cx3cr1-CreERT2:R26-DTR mice were treated with diphtheria toxin (DT) at different timepoints to deplete microglia and BAMs, but also other peripheral resident myeloid cells. On the one hand, at 2 weeks post-infection, mice depleted of resident myeloid cells including BAMs displayed increased parasite load in the dura. On the other hand, if brain myeloid cell depletion was induced at 4 weeks post-infection, mice displayed decreased leukocyte counts and cytokine concentrations (e.g., CCL5) in the CSF, compared to brain myeloid-intact control mice [16] (Figure 1). Single-cell RNA sequencing (scRNA-seq) of immune cells indicated that the transcriptomic changes were more pronounced in BAMs [e.g., sustained expression of cytokines and chemokines such as Ccl5] when compared to microglia, even 9 weeks after Trypanosoma brucei clearance [16]. Altogether, these data suggest that activated brain myeloid cells (microglia and/or BAMs) can harbor an antimicrobial effect yet sustain an inflammatory phenotype even after pathogen clearance.

Of note, as observed during viral or parasitic infections, the depletion of BAMs by intracisternal injection of clodronate-laden liposomes after intravenous infection with Streptococcus agalactiae has also led to decreased immune cell recruitment and increased bacterial load in the dura compared with untreated mice [14] (Figure 1).

Overall, in the mouse infection models discussed (viral, parasitic, or bacterial), fighting the infection came at a cost. Indeed, BAMs, unlike microglia, retained a long-lasting scar of the inflammatory insult even long after pathogen clearance. We propose that multiple rounds of infection, alongside other inflammatory insults, might cause long-lasting reprogramming of brain BAMs, which might accelerate their functional deterioration and foster the development of neurological disorders. In humans, long-term alterations in myeloid cell activation following microbial infections are suspected to influence some chronic inflammatory diseases [37–40]. A growing body of data also suggests a link between infections and AD. For instance, in mice and humans, viral (e.g., herpes simplex virus type 1) and bacterial (e.g., Porphyromonas gingivalis) infections have been associated with worsening of AD-like symptoms and brain Aβ deposition [41–45]. Likewise, 12 months after SARS-CoV-2 infection, patients have shown a higher risk for cognitive impairment and AD diagnosis, especially upon severe acute illness, compared with matched healthy individuals [46,47]. Whether infections have long-term effects on BAMs in humans, and whether this inflammatory scar sets the ground for neurodegenerative diseases in open for debate.

Microglia versus BAMs in brain aging and Alzheimer’s disease

Innate immunity in Alzheimer’s disease

AD is the most common neurodegenerative disorder amongst the elderly and causes marked behavioral alterations, including severe mood changes and cognitive impairment (reviewed in [48,49]). The brains of patients diagnosed with AD present two main pathological hallmarks: extracellular and vascular Aβ plaques and intracellular neurofibrillary tau tangles. These hallmarks are often accompanied by exacerbated glial activation and evident neuronal loss (reviewed in [48,49]; Figures 2, 3, 4). Aging is the main risk factor for late-onset AD [49]. However, apart from aging, multiple environmental (e.g., infections [41–45]) and genetic factors can increase the likelihood of developing late-onset AD. Amongst the ever growing list of common or rare risk genes and gene loci identified in AD genome wide association studies (GWAS), apolipoprotein E (APOE), triggering receptor expressed on myeloid cells 2 (TREM2), CD33, and complement receptor 1 (CR1) genes—to name a few—encode proteins that can be highly expressed by peripheral, CNS-border, and CNS parenchyma innate immune cells [9,11,12,50–57]. Many of these GWAS hits have placed microglia under the spotlight when it comes to brain immunity in AD. Consequently, the development and widespread implementation of advanced RNA-seq techniques has led to the identification of a unique state of chronically activated microglia, named disease-associated microglia (DAM), in transgenic mouse models of AD, such as those of brain amyloidosis (e.g., the 5xFAD mice) [58,59]. Researchers have shown via scRNA-seq that the acquisition of a DAM signature in the murine brain follows a two-step process that is highly dependent on Trem2 and Apoe expression [58–60] (Figure 2). Indeed, Apoe^−/− and Trem2^−/− AD transgenic mice show impaired microglial clustering around Aβ plaques, which then become more diffuse and neurotoxic, when compared to their wild type counterparts [61–64]. suggesting that both TREM2 and APOE might modulate a protective response by DAMs in the context of increased brain Aβ burden. Of note, murine brain BAMs also exhibit high expression or Trem2 and Apoe (reviewed in [9,10]), but less attention has been dedicated to their putative role(s) in models of AD. As we relay below, recent evidence supports the notion that when compared to microglia, altered BAM function should be considered as a distinct and integral component of the immune response in the aged and AD brain.

Figure 2. Scheme depicting the molecular signatures of parenchymal microglia and recruited monocyte-derived macrophages in the vicinity of amyloid plaques in AD transgenic mice.

One of the main brain pathological hallmarks of AD is the extracellular deposition of amyloid plaques (rich in aggregated Aβ species), a feature that is mimicked in AD transgenic mice [48]. Aging and the development of brain Aβ pathology are linked to the appearance of DAM that encircle plaques and the blood monocyte-derived DIMs that secrete TNF [58–60,63,64,74]. The possible contribution of skull bone marrow-derived myeloid progenitors to brain BAMs and DIMs in AD warrants further investigation. Green arrow represents a potentially protective mechanism. Red arrow represents a potentially deleterious mechanism. Dashed arrow with a question mark represents an unexplored connection. Aβ, amyloid beta; APOE, apolipoprotein E; CD83, cluster of differentiation 83; DAM, disease-associated microglia; DIMs, disease inflammatory macrophages; SBM, skull bone marrow; TNF, tumor necrosis factor; TREM2, triggering receptor expressed on myeloid cells 2. Figure created with Biorender.com.

Figure 3. Glymphatic function and microglial activation are modulated by brain BAMs in mice.

Exacerbation of brain Aβ pathology with age in transgenic mice is linked to an altered BAM activation signature, with less LYVE-1 and more MHC-II expression at the surface [20,34]. Activated leptomeningeal BAMs alter the perivascular extracellular matrix leading to defective CSF/interstitial fluid (ISF) flow via the glymphatic system. Therapeutic delivery of CSF-1 into the murine brain rejuvenates BAMs, which upregulates LYVE-1, and leads to improved CSF glymphatic flow [34]. Activated BAMs, particularly PVMs, secrete osteopontin, which further exacerbates microglial activation and excessive neuronal synaptic pruning [77]. Green arrow represents a potentially protective mechanism. Red arrow represents a potentially deleterious mechanism. Aβ, amyloid beta; BAMs, border-associated macrophages; CSF, cerebrospinal fluid; CSF-1, colony stimulating factor-1; ISF, interstitial fluid; LYVE-1, lymphatic vessel endothelial hyaluronan receptor-1; MHC-II, major histocompatibility complex type-II; NVU, neurovascular unit. Figure created with Biorender.com.

Figure 4. Scheme depicting newly described roles of brain innate immune cells in tau-mediated neurodegeneration.

The formation of intracellular neurofibrillary tangles and extracellular hyperphosphorylated tau aggregates is one of the brain pathological hallmarks of secondary tauopathies such as AD [48]. In mouse models of AD-like tauopathy expressing human APOE4, namely the PS19:APOE4 mice, increased microglial expression of MHC-II has been associated with increased recruitment of neurodegeneration-promoting activated T cell clones [78,79,84]. Yet, little is known about the contribution of MHC-II^high BAMs to this phenomenon. Red arrow represents a potentially deleterious mechanism. Dashed arrow with a question mark represents an unexplored connection. APOE4, human apolipoprotein E4 gene; BAMs, border-associated macrophages; MHC-II, major histocompatibility complex type-II; p-tau, hyperphosphorylated tau. Figure created with Biorender.com.

BAMs in brain aging and AD

The effects of aging on brain BAMs are far from being fully understood. Upon analyses of bulk RNA-seq data from whole meningeal dural preparations from young (2 to 3 months of age) and aged (22 months of age) mice, two independent studies identified significant alterations in gene expression modules and pathways related to monocyte and macrophage functions [65,66]. Compared with BAMs from young mice, aged dural and leptomeningeal BAMs concomitantly retained surface expression of CD206, expressed less lymphatic vessel endothelial hyaluronan receptor 1 (LYVE-1), became MHC-II^high and upregulated Trem2 gene expression [20,34,35,65]. When compared to young mice, this altered activation signature of brain meningeal and perivascular BAMs in aged mice was linked to lower extracellular matrix remodeling at the basement membrane of pial blood vessels, reduced arterial motion, and impaired CSF influx through the glymphatic pathway [34] (Figure 3). Notably, the rejuvenation of aged brain BAMs in mice (and possibly other resident myeloid cells like microglia) via the therapeutic delivery of CSF-1, reverted the perivascular extracellular matrix abnormalities and improved glymphatic function [34]. This newly described role of brain BAMs in regulating glymphatic CSF recirculation also prompts the investigation of possible deleterious interactions between dysfunctional MHC-II^high BAMs and the meningeal lymphatic vasculature in the dura of aged mice [18,67,68]. Perhaps a rejuvenation of aged dural BAMs via CSF-1 can exert beneficial effects on glymphatic CSF influx/efflux via enhanced brain lymphatic drainage. Accordingly, it will be important to delve further into the mechanistic underpinnings of the functional crosstalk between BAMs and other brain perivascular, vascular, and parenchymal cells, particularly neurons and glia. Identifying BAM-mediated mechanisms of accelerated neuronal and glial aging might lead to new candidate therapeutic strategies to delay or perhaps prevent, neuronal dysfunction and cognitive decline.

Regarding the development of brain amyloid pathology, despite pending debates, earlier studies showed that promoting the recruitment of monocytes into the brain (e.g., via immunotherapy against programmed death-ligand 1 (PD-L1) led to improved clearance of vascular Aβ and memory function in mouse models of brain amyloidosis, namely Trem2^−/− 5xFAD mice [69–75]. Likewise, and in agreement with the positive impact of BAMs on neurophysiology, depletion of ~75% of brain and leptomeningeal BAMs in 5xFAD mice resulted in reduced glymphatic function and increased brain Aβ plaque load [34]. However, recent studies also point towards potential deleterious roles of brain BAMs in models of AD-like brain amyloidosis. Recently published data obtained using 5xFAD mice uncovered a previously unappreciated phenomenon of increased monocyte recruitment into the brain’s perivascular spaces and parenchyma. Via fate-mapping models and the integrated analyses of different published scRNA-seq datasets, the authors demonstrated the existence of the so-called disease inflammatory macrophages (DIMs), which were distinct from yolk sac-derived DAMs or steady-state BAMs [60]. DIMs express Trem2 but, contrarily to the apparently neuroprotective DAM [61–63], expand in Trem2^−/− mice, upregulate CD83, and produce TNF, suggesting a potentially deleterious role by promoting a neurotoxic inflammatory environment, although this remains conjectural [60] (Figure 2). In fact, the recruitment and engraftment of monocyte-derived macrophages in the brains of 5xFAD transgenic mice remains a topic of debate. Another study employing the inducible Ccr2-CreERT2:R26-tdTomato reporter line on a 5xFAD background failed to observe any tdTomato⁺ myeloid cells around amyloid plaques, suggesting that circulating CCR2⁺ monocytes do not engraft the brains of AD transgenic mice [76].

Using the newly developed mutated amyloid precursor protein knock-in (APP-KI) mice, another study showed that brain PVMs secreted osteopontin in response to brain Aβ pathology [77]. Constitutive genetic deletion of osteopontin in APP-KI mice diminished microglial activation and prevented excessive neuronal synaptic pruning [77], suggesting that preventing osteopontin secretion by PVMs might be a promising therapeutic strategy to prevent neuronal decay in the context of exacerbated brain amyloidosis (Figure 3). However, additional experiments involving models of conditional gene deletion specifically in BAMs will be essential to support the claim that osteopontin produced by brain PVMs underlie the observed neuronal synaptic damage in AD.

From another perspective, in a mouse model of CNS tauopathy expressing human APOE4 (PS19:APOE4 mice), used to mimic the secondary tauopathy observed in AD [48], widespread ablation of brain myeloid cells via CSF1R inhibitors PLX3397 or PLX5622 prevented tau-mediated neurodegeneration almost completely, in a process that was independent of TREM2 [78–80]. However, this effect cannot be attributed to the depletion of microglia alone, because PLX compounds have been shown to cause the depletion of non-microglial innate immune cells, namely brain BAMs, as well as peripheral dendritic cells, monocytes, and neutrophils [81–83]. Recently, researchers also observed accumulated clones of activated CD4⁺ and CD8⁺ T cells in the hippocampus of PS19:APOE4 mice, in close proximity to innate immune cells resembling microglia with high expression of MHC-II [84]. Notably, and again employing a similar strategy, PLX3397-mediated widespread depletion of myeloid cells in the brain and its border tissues, reduced T cell infiltration and curbed neurodegeneration were noted relative to control mice[84]. This strongly suggested that antigen presentation by CNS MHC-II^high innate immune cells was necessary for activated T cell extravasation, retention, and/or effector function, at least in this mouse model of tau-mediated neurodegeneration [84] (Figure 4). However, whether this outcome was exclusively mediated by microglia, BAMs, or both, warrants further rigorous experimentation using BAMs-specific Mrc1-CreERT2 and Lyve1-NCre:Cx3cr1CCre, or microglial-specific Hexb-CreERT2, P2ry12-CreER and Tmem119-CreER mouse lines [17,85–87]. In fact, the need to disentangle the specific roles of microglia and BAMs in neurodegenerative diseases is further emphasized by recent findings showing that MHC-II expression by BAMs, rather than by microglia, is required to promote the entry of encephalitogenic CD4⁺ T cells into the brains of mice injected with a viral vector expressing human αSyn (a mouse model of Parkinson’s disease) [88,89].

Of note, the high disease penetrance in animal models overexpressing mutated APP and presenilin genes (causative of familial forms of AD), like the 5xFAD mice, does not favor the search for mechanisms underlying sporadic late-onset AD etiology, and rather favors the identification of factors influencing the acceleration (or delay) of artificial brain pathological features. Moreover, as most of the data discussed here were obtained with animal models, investigating in depth the translational relevance of most of the above-mentioned preclinical findings is necessary.

In sum, the verdict is still out when it comes to the detrimental or protective roles of recruited monocyte-derived DIMs and brain BAMs in the context of AD pathophysiology. Nevertheless, we posit that the findings discussed here support an initial beneficial role of brain BAMs in maintaining brain function and preventing neurodegeneration induced by Aβ accumulation and deposition. However, with advanced aging and exacerbated amyloidosis and/or tauopathy, inflammatory monocytes may engraft the CNS and BAMs may acquire deleterious activation signatures that promote unwanted brain immune responses, neuronal synaptic dysfunction, neurodegeneration, and accelerated cognitive decay. These possibilities certainly merit further attention.

Concluding remarks

Significant effort must be invested in better understanding the putative roles of brain BAMs and the downstream consequences of their function/malfunction in steady-state or disease. Due to their privileged spatial localization, BAMs serve as the first sentinels that prevent widespread brain infection by phagocytosing microbes and triggering specific adaptive immune responses [13]. Brain BAMs may also play important effector roles in aging and in mouse models of AD, affecting vascular pulsation and CSF flow, brain parenchymal immunity (via direct signaling to microglia), and the development of pathology and neurodegeneration [7–9,18]. We hypothesize that severe consecutive or chronic infections may accelerate the aging of BAMs, pushing them into a maladaptive activation state that no longer provides neurotrophic support. This may in turn further fuel unwanted neuroinflammation and favor the accumulation of toxic protein aggregates in the AD brain. Despite the need for rigorous investigation (see outstanding questions), if we are right, developing therapies that can specifically target BAMs, instead of all CNS innate immune cells indiscriminately, should hold promise in preventing the long-term neurological sequelae of infections, including the risk in the elderly population of cognitive decay and dementias such as AD.

Outstanding questions box.

Can BAMs establish a molecular crosstalk with brain parenchymal cells other than microglia? Is the nature of this communication different across different anatomical locations?

What is the exact contribution of circulating blood monocytes versus skull bone marrow-derived hematopoietic progenitors to the pool of meningeal macrophages upon infection? Is the nature of dural macrophage replenishment the same in different models of neurodegenerative diseases? Are these processes, and the mechanisms governing them, distinct in males and females?

Can skull bone marrow-derived hematopoietic progenitors replenish brain PVMs in disease conditions where the meningeal arachnoid barrier is disrupted? Can this also be interrogated for parenchymal microglia?

What are the consequences of multiple infections to brain BAM turnover and function? Is it possible to prevent or revert the long-lasting transcriptional and functional changes observed in BAMs upon CNS pathogenic infections?

What are the exact molecular triggers of accelerated BAM aging? Do long-lived brain BAMs become senescent? Besides vascular function and glymphatic flow, what other neurophysiological processes are modulated by BAMs?

Can AD-like amyloid or tau pathology progression be affected via the specific modulation of brain BAM responses? Are MHC-II^high BAMs involved in the recruitment of neurodegeneration-promoting T cells to the AD brain?

Highlights.

• Border-associated macrophages (BAMs) in the brain are transcriptionally and phenotypically distinct from microglia as shown in mice and humans.

• BAMs promote brain glymphatic cleansing at steady state, and their homeostatic polarization correlates with proper cognitive function in mice.

• BAMs play an active role during murine infections by limiting microbial dissemination into the brain. Accordingly BAMs can be replaced by peripheral monocytes and undergo a long-lasting transcriptional reprogramming which affects their homeostatic signature and could compromise their ability to sustain brain function.

• With aging, BAMs become MHC-II^high and cease to degrade the extracellular matrix, contributing to defects in vascular and glymphatic functions in mice.

• In mice and humans, BAMs, like microglia, express genes that alter the risk for AD, including TREM2 and APOE. The depletion of BAMs in a mouse model of AD-like brain amyloidosis led to worsened Aβ accumulation and deposition.

• By contrast, BAMs respond to increased Aβ pathology by secreting osteopontin that exacerbates microglial activation and synaptic pruning in mice.

• We propose that BAM dysfunction following serial infections or other inflammatory insults contributes to accelerated brain aging and a higher risk for AD development in mice and humans.

Significance Box.

Mammalian central nervous system non-parenchymal macrophages (BAMs) play distinct roles from parenchymal microglia in supporting neurophysiology during adulthood and aging; they help prevent the dissemination of infectious agents, or respond to Alzheimer’s disease (AD)-like brain pathology and inflammation. We postulate that the acquisition of an enduring deleterious phenotype by brain BAMs (in response to repeated infections or inflammatory insults) is at the genesis of accelerated cognitive decline in the elderly and the development of AD.

Glossary

Alpha synuclein protein:

abundant brain pre-synaptic neuronal protein that can misfold and polymerize to form toxic fibrils coalescing into pathological neuronal inclusions in Parkinson’s and Alzheimer’s diseases.

Amyloid beta protein:

protein resulting from misprocessing of the membrane amyloid precursor protein, can form beta sheet-rich fibrils, depositing in the brain as amyloid plaques. The extracellular amyloid plaques are a pathological hallmark of Alzheimer’s disease (AD), are markedly immunogenic, can disrupt neuronal communication, and contribute to neurodegeneration.

Border tissues:

in the central nervous system (CNS), comprise the perivascular spaces, choroid plexus, and meningeal layers.

Border-associated macrophages:

in the brain, this group comprises all non-parenchymal macrophages, including perivascular, choroid plexus, and meningeal (dural and pial) ones.

Brain ventricles:

system of interconnected cavities within the brain that are filled with cerebrospinal fluid (CSF).

Cerebrospinal fluid:

clear fluid with low protein content, whose main molecular components result from a controlled and selective filtration of plasma, and active secretion from choroid plexus epithelial cells. It fills the brain’s ventricles, cisterns, and sulci, the spinal cord central canal, and the subarachnoid spaces; provides physical and trophic support to neural tissues, playing a crucial role in CNS cleansing.

Choroid plexus:

membranous tissue; attached to the ependymal wall of each brain ventricle and formed by a monolayer of specialized epithelial cells bound together by tight junctions (forming the blood-CSF barrier) enclosing a stromal compartment irrigated by fenestrated blood vessels.

Central nervous system (or brain) parenchyma:

structure formed by neurons and glial cells, embedded in a connective tissue matrix, delineated by a subpial glia limitans and ventricular ependyma, and irrigated by specialized blood vessels that form a tight barrier between the blood circulation and the perivascular spaces.

Disease inflammatory macrophages:

population of peripheral monocyte-derived macrophages expressing TREM2, CD83, TNF, and other proinflammatory molecules; found in greater numbers in aged and in amyloid plaque-containing brains.

Disease-associated microglia:

two-step, Trem2- and Apoe-dependent, microglial activation signature found in the brain parenchyma; expands considerably in the presence of overt AD-like brain amyloid plaque or tau pathology, or at a lower extent during aging.

Glymphatic system:

brain waste clearance system relying on CSF flow through periarterial spaces, diffusion into the parenchyma, and exit of interstitial fluid from the brain via perivenous spaces back into the subarachnoid CSF sink. This system removes parenchymal waste products and toxins.

Meninges:

tissue wrapping the brain and spinal cord; composed by the outmost dura mater, intermediate arachnoid, and inner pia mater. In the cranium, the dura is further subdivided into the skull-attached periosteal layer and the inner meningeal layer, between which exist the venous sinuses, lymphatic vessels, and dural resident immune cells. The avascular tight arachnoid layer with barrier-like properties, and the highly vascularized inner pia are connected by trabeculae, delimitating the subarachnoid space filled with flowing CSF. The arachnoid and pia together form the leptomeninges.

Microglia:

CNS parenchyma resident innate immune cells originating exclusively from prenatal yolk-sac, non-monocytic, c-Myb-independent, c-KIT⁺ erythromyeloid progenitors.

Myelinated white matter regions:

composed mainly of nerve fibers, coated with an insulating lipid-rich substance called myelin (pale color, hence “white matter”) produced by oligodendrocytes in the CNS.

Neuronal synaptic pruning:

neuronal synaptic remodeling modulating the efficiency of neuronal electrical transmissions (neuronal autonomous or mediated by other brain cells, namely microglia).

Non-immune privileged:

tissue with proper immunosurveillance from the existence of a draining lymphatic system, the absence of blood vascular tight junctions, facilitated immune infiltration, and the expression of antigen-presenting molecules such as major histocompatibility complex class I.

Perivascular macrophages:

located in the perivascular spaces of the CNS; more abundant around penetrating arteries and arterioles.

Tau protein:

microtubule-associated; can become hyperphosphorylated, aggregate, and form the toxic neurofibrillary tangles that are a pathological hallmark of primary tauopathies and AD.