Parenchymal microglia are the brain-resident immune cells and they play crucial roles in the development, maintenance of homeostasis, and diseases in the CNS. Microglia are crucial in brain development and regulate many mechanisms including synaptic pruning and maturation and angiogenesis (Arnold and Betsholtz, 2013). Under physiological conditions in adult life, microglia are constantly monitoring their surroundings thanks to their fine ramified motile processes (Nimmerjahn et al., 2005). Once microglia encounter harmful substances, such as infiltrating components from blood, burden abnormal proteins, or cell debris, they become activated to phagocyte these harmful substances or to protect the damaged cells (Hu et al., 2015). Furthermore, microglia can promote angiogenesis in both physiological (Foulquier et al., 2019) and pathological conditions such as ischemic stroke, AD, multiple sclerosis (MS), and Parkinson’s disease (PD) (Zhao et al., 2018), highlighting the important cross-talk of microglia with the cerebral vasculature (VAM). There is, however, a lack of molecular studies to differentiate VAM from the general population of parenchymal microglia, and we suggest that an advanced molecular characterization should be undertaken to reveal their true nature and assess their potential protective and/or deleterious functions in the context of cerebrovascular diseases.

Apart from microglia, CNS macrophages are also involved in the maintenance of brain homeostasis, but their role is limited to its borders. CNS macrophages reside in the non-parenchymal perivascular space, subdural meningeal spaces, and choroid plexus spaces – namely, PVMs, meningeal macrophages (MMs), and choroid plexus macrophages (CPMs), respectively (Iadecola, 2017; Figure 1A). PVMs and MMs persist throughout the life of the organism due to their longevity and their capacity of self-proliferation, rather than the infiltration of peripheral myeloid cells. In contrast, CPMs mainly depend on blood-derived immigrating Ly6C^hi monocytes after birth (Goldmann et al., 2016). While microglia and macrophages share many functions and markers, previous studies have revealed differential marker expressions useful for their distinction (next paragraph, Table 1).

Studies on VAM or PVMs and their involvement in cerebrovascular dysfunction were identified from electronic searches exclusively done by using the PubMed database. We used the following MeSH and free search terms to identify peer-reviewed original articles in English: for VAM: microglia AND (blood–brain barrier OR cerebral small vessel) AND (cerebrovascular disorders OR cSVD OR stroke); and for PVMs: (perivascular macrophages OR CNS macrophages OR brain macrophages) AND (cerebrovascular disorders OR cSVD OR stroke). In both searches, we excluded studies matching the following terms: review, infection, epilepsy, and hemorrhage. Furthermore, studies were excluded if they referred to microglia or macrophages in the context of brain tumor/metastasis, non-CNS diseases, infectious diseases, and drug or alcohol abuse. Screening and extraction of articles were done by TK under the guidance of SF. For each study, the following variables were recorded: (a) year of publication, (b) type of disease, (c) animal model, (d) microglia or macrophages markers, and (e) results. The searches were performed on October 30, 2019, and the results are described in a flow diagram (Figure 2).

Perivascular macrophages have been shown to contribute to an increased vascular permeability and an increased granulocyte recruitment in the acute phase of stroke using the transient middle cerebral artery occlusion (tMCAO) model. Their depletion by the administration of clodronate-containing liposomes (CCLs) was indeed able to attenuate the vascular permeability, as evidenced by the reduced Evans blue extravasation 24 h post-ischemia, and to reduce the granulocyte infiltration in the ischemic cortex (Pedragosa et al., 2018). In an acute hypertension model induced by Angiotensin II infusion (2 weeks), the increased BBB permeability was shown to result from the generation of reactive oxygen species (ROS) mediated by the Angiotensin II type 1 receptor (AT₁R) and the subsequent activation of Nox2 in PVMs. Indeed, the depletion of PVMs or the deletion of AT₁R or Nox2 from PVMs, using bone marrow chimeras, were able to prevent BBB dysfunction, restore the neurovascular coupling and prevent cognitive dysfunction (Faraco et al., 2016). In SHR-SP, a chronic hypertensive model, PVM depletion with CCLs improved the endothelium-dependent dilation of the MCA, and prevented MCA structural remodeling induced by hypertension (Pires et al., 2014). Post-mortem investigation of brains from CADASIL patients (cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy), a hereditary cSVD form, has revealed the presence of PVM-like cells with a phagocytic morphology around granular osmiophilic material depositions in arteries and arterioles, suggesting a role for PVMs to clear abnormal depositions in the perivascular space to prevent vascular remodeling (Yamamoto et al., 2013).

The contribution of PVM to cerebrovascular dysfunction was also evidenced in few studies on AD. PVM depletion using CCLs increased the number of amyloid depositions in cortical vessels of an AD mouse model an AD mouse model (Hawkes and McLaurin, 2009). This highlights again the crucial phagocytic role of PVMs to preserve the cerebrovasculature. This study reported for the first time that modulating PVM density can influence the clearance of amyloid from the cerebral vasculature. Unfortunately, it is highly possible that an increased clearance capacity can be accompanied by an increased ROS production. Park et al. (2017) have shown that the selective depletion of PVMs using intracerebroventricular injection of CCLs reduced ROS production and cerebrovascular dysfunction induced in different AD models. By using bone marrow transplantation, the involvement of CD36 and Nox2 from PVMs was demonstrated to contribute to the neurovascular dysfunction associated with the amyloid deposition (Park et al., 2017). These results indicate a crucial role for PVMs in clearing Aβ peptides from perivascular spaces and for preventing Aβ accumulation in cerebral vessels but it could be at the expense of a generation of ROS deleterious for the integrity of cerebral vessels in the long term.

In summary, while PVMs seem to exert protective effects at first to halt the progression of pathological events such as removal of harmful proteins, we suggest that their repeated activation and exposure to danger signals may lead over time to different brain disease-specific deleterious effects. The biology and pathobiology of PVMs in other brain diseases and in other tissues have been described in other recent reviews (Faraco et al., 2017; Lapenna et al., 2018).

While the activation of parenchymal microglia in the presence of BBB leakages in hypertensive cSVD models is well known (Kaiser et al., 2014; Foulquier et al., 2018) evidences for the contribution of VAM to the initiation of cerebrovascular dysfunction remain limited.

In the genetic hypertensive rat model (Cyp1a1-Ren2), the microglia density increased in 6-month-old rats with a higher number of VAM (presumably VAM based on their shape, Iba1 immunoreactivity, and location), prior to any cerebrovascular lesions. This study further indicated that a modest but chronic blood pressure elevation can induce the regulation of growth factors and inflammatory genes prior to vascular remodeling, suggesting a role for VAM in the progression of cerebrovascular dysfunction (Pannozzo et al., 2013). In DOCA-salt rat, a sub-chronic hypertension model, while PVM phenotype did not change, VAM exhibited dynamic phenotypic changes; proliferative parenchymal and VAM proliferated before switching to a pro-inflammatory state and before BBB impairment and the occurrence of cerebrovascular lesions (Koizumi et al., 2019). In a post-mortem study on CAA, Carrano et al. revealed that major changes in tight-junction protein expressions (claudin-5, Occludin, ZO-1) were observed in CAA-affected capillaries engulfed by NADPH oxidase-2 (NOX-2)-positive activated microglia, and this was observed in association with BBB disruption (Carrano et al., 2012). In addition, Aβ induced ROS formation by binding RAGE (receptor of advanced glycogen product), an Aβ transporter. In vitro, blocking RAGE or inhibiting NOX-2 reduced the toxic effect of Aβ on endothelial cells (Carrano et al., 2011, 2012), supporting the evidence that the increased expression of NOX-2 in VAM could also affect the cerebral small vessels, similarly to the findings on PVMs.

Following an artificial BBB leakage induced by focal laser-injury, the immediate accumulation of processes from VAM toward the laser-injured capillary was capable of closing the BBB. This highly migratory behavior of VAM relied on the function of P2RY12 receptors as their inhibition using clopidogrel or their genetic ablation suppressed VAM motility and thereby led to a prolonged delay before BBB closure (Lou et al., 2016). This highlights the importance of VAM for BBB repair. In fact, the ultrastructural analysis of the laser injury by electron microscopy revealed that the aggregation of densely packed processes completely sheathed the site of injury. Immuno-labeling revealed that VAM processes, extending toward the laser-injured site, exhibited high P2RY12 expression (Lou et al., 2016). In this article, the authors used the term “juxtavascular microglia” to describe the vessel-associated Cx3Cr1-positive microglia at the capillary level with extended and fine processes, i.e., VAM. However, they wrongly defined juxtavascular microglia being largely localized within the perivascular space, which would correspond to the definition of PVMs. Another evidence has pointed toward a BBB repair role for VAM.

In tMCAO mice, Iba-1-positive cells with ramified processes start to cluster around vessels in the penumbra area within 1 h after the initiation of the ischemic insult. The attracted cells enwrap blood vessels in the penumbra 24 h post tMCAO and harbor an amoeboid shape and a high CD68 expression (Jolivel et al., 2015). The hypoperfusion induced by tMCAO in the area surrounding the ischemic core may continuously produce pro-inflammatory components, such as DAMPs, ROS, and inflammatory cytokines that may themselves, or after the leakage of plasma immunoglobulins or proteins, induce the infiltration of circulating monocytes or neutrophils into the brain parenchyma (Benjelloun et al., 1999; Krueger et al., 2019). The rapid mobilization of microglia after the ischemic insult suggests that VAM could be the first to initiate the BBB repair in the early phase of the injury before the migration of other parenchymal or systemic macrophages in the following hours/days.

In summary, parenchymal and VAM are quickly mobilized and accumulate around cerebral vessels following BBB dysfunction. In addition, they appear to be already present and activated in some conditions in the absence of BBB dysfunction in the early phase of cerebrovascular diseases and could therefore be targeted to prevent disease’s progression. This will however only be achieved thanks to a refined molecular characterization of their activation dynamics.