Mural cells are cellular components with contractile properties in the NVU, that enable them to regulate vascular tone and the CBF [22]. As pial arteries branch into arterioles and capillaries after penetrating into parenchyma, the mural cell population composed of the NVU changes [51]. Penetrating arteries consist of one to three layers of VSMCs while arterioles contain only one layer [41, 51]. After descending to capillary level, pericytes replace VSMCs and embed within the endothelial BM [51]. Previously, the CBF was shown to be regulated solely by VSMCs [52]. However, with the study of pericytes, this viewpoint has been challenged.

In a series of studies, pericyte were confirmed to constrict or dilate in response to neurotransmitters [53], for example, glutamate evokes pericyte dilation, and pericytes constrict in response to a gamma-aminobutyric acid (GABA) receptor blocker suggesting that pericytes participate in the regulation of the CBF [52, 54, 55, 56, 57]. Moreover, an in vivo experiment in which mice expressed DsRed in pericytes, revealed that capillary dilation precedes penetrating arterioles, demonstrating that capillary dilation is a result of active relaxation of pericytes rather than a passive response to elevated blood pressure caused by arteriole dilation [52]. In ischemic stroke, capillaries constrict segmentally at regions near pericytes, after which pericytes contract and subsequently die rigidly [58]. Damages to pericytes contributes to long-lasting microcirculatory reflow impairment even after reperfusion [59].

Pericytes degeneration and neurovascular dysfunction have been observed in AD [51, 52]. Moreover, the oxidative stress caused by A$\beta$ leads to capillary constriction. A reduction in the CBF reduces the oxygen supply and glucose availability to the brain, resulting in the impairment of neurons and neurodegenerative changes [60]. A recent study showed that white matter lesions (WMLs) induced by persistent cerebral hypoperfusion is a driving factor for dementia [61]. Moreover, the two-hit vascular hypothesis suggests that prior to neurodegeneration and cognitive impairment, genetic, vascular and environmental factors cause vascular damage (hit1), and neurovascular dysfunction contributes to the accumulation of A$\beta$ (hit 2) [40, 41]. A previous study using APP${}^{\text{sw/0}}$ Pdgfr$\beta$${}^{+/-}$ mice found that the loss of pericytes results in a series of AD-like neurodegeneration pathological changes including accelerated A$\beta$ deposition, tau pathology and neuronal dysfunction [62]. Notably, vascular damage (hit1) has been observed in pericyte-deficient APP${}^{\text{sw/0}}$ mice, indicating that vessel damage and pericyte degeneration may be mutually causal [62]. Together, these findings suggest that pericyte degeneration is an early and key event in AD neurodegeneration.

However, due to differences in stimulation methods, transgenic mice, and other factors, the function of pericytes in regulating the CBF has not been fully elucidated.

The blood-brain barrier (BBB) is a special protective barrier that exists between capillaries and the brain, and is composed of ECs, endothelial tight junctions (TJs), the BM, pericytes and astrocyte endfeet [63, 64]. The BBB protects the brain from invasion by blood-derived harmful factors, thus maintaining the homeostasis of the CNS [65]. Using quail-chick transplantation chimeras, BBB was shown to develop in response to the neural tissue environment [66]. During embryogenesis in rodents, astrocytes and pericytes are required to wrap immature vessels. It is widely accepted that immature vessels are covered preferentially by astrocytes postnatally. However, Daneman et al. [67] reported that pericytes were recruited during embryogenesis, more than one week before the generation of astrocytes, thus revealing the role of pericytes in the formation of the BBB. The permeability of the BBB is increased in pericyte-deficient mice, indicating the essential role of pericytes in maintaining BBB integrity [49]. Pericytes maintain the integrity of the BBB through two pathways, forming and preserving the TJs of ECs, and transcytosing in the CNS ECs.

The breakdown of the BBB has been observed in AD [68], which is attributed to the loss and detachment of pericytes destroying the integral structure of the BBB and increasing BBB permeability [69]. Interestingly, sirtuin-1 (SIRT1), an anti-aging gene, is markedly suppressed exposed to A$\beta$ [70]. The decreased expression of SIRT1 also increased the permeability of the BBB and accelerated the process of senescence [71].

Angiogenesis is the formation of new blood vessels from existing vessels. Angiogenesis involves three steps: initiation, sprouting/migration, and maturation. In angiogenesis, the complex signaling pathways between ECs and pericytes which include the platelet-derived growth factor (PDGF)-BB/PDGF receptor beta (PDGFR$\beta$) pathway [72, 73], the angiopoitin1/tyrosine protein kinase receptor (Tie2) signaling pathway [74], the vascular endothelial growth factor (VEGF)/VEGF receptor (VEGFR) pathway [46], the sphingosine-1-phosphate (S1P) signaling pathway [75], transforming growth factor beta (TGF-$\beta$) [76] are the foundation for the formation and stabilization of new blood vessels. Pericytes regulate the expression of VEGF, resulting in the instability of blood vessels and initiating the angiogenesis [77]. Pericytes detach from blood vessels to pave the way for endothelial sprouting [78]. The migrating ECs secrete VEGF to stabilize nascent vessels and signal to pericytes to recruit VEGF. Moreover, the recruited pericytes communicate with ECs to promote the stabilization and maturation of new blood vessels [78]. The coverage of pericytes is a marker of vascular maturation and a lack of pericytes results in vascular hyperplasia [79].

LDL receptor-related protein-1 (LRP-1) is an apoE receptor that mediates the clearance of A$\beta$. Using freshly isolated cortical slices incubated with A$\beta$, Ma et al. [80] showed that pericytes rapidly remove Cy3-A$\beta$42. In addition, in AD and APP${}^{\text{swe/0}}$ mice, the abundant accumulation of A$\beta$ in pericytes indicates the important role of these cells in clearing A$\beta$ at the BBB. Additionally, in an LRP-1 conditional knockout model, Cy3-A$\beta$42 uptake by pericytes was reduced by 80% compared with that in the control group. The process of clearing A$\beta$ by pericytes can be inhibited by antibodies against LRP-1 [80, 84], further confirming that the clearance of A$\beta$ is mediated by LRP-1 (Fig. 1A). Compared with that in adult wild-type mice, clearance in apoE knockout mice is substantially reduced, indicating that LRP-1-mediated transport can be influenced by apoE, a definite risk factor for AD [85]. ApoE is required for A$\beta$ clearance and is isoform- specific. A study revealed that apoE3, but not apoE4, normalizes A$\beta$ clearance in mouse pericytes with silenced mouse apoE [80], while another study revealed that the binding of A$\beta$ to apoE3 reduces its clearance rate at the BBB. It has also been reported that A$\beta$ binding to apoJ significantly accelerates the BBB clearance rate [86, 89].

Fig. 1.

The interactions between A$\beta$ and pericytes. (A) LRP-1 and P-gp mediate the efflux of A$\beta$ while RAGE regulate the influx of A$\beta$ into brain parenchyma. BACE1 on pericytes degrades A$\beta$1-40 into A$\beta$34 intermediates and the A$\beta$ (1-42)-RAGE interaction induces the generation of VEGF and MCP-1 contributing to the vascular remodeling. (B) CD36 mediates the clearance of A$\beta$1-40 by pericytes. The reduced expression of CD36 promotes the deposition of A$\beta$1-40 resulting in CAA. (C) Oligomeric A$\beta$1-42 activates NOX4 in pericytes to produce ROS and ET in sequence, and ET binds to ET${}_{\text{A}}$-R on pericytes, triggering capillary constriction. Capillary constriction results in the reduction of CBF and the glucose and oxygen it contains. Hypoxia in turn upregulates the expression of BACE1, further increasing the generation of A$\beta$ and forming an amplified positive loop, ultimately leading to synapse dysfunction and neuron loss. (D) Fibrillar A$\beta$1-42 activates MMP-9 to induce NG2 sheds from pericytes leading to the detachment of pericytes and the destruction of endothelial TJs which is the significant part of BBB. (E) A$\beta$1-40 induces pericytes mitophagy through the CD36/PINK1/Parkin pathway and increases oxidative stress in pericytes. The increased lipid ROS and iron ions caused by oxidative stress in pericytes inducing pericytes ferroptosis dependent on mitochondrial autophagy. (F) Fibrillar A$\beta$1-40 reduces the viability and proliferation of pericytes, and increases the activity of the key apoptotic proteins caspase3/7 while the effects of monomer A$\beta$1-40 are completely opposite. LRP-1, LDL receptor-related protein-1; RAGE, receptors for advanced glycation end products; A$\beta$, beta-amyloid; BACE1, $\beta$-site amyloid precursor protein (APP) cleaving enzyme 1; VEGF, vascular endothelial growth factor; CAA, cerebral amyloid angiogenesis; NOX4, nicotinamide adenine dinucleotide phosphate oxidase 4; ROS, reactive oxygen species; ET, endothelin; CBF, cerebral blood flow; MMP-9, matrix metalloproteinase-9; NG2, neural/glial antigen 2; TJs, endothelial tight junctions; BBB, blood–brain barrier; PINK1, PTEN-induced putative kinase 1; sAPP$\beta$, soluble amyloid precursor protein beta; MCP-1, monocyte chemoattractant protein-1; ET${}_{\text{A}}$-R, endothelin receptor type A.

The expression of LRP-1 is downregulated in AD patients exacerbating A$\beta$ pathology [90]. By injecting pericytes into APP/PS1 mice, Tachibana et al. [91] showed that microcirculation improved in the pericyte-injected hemisphere and that the deposition of A$\beta$ decreased in a manner dependent on the expression of LRP-1 on pericytes. However, in a recent phase I clinical trial in which mesenchymal stem cells (MSCs) were stereotactically injected to the brains of AD patients, no significant effects on cognitive function were observed. The use of transplanted pericytes or MSCs in the brain to prevent or treat AD has not been validated [92].

Amyloid precursor protein (APP) is cleaved by $\beta$-Site Amyloid Precursor Protein (APP) Cleaving Enzyme 1 (BACE1) and $\gamma$-secretase, which sequentially results in the formation of A$\beta$ peptides, including A$\beta$40 and A$\beta$42 [93]. In addition to “amyloidogenic” activity, BACE1 also possesses “amyloidolytic” activity, whereby it degrades longer A$\beta$ isoforms at position 34 into A$\beta$34 intermediates [94] (Fig. 1A). A previous study demonstrated that BACE1 is expressed in pericytes [95]. Treating pericytes with a BACE inhibitor resulted in a dose-dependent decrease in A$\beta$34 levels, indicating the role of BACE1 in the cleavage of A$\beta$ peptides and the formation of A$\beta$34 in pericytes [95]. Notably, the substrate of BACE1 is A$\beta$40, but not A$\beta$42 [94, 95].

In AD, the A$\beta$34/A$\beta$40 ratio is decreased significantly, and the level of A$\beta$34 is correlated with disease progression [96]. The progression of AD can be divided into six stages. AD in Braak stage I-II is clinically silent, AD in Braak stage III–IV is incipient and AD in Braak stage V–VI is fully developed [97, 98, 99]. Kirabali et al. [95] revealed that in Braak stage II, A$\beta$34 levels peak, while in Braak stages III and IV, the immunoactivity of A$\beta$34 significantly decreases which explains the dysfunction and loss of brain pericytes in AD pathogenesis. Moreover, analysis of PDGFR$\beta$ immunoactivity revealed that the loss of pericytes had already occurred at Braak stage II [37].

AGEs (advanced glycation end-products) are the final products of the nonenzymatic glycation of proteins (Maillard reaction) [100], which is irreversible. AGEs can bind and destroy various histocytes through a process called cross linking. A series of studies have shown that AGEs accelerate aging and cause neurodegenerative disorders including AD [101, 102, 103]. Receptors for advanced glycation end products (RAGE), a multiligand receptor of the immunoglobulin family, can not only specifically bind to AGEs, but also bind to various ligands such as high-mobility group box-1 (HMGB1), S100 and A$\beta$ [104, 105], and play vital roles in the occurrence and development of various diseases. RAGE plays a critical role in regulating the influx of circulating A$\beta$ into the brain via the BBB as a transporter [106]. Moreover, RAGE promotes the generation and accumulation of A$\beta$ by enhancing the activity of $\beta$-secretase and $\gamma$-secretase [107]. In addition, RAGE induces dysfunction of synapses and neuronal circuits, which is the structural foundation of cognition [108, 109]. According to its structure, RAGE is classified into three isoforms: N-truncated, and C-truncated, which are also called endogenous secretory RAGE (esRAGE) [110]. Moreover, RAGE can be cleaved by proteolytic enzymes to form cRAGE. The soluble form of RAGE (sRAGE) is composed of esRAGE and cRAGE. Notably, sRAGE can interact with A$\beta$ to form sRAGE-A$\beta$ interactions which can inhibit the neurotoxicity of RAGE and promote A$\beta$ clearance from the brain [110].

It has been proven that RAGE is expressed on pericytes [111]. Using small interfering RNA (siRNA) technology to suppress the expression of the pericyte RAGE gene, Lue et al. [111] showed that the level of A$\beta$ dramatically decreased, indicating that the A$\beta$ (1-42)-RAGE interaction may function by tethering A$\beta$ to the cell surface of pericytes, advancing the A$\beta$-A$\beta$ interaction and further promoting fibrillogenesis. By blocking RAGE with an anti-RAGE antibody, the levels of A$\beta$-induced VEGF and monocyte chemoattractant protein-1 (MCP-1) were decreased, indicating that RAGE-A$\beta$ interactions in pericytes contribute to the vascular remodeling that is observed in AD (Fig. 1A). The combination of A$\beta$ and pericyte RAGE can induce oxidative stress and a subsequent inflammatory response by activating a variety of signaling pathways, including mitogen-activated protein kinase (MAPK), glycogen synthase kinase 3 (GSK-3) and nuclear factor kappa-B (NF-$\kappa$B) [112]. Oxidative stress and inflammatory reactions can thicken the BM and increase the deposition of A$\beta$, eventually leading to vascular amyloidosis and disruption of the BBB [111]. In addition, the A$\beta$-RAGE interaction results in cognitive impairment by accelerating the aging process and inducing oxidative stress [113]. In AD, elevated RAGE levels may account for neuronal death and cognitive impairment. However, the level of sRAGE is lower [114]. In view of this, RAGE inhibitors may be potential targets for treating AD and these agents have been proven to be effective in preclinical and clinical studies [115], although the results have been unsatisfactory.

P-gp, a subtype of the ATP binding cassette (ABC) transporter family, is an ATP-dependent transporter responsible for the efflux of various substrates from the brain to the blood [116, 117]. Using an immunogold technique with monoclonal anti-P-gp antibodies, Bendayan et al. [118] showed that gold particles are present in ECs, astrocytes and pericytes, suggesting that pericytes can express P-gp. In human studies, the expression level of P-gp was shown to be negatively correlated with the accumulation of A$\beta$ [119, 120].

Moreover, it has been proven that inhibiting P-gp leads to increased intracellular deposition of A$\beta$ in brain capillaries [121, 122], suggesting that P-gp plays a crucial role in the clearance of A$\beta$. However, the mechanism by which P-gp affects A$\beta$ transport remains controversial. A variety of studies support the notion that A$\beta$ stimulates the ATPase activity of P-gp [123, 124]. However, Bello et al. [125] reported that A$\beta$ has no effect on the adenosine triphosphate (ATP) hydrolysis activity of P-gp. In previous research, McCormick et al. [123] noted that the activation of P-gp ATPase by A$\beta$ depends on the lipid environment, which may account for the differences between those studies. In addition, A$\beta$ can affect P-gp conversely. By treating transgenic human amyloid precursor protein (hAPP) overexpressing mice with an irreversible inhibitor of the ubiquitin-activating enzyme E1, Hartz et al. [126] showed that retained P-gp results in a decreased level of A$\beta$, suggesting that A$\beta$ induces P-gp degeneration through the ubiquitination pathway. In addition, the brains of AD patients exhibit marked decrease in P-gp and a significant increase in A$\beta$ deposition and ubiquitinated A$\beta$ [126, 127, 128].

CD36, a glycosylated membrane protein, is widely expressed in the nervous system, including in pericytes [129]. CD36 is involved in a variety of pathological processes, such as vascular oxidative stress, the inflammatory response, mitochondrial dysfunction and neurovascular uncoupling [129, 130, 131].

Immunofluorescence staining revealed that, CD36 and A$\beta$1-40 colocalize with PDGFR$\beta$, a marker of pericytes, suggesting that CD36 may be involved in the clearance of A$\beta$1-40 by pericytes [132]. In transgenic mice lacking CD36, Li et al. [132] showed a reduction in A$\beta$1-40 and cerebral amyloid angiopathy (CAA), suggesting that CD36 promotes the deposition of A$\beta$1-40 resulting in vascular dysfunction (Fig. 1B). Moreover, the transcription and expression levels of CD36 in pericytes treated with A$\beta$1-40 increased in a concentration-dependent manner aggravating vascular dysfunction [132]. A$\beta$1-40 increases the permeability of the BBB in vitro, which can be reversed by inhibiting the expression of CD36 in pericytes, suggesting that inhibiting the expression of CD36 increases BBB tightness [133], providing a new therapeutic target for preventing BBB destruction during AD progression.

Previous studies have shown that a decrease in the CBF is the earliest change in AD patients [19], and capillaries exhibit focal constriction [142]. Vascular resistance in the brain mainly occurs in capillaries, and the CBF is regulated by pericytes, indicating that pericytes dysfunction contributes to vascular disturbances in AD [22]. One of the most characteristic changes in AD is the aberrant deposition of A$\beta$, which results in the formation of A$\beta$ plaques in the brain parenchyma, suggesting that A$\beta$ may be the latent culprit.

In human brain slices, A$\beta$1-42 (oligomeric and monomeric) can trigger a slowly progressive constriction of capillaries near pericytes, suggesting that A$\beta$1-42 induces the constriction of human pericytes in a concentration-dependent manner within limits [36]. Reactive oxygen species (ROS) are generated by nicotinamide adenine dinucleotide phosphate (NADPH) oxidase and can be removed by superoxide dismutase 1 (SOD1) [134]. Recently, Nortley et al. [36] discovered that oligomeric A$\beta$1-42 evoked capillary constriction could be blocked by endothelin receptor type A (ET${}_{\text{A}}$-R), SOD1 and NADPH oxidase inhibitors, suggesting that ROS and endothelin (ET) participate in the constriction induced by A$\beta$1-42. Moreover, evidence has shown that ROS in pericytes are produced by reduced nicotinamide adenine dinucleotide phosphate oxidase 4 (NOX4) [135] rather than other isoforms. After ET was applied along with SOD1, the capillaries remained constricted, demonstrating that ET functions downstream of ROS. Moreover, ET combined with ET${}_{\text{A}}$-Rs on pericytes triggers intense constriction [143]. In summary, A$\beta$1-42 activates NOX4 in pericytes to generate ROS, and these ROS induce the generation of downstream ETs, which then interact with ET${}_{\text{A}}$-Rs on pericytes, triggering strong capillary constriction [36, 144]. The constriction of capillaries increases vascular resistance, causing a reduction in the CBF, which leads to a decrease in glucose and oxygen supplies, ultimately leading to synapse dysfunction and neuron loss [145]. In addition, hypoxia in turn upregulates the expression of BACE1, further increasing the generation of A$\beta$ and forming an amplified positive loop [143] (Fig. 1C). The discovery of this mechanism overturns the previous view that the reduction in the CBF is the result of arteriole constriction.

It has been observed that the level of ET increases in AD with the upregulation of enzymes responsible for synthesizing ET [136]. Similarly, compared to those in nondemented controls, Sengillo et al. [69] reported a remarkable loss of pericytes in the cortex and hippocampus of AD patients, which may be attributed to the fact that chronic exposure to A$\beta$1-42 results in the constriction and rigid death of pericytes. Fortunately, blocking NOX4 and ET${}_{\text{A}}$-Rs prevent further contraction caused by A$\beta$, although the capillary diameter does not return to baseline levels. Moreover, C-type natriuretic peptide (CNP) successfully reverses the A$\beta$-evoked constriction [136] and may be applied in the treatment of AD in the future.

Neural/glial antigen 2 (NG2), a transmembrane proteoglycan, is an original marker of pericytes [30]. NG2 not only plays a significant role in the proliferation and motility of pericytes, but also promotes the formation and maturation of endothelial TJs [38].

A$\beta$1-42 is the principal component of neuritic plaques characterized in AD. Different aggregated forms of A$\beta$1-42 have been proven to influence the shedding of NG2 from pericytes differently. NG2 sheds pericytes to form soluble NG2 (sNG2). Fibrillar A$\beta$1-42 decreases the level of sNG2, while the level of sNG2 increases after exposure to oligomer-enriched preparations of A$\beta$1-42 [137]. After exerting inhibitors of matrix metalloproteinase (MMP, an angiogenic factor secreted by pericytes) [138], the consequences of fibrillar A$\beta$1-42 remain, while increase in sNG2 resulting from oligomeric A$\beta$1-42 is eliminated, suggesting that the shedding of NG2 induced by A$\beta$ is mediated by MMP-9. Moreover, fibrillar A$\beta$1-42 decreases the activity of MMP-9 while oligomeric A$\beta$1-42 increased MMP-9 activity [137]. However, oligomeric A$\beta$1-42 does not alter the concentration of MMP-9, indicating that A$\beta$1-42 affects the activity rather than the secretion of MMP-9.

Taken together, in the early stage of AD, A$\beta$1-42 exists as oligomers and activates MMP-9, which subsequently increases the level of sNG2. The finding coincides with the discovery of increased MMP-9 in cerebrospinal fluid (CSF) during early AD pathogenesis [139]. sNG2 has been demonstrated to promote angiogenesis, resulting in unstable blood vessels and dysfunction of the BBB [146], and the level of sNG2 is increased in the CSF of AD patients [147]. In addition, the release of NG2 from the cell surface of pericytes drives pericyte detachment and contributes to the loss and dysfunction of TJs. Even worse, this effect could be aggravated by the enhanced degradation of the extracellular matrix and TJs caused by increased MMP-9 activity [148]. The detachment of pericytes and destruction of endothelial TJs destroy the integral structure of the BBB and increase the permeability of the BBB (Fig. 1D). Subsequently, the hazardous substances enter the brain and neurotoxicity caused by A$\beta$ could cause neuronal loss and cognitive decline. As the disease progresses, fibrillar A$\beta$1-42 accumulates, which may be the protective mechanism of the body [149].

Researchers found that the SIRT1 activator, resveratrol, reduced serum MMP-9 levels in AD patients, thus reducing neuro-inflammation [141]. Additionally, resveratrol was able to slow down the progressive decline in daily living scores (ADLs) in AD patients [141]. In spite of this, the relationship between the delay in cognitive decline and the decrease in MMP-9 remains unknown. The discovery that A$\beta$1-42 influences the shedding of NG2 on pericytes via MMP-9 may explain the pathology of AD and provide a new therapeutic strategy, such as specifically inhibiting MMP-9, to prevent AD dysfunction.

A$\beta$1-42 (10%) and A$\beta$1-40 (80%) are two typical soluble monomeric subtypes of A$\beta$ [49]. A$\beta$1-42 is more likely to form insoluble aggregates than A$\beta$1-40 [6], but A$\beta$1-40 can be deposited in the vascular system to form CAA [150]. Recently, it has been reported that more than 90% of patients with AD have CAA and CAA usually precedes the formation of neuritic plaques that are composed mainly of A$\beta$1-42 [151]. A$\beta$1-40 is presumed to have a particular toxicity to pericytes [152] and could accelerate the disruption of the BBB.

A recent study showed that pericytes treated with A$\beta$1-40 exhibit a prominent decrease in proliferation and a marked increas in mitochondrial ROS (Mito SOX) in a manner dependent on both time and concentration [132]. The results also demonstrated that A$\beta$1-40 activates mitochondrial autophagy through the CD36/PINK/Parkin pathway (PINK, PTEN-induced putative kinase) [132]. Mitochondrial damage and autophagy induced by A$\beta$1-40 often lead to apoptosis [140]. Notably, A$\beta$1-40 rather than apoptosis increases oxidative stress in pericytes. An increase in lipid ROS is accompanied by an increase in the concentration of iron ions, indicating that A$\beta$1-40 may induce pericyte ferroptosis [153]. The morphological features of pericytes exposed to A$\beta$1-40 according to transmission electron microscopy (TEM) also correspond with ferroptosis. Moreover, inhibiting mitochondrial autophagy prevents pericyte ferroptosis and ferroptosis inhibitors could prevent mitochondrial autophagy evoked by A$\beta$1-40, suggesting that ferroptosis is dependent on mitochondrial-related autophagy [132] (Fig. 1E).

The expression of pericyte markers is dynamic, and depends on the functional state of pericytes; for example, PDGFR$\beta$ labels relatively immature pericytes [33], NG2 stimulates the proliferation and migration of pericytes [34], and laminin is expressed in active or mature subsets of pericytes [154]. In the hippocampus of AD patients, a significant reduction in the number of pericytes expressing NG2 [37], CD13 and PDGFR$\beta$ [70] is observed, while the number of pericytes expressing other markers is not affected. It can be concluded that the subsets of pericytes that participate in activation, migration, and proliferation are affected by the pathology of AD. A previous study demonstrated that the CAA formed by A$\beta$1-40 is associated with pericyte degenerative changes [150]. Interestingly, A$\beta$1-40 levels are associated with the number of pericytes [37]. Since aggregated A$\beta$1-40 is toxic [150], it is speculated that monomeric A$\beta$1-40 may be beneficial.

Schultz et al. [37] showed that fibrillar A$\beta$1-40 reduces the viability and proliferation of pericytes in vitro, and increases the activity of the key apoptotic proteins caspase3/7. These findings are consistent with the discovery of pericyte degeneration near fibrillar A$\beta$1-40 [150]. Moreover, the monomer A$\beta$1-40 decreases the mortality of pericytes by decreasing the activity of caspase3/7, and promoting pericyte proliferation, which could explain the positive correlation between the levels of monomeric A$\beta$1-40 and the number of pericytes (Fig. 1F).

The occurrence of CAA and AD largely overlap [13, 14, 15] and the occurrence of advanced CAA is related to more severe cognitive impairment in patients with AD [15, 16]. Aggregated A$\beta$1-40, the major component of deposition in the CAA may account for the loss of pericytes in AD patients [155].