Mechanisms of high-density lipoprotein in regulating blood-brain barrier function: insights and implications

Abstract

High-Density Lipoprotein (HDL) not only mediates reverse cholesterol transport in the periphery but also significantly influences Blood-Brain Barrier (BBB) function within the central nervous system by regulating lipid metabolism, inflammatory responses, oxidative stress, Aβ trans-barrier clearance, and endothelial nitric oxide signaling. Variations in structural composition and biological properties between plasma HDL and brain-derived HDL-like particles enable them to perform both synergistic and distinct roles in maintaining cholesterol homeostasis, protecting tight junctions, regulating barrier permeability, facilitating Aβ efflux, and stabilizing the neurovascular unit. Key components such as ApoA-I, the ApoM/Sphingosine-1-Phosphate (S1P) signaling axis, and ApoE preserve BBB integrity by enhancing endothelial and pericyte functions, stabilizing intercellular junctions, inhibiting matrix metalloproteinase activity, and reducing neuroinflammation. Additionally, HDL/HDL-like particles alleviate barrier stress by facilitating the clearance of metabolic waste through the glymphatic system. Existing research collectively suggests that HDL-like particles, as a multifaceted regulator of BBB homeostasis, play a crucial role in maintaining neurological health and influencing disease processes.

Introduction

HDL, a class of multifaceted lipoprotein particles, has attracted increasing attention in recent years for its role in central nervous system health and disease. Beyond its traditionally recognized function in reverse cholesterol transport, HDL/HDL-like particles profoundly influence the structure and function of the BBB through multiple mechanisms, including regulation of endothelial lipid metabolism, suppression of inflammatory responses, alleviation of oxidative stress, and maintenance of cellular energy homeostasis. Extensive research shows that HDL/HDL-like particles not only interact directly with brain microvascular endothelial cells and pericytes through their major apolipoproteins and associated enzymes but also indirectly support BBB integrity and repair by modulating lipid metabolism and inflammatory responses in glial cells [1]. The dual role of plasma HDL and brain-derived HDL-like particles in both protecting and disrupting BBB function may provide a critical entry point for understanding the pathological processes in neurodegenerative diseases, stroke, and Cerebrospinal Fluid (CSF) metabolism disorders [2]. This article will systematically review the mechanisms by which HDL/HDL-like particles and their key components regulate BBB homeostasis and explore their potential applications in the prevention and treatment of neurological disorders.

Search strategy

In the process of writing this review, we conducted a systematic literature search. The primary databases searched were PubMed, with the search time frame set from January 2010 to April 2025. The search keywords included “high-density lipoprotein”, “brain”, “blood–brain barrier”, “apolipoprotein A”, “apolipoprotein M”, “sphingosine-1-phosphate”, “apolipoprotein E”, “apolipoprotein J”, “glymphatic system”, “neurodegenerative diseases”, “glial cell polarization”, and “pericytes” among others. Boolean logic (AND, OR) was used to combine the keywords during the search, with the search fields limited to title and abstract.

Structural and functional characteristics of plasma high-density lipoprotein

HDL is a highly heterogeneous and functionally diverse plasma lipoprotein. It consists of several HDL subfractions of different sizes and surface charges, which are found in the vicinity of most cells [3]. Ultracentrifugation is the traditional method for separating HDL subfractions based on their density into two main categories: HDL2, which is mainly lipid-rich and characterized by larger particles, and HDL3, which has a higher protein content and smaller particles [4]. Plasma spherical HDL particles likely share the same overall structure: a water-insoluble neutral lipid core mainly composed of cholesterol esters and triglycerides, surrounded by a surface monolayer of lipids (mainly phospholipids and some unesterified cholesterol), with apolipoproteins embedded within this layer [5]. Lecithin-cholesterol Acyltransferase (LCAT) mediated cholesterol ester buildup gradually occurs within disc-shaped HDL, eventually forming spherical HDL, which is the dominant form in normal human plasma [6].

ApoA-I, ApoA-II, ApoA-IV, ApoC-I, ApoC-II, ApoC-III, and ApoE mainly consist of tandem amphipathic α-helices. In the unlipidated state, they partially fold into helix bundles or dimers. When they bind to HDL, these helices align in a ribbon-like pattern along the phospholipid monolayer of HDL, thereby stabilizing the particles and mediating various metabolic functions [7]. About 70% of the total HDL protein in plasma is apolipoprotein ApoA-I, which is found in nearly all HDL particles. The second most abundant protein is ApoA-II, making up 15–20% of the total plasma HDL protein, but not all HDL particles contain ApoA-II. ApoA-I and ApoA-II are the main “scaffolding” proteins that determine the structure of HDL particles [8]. In human spherical LpA-I (with ApoA-I as the main scaffolding protein and without ApoA-II), ApoA-I supports a cage-like scaffold formed by three molecules arranged in a trefoil-like configuration. Discoid HDL particles typically exhibit a double-belt pairing [9]. When ApoA-II participates, it may disrupt ApoA-I pairing and localization, thereby increasing the freedom of inter-helix slippage, which results in a more continuous particle size distribution and increased heterogeneity [9]. In vitro experiments using human recombinant High-Density Lipoprotein (rHDL) particles revealed that LCAT activation involves adjacent segments from two ApoA-I molecules, a finding supported by the disc-shaped “double-band” model. Whether trefoil contains equivalent interfaces remains to be further validated [10]. Compared to ApoA-I and ApoA-II, ApoA-IV, ApoC, and ApoE are predominantly surface-bound and exchangeable components, capable of dynamic distribution between HDL and Triglyceride-rich Lipoproteins (TRL, including chylomicrons and VLDL) [11–13]; ApoM anchors to the phospholipid monolayer on the HDL surface via its unprocessed N-terminal signal peptide, delaying ApoM export and promoting the formation of large-particle HDL enriched with ApoM/S1P [14]. Regarding particle size heterogeneity, the ApoA-I copy number and banded arrangement determine the fundamental HDL particle size and mediate the maturation transition from discoid to spherical particles [8]. ApoA-II synergistically enhances particle stability, favoring the generation of smaller, denser HDL3 particles [15]; ApoE is frequently enriched in larger, cholesterol ester-rich HDL2 particles and is involved in particle receptor/matrix interactions and in vivo clearance [16]. ApoA-IV is primarily localized to disc-shaped HDL during fasting and possesses receptor adhesion capabilities [17]. The stability and exchangeability of HDL surface apolipoproteins are primarily governed by intrinsic factors (the hydrophobic surface area and length of amphipathic α-helices, multivalent binding/oligomeric states, specific anchoring motifs, flexible disordered regions, and post-translational modifications) and extrinsic factors (particle curvature and lipid composition, surface tension, and remodeling processes such as ATP-Binding Cassette Subfamily A Member 1, ABCA1/LCAT/ Cholesteryl Ester Transfer Protein, CETP/ Phospholipid Transfer Protein, PLTP/Phospholipid Transfer Protein) [6, 18].

Proteins at lower concentrations on HDL can be further divided into lipoprotein-specific proteins and auxiliary proteins. Lipoprotein-specific proteins include LCAT, CETP, and Paraoxonase 1 (PON1) (each present on average fewer than one copy per particle) [5]. Auxiliary proteins are more heterogeneous in type and abundance, although they are present in smaller amounts and are mostly loosely associated with HDL [19, 20]; These proteins play key roles in infection defense, inflammatory response, hemostasis, and tissue repair, including α-1-antitrypsin, α-2-HS glycoprotein, C-reactive Protein (CRP), complement C, heme-binding protein-related protein, Platelet-activating Factor Acylhydrolase (PAF-AH), and Glutathione Peroxidase 3 (GPx3) [21, 22]. As of now, 285 proteins have been identified within HDL, and this number is still expanding as more proteins are identified (https://homepages.uc.edu/%7Edavidswm/HDLproteome.html). The lipid components carried by HDL mainly include phospholipids, sphingomyelins, steroids, cholesterol esters, triglycerides, and trace amounts of diverse rare lipids. Among these, phosphatidylcholine exhibits molecular diversity, with a relatively high proportion of polyunsaturated fatty acids (e.g., phosphatidylcholine 36:4), which may provide stronger antioxidant and anti-inflammatory effects [23]. Additionally, HDL also carries lipid-soluble antioxidants such as tocopherols, carotenoids, and Coenzyme Q10, along with various lipophilic vitamins and antioxidant molecules. These components work together to confer HDL's unique biological activity [24–27].

The biological effects of HDL are mediated by the entire HDL particle or by specific proteins or lipids carried by HDL, rather than being determined by the plasma cholesterol content of HDL [3]. Based on human genetic and functional evidence of Scavenger Receptor Class B Type I (SR-BI) deficiency, HDL-C, as a static concentration metric, cannot adequately replace assessments of HDL flux or function. Its variability is influenced not only by HDL particle size and quantity but also by SR-BI-mediated selective uptake and regulation of the RCT metabolic pathway [28]. Therefore, compared to the traditional focus on HDL-C levels, evaluating the functional state of HDL may better reflect its true role in disease progression [29, 30].

HDL-like particles in the central nervous system

HDL-like particles specific to the brain not only serve as key carriers for cholesterol homeostasis in the central nervous system but also perform multiple biological functions related to neuroprotection, Aβ clearance, and neuroinflammation regulation [31].

Research has shown that human CSF lipoproteins form spherical particles, while secretory granules from rat primary astrocytes consist of immature disc-shaped particles lacking a core lipid layer [32]. Furthermore, HDL-associated enzymes and membrane proteins, such as LCAT and PLTP, are found in CSF and demonstrate biological activity [33, 34], ABCA1 mRNA expression has been observed in human central nervous system tissues [35]. In vivo and in vitro mouse studies suggest that ABCA1 is a crucial regulator of ApoE esterification and secretion within the central nervous systems [36, 37]. Several studies have detected LCAT and its cholesterol esterification activity in CSF, confirming that cholesterol esters are generated in CSF lipoproteins in both humans and animal models, with about 60–70% of cholesterol in CSF present in esterified form [33, 38, 39], suggesting that mature HDL-like particles may contain a neutral lipid core mainly composed of cholesterol esters. CSF predominantly contains ApoE-enriched HDL-like particles (CSF-Lp) [40]. Density and particle size fractionation, as well as proteomics studies of human CSF, revealed a dominant size band concentrated between 13 and 22 nm [41]. Separation experiments showed that the main peak for total cholesterol/phospholipids highly overlapped with ApoE distribution, suggesting that ApoE serves as the primary scaffold for these particles [42]. In contrast, ApoA-I signals were weaker and did not fully coincide with the lipid peak. At the same time, a CSF-LpA subpopulation (13–18 nm), with particle sizes similar to plasma HDL, exists, mainly composed of ApoA-I/ApoA-II (although its overall abundance in CSF is much lower than in plasma). Smaller sCSF-Lp (10–12 nm) subpopulations contain negligible ApoA-I/II [41]. Overall, although ApoA-I/ApoA-II is also present in CSF-Lp and some particles structurally resemble plasma HDL, a larger proportion use ApoE as the main scaffold. Their size distribution ranges roughly from 8 to 22 nm, with the majority of particles between 13 and 22 nm [43]. This paper collectively refers to them as “HDL-like particles (CSF-Lp).”

In mouse models, the normal brain microvascular endothelium depends on Major Facilitator Superfamily Domain Containing 2A (MFSD2A) to maintain extremely low levels of transcytosis, particularly by inhibiting caveolae-mediated transcytosis and vesicular trafficking, thereby restricting the entry of plasma macromolecules such as albumin [44, 45]. This mechanism might also limit the crossing of large particle complexes, such as intact HDL. Although BBB endothelium does express several lipoprotein receptors (e.g., SR-BI, Low-Density Lipoprotein Receptor-Related Protein 1/LRP1), existing evidence suggests that SR-BI's primary role is mediating selective HDL lipid uptake rather than whole particle endocytosis [46]. In vitro experiments show that ApoA-I/HDL exhibits clathrin-independent, caveolin-independent, and cholesterol-dependent transmembrane transport in brain endothelial cells, reaching saturation. While ApoA-I has been shown to cross the BBB in animal models. However, the ability of HDL to cross the BBB in vivo remains unconfirmed by experimental evidence [47, 48]. LRP1 has been extensively demonstrated to participate in Aβ clearance across the BBB and endothelial cells homeostasis [49]. Endothelial-specific models clearly show that at physiologically relevant concentrations (0.05–5 nM), LRP1-deficient primary cerebral microvascular endothelial cells demonstrate significantly reduced transmembrane transport of [^{1 2 5} I]-Aβ₁₋₄₂, with about a 50% reduction at 0.1 nM. Furthermore, the relative contribution of LRP1 increases at higher concentrations. In vivo, deletion of LRP1 in brain endothelial cells significantly impaired brain-to-blood Aβ efflux, with BBB-dependent efflux reduced by approximately 48% at 15 minutes and about 27% of total clearance mediated by endothelial LRP1, and in 5xFAD models, it impaired spatial learning and memory deficits, suggesting that LRP1-mediated trans-BBB clearance plays a crucial role in maintaining Aβ homeostasis in the brain [50]. Studies have shown that the rapid transcellular transport of Aβ across the BBB relies on the synergistic interaction between LRP1 and ABCB1/P-glycoprotein (P-gp). Following Aβ-LRP1 endocytosis, Phosphatidylinositol Clathrin Assembly Lymphoid Myeloid Leukemia (PICALM) directs the complex to Rab5/Rab11-positive sorting endosomes, where it associates with ABCB1/P-gp to facilitate relay-style transmembrane transport of Aβ. Immunoprecipitation, colocalization, and dual inhibition experiments all demonstrate that LRP1 and ABCB1/P-gp are functionally tightly linked, with both being indispensable for Aβ clearance [51]. Furthermore, LRP1 influences Aβ generation by promoting the endocytosis of Amyloid Precursor Protein (APP). In mice with inactivated NPxY2 endocytic motif (impairing overall LRP1 endocytic function), despite suppressed Aβ clearance, APP processing significantly shifted from the β-secretase pathway to the α-secretase pathway. This manifested as elevated sAPP-α, reduced soluble Aβ, and decreased plaque burden, suggesting that LRP1 exerts dual “clearance + production” effects on Aβ metabolism, with the reduction in production dominating in this model [52]. Although LRP1 is expressed on BBB endothelial cells, there is no direct evidence showing that it mediates the high-throughput crossing of ApoA-I or whole HDL across the BBB. However, LRP1 is not just an endocytic receptor. Its signaling function is equally important for maintaining BBB homeostasis. Studies indicate that mice with brain-endothelial cell-specific LRP1 knockout show significantly increased BBB permeability, along with higher cyclosporine A levels, increased Matrix Metalloproteinase-9 (MMP-9) activity, degradation of tight junction proteins such as claudin-5, and reduced P-glycoprotein expression. These findings suggest that LRP1 helps maintain endothelial tight junctions by inhibiting the CypA-MMP-9 pathway and acts as a co-activator of PPARγ to regulate the expression of ABC transporters such as P-glycoprotein [53].

Due to the strict permeability restrictions of the BBB, the lipoprotein composition in the nervous system differs significantly from that in peripheral blood [54, 55]. According to Koch et al., in individuals with an intact BBB, CSF contains approximately 0.26% of plasma ApoA-I (about 1/385) and 4.4% of plasma ApoE (about 1/23). Regarding lipids, the total cholesterol in CSF is approximately 0.3% of plasma levels, and total fatty acids are about 0.54% (without distinguishing CE/FC; CSF triglycerides are present in trace amounts [41]; apolipoprotein levels in CSF account for only a very small proportion of plasma levels (typically less than 1%), and the overall content is much lower than that in peripheral blood.

In terms of lipid composition, in comparison to plasma HDL, the ratios of phospholipids and sphingolipids in HDL-like particles differ, especially since HDL-like particles contain higher levels of neuron-specific lipids, such as sphingomyelin, which are less common in plasma HDL [56]. Lipoproteins in the CSF are mainly composed of ApoE, ApoA-I, and ApoJ [42], followed by smaller amounts of ApoD, ApoA-II, ApoA-IV, ApoC, and ApoH [57]. ApoD [58], ApoH [59], ApoJ [60] can be synthesized in the brain. ApoC-I may also be synthesized by astrocytes in the brain [61]. The current consensus suggests that ApoB is absent from CSF, as it is believed to be unable to cross the BBB [43]. This suggests that the central nervous system may lack lipoproteins like peripheral Chylomicrons (CMs), Very Low-density Lipoproteins (VLDL), and Low-density Lipoproteins (LDL). Although experimental evidence shows that ApoA-I can enter the brain in small amounts via receptor-mediated transcellular transport, its flux remains extremely low. The specific mechanism will be detailed in the ApoA-I section. According to the Human Protein Atlas data, the mRNA and protein expression of ApoA-II and ApoA-IV in human brain tissue are extremely low or undetectable, while they are highly enriched in liver tissue [62]. According to Elliott et al., peripherally derived ApoA-II may enter the CSF via the choroid plexus [63]. To date, no in vivo or in vitro experiments have directly demonstrated that ApoA-II, ApoA-IV, ApoC-II, or ApoC-III can cross the BBB via specific receptor-mediated transport mechanisms. Sensitivity limitations and sample volume significantly limit the detection and characterization of HDL-like particles and their apolipoproteins in CSF, as their levels are generally much lower than in peripheral blood, substantially increasing research challenges. Therefore, the composition, structure, and functional characteristics of apolipoproteins in human CSF remain incompletely elucidated. Existing evidence suggests that HDL-like particles primarily depend on de novo synthesis in the brain, with a minor contribution from peripheral apolipoproteins [40].

ApoE, the primary apolipoprotein of lipoproteins resembling HDL, can be synthesized independently in both the peripheral and central nervous systems [64]. However, ApoE synthesized in the liver cannot cross the BBB [63]. Based on large-scale human brain monocyte transcriptomic maps and specialized human brain vascular cell atlases, multiple brain cell types possess ApoE synthesis capacity [65, 66]. Astrocytes are the main source of ApoE. Neurons, microglia, and perivascular macrophages exhibit low expression levels under homeostasis but show significant upregulation under inflammatory or pathological conditions (e.g., Alzheimer’s disease, brain injury). Endothelial cells, pericytes, vascular smooth muscle cells, and vascular stromal cells (e.g., meningeal and vascular-associated fibroblasts) also express ApoE, albeit at lower levels [67]. Although these cells contribute much less than astrocytes, they may still participate in local ApoE production under certain conditions, suggesting that ApoE sources may vary by cell type and pathological state. In vitro animal experiments revealed substantial differences in secretion levels between astrocytes and microglia. Researchers isolated and cultured primary astrocytes and microglia from mice with a targeted replacement of the human ApoE allele. They found that astrocytes secreted at least six times more ApoE than microglia, and astrocytes carrying the ApoEε2 allele secreted more ApoE than those carrying the ApoEε4 allele [68]. When isolating mouse perivascular cells carrying ApoEε3 or ApoEε4 for in vitro culture and co-culture with HUVECs, results showed that perivascular cells secreted approximately 25–30% of the lipidated ApoE produced by astrocytes. Furthermore, perivascular cells carrying ApoEε4 displayed a reduced capacity to induce endothelial expression of basement membrane proteins [69]. In vitro studies using human induced pluripotent stem cell (iPSC)-derived brain microvascular endothelial cells showed that ApoEε4-carrying endothelial cells secreted about 15% less ApoE than ApoEε3-carrying cells [70]. In APOEε3/ε4 knock-out mice and bone marrow transplant models, researchers confirmed that ApoEε4-positive Border-associated Macrophages (BAMs) are the key source and effector cells responsible for neurovascular dysfunction [71]. Single-cell transcriptomic analysis of human cerebral vascular samples revealed significant ApoE expression in smooth muscle cells and meningeal fibroblasts [72]. In vivo animal studies have demonstrated that neurons do not typically express ApoE but induce robust expression following excitotoxic injury (e.g., kainic acid treatment). Moreover, most microglia do not regularly express ApoE, with only about 6–10% expressing it after injury. In contrast, vascular smooth muscle cells and non-astrocytic cells surrounding small vessels stably express ApoE under normal conditions [73–75]. Different ApoE subtypes have a significant impact on the function of HDL-like particles. For instance, ApoE4 exhibits markedly reduced cholesterol transport efficiency [76], receptor affinity, and anti-inflammatory/antioxidant activity compared to ApoE3 and ApoE2 [77], and is closely associated with a high risk of neurodegenerative diseases such as Alzheimer’s disease [78–80]. Abnormal alterations in HDL-like particles have been observed in various neuropathological conditions, such as stroke, traumatic brain injury, Alzheimer’s disease, and other neurodegenerative disorders, where both particle abundance and function are significantly impaired [81–83]. This functional impairment of HDL-like particles not only disrupts cholesterol homeostasis but may also worsen neuroinflammation and BBB dysfunction, thereby promoting disease progression [84–86]. (See Table 1).

Table 1.

Comparative characteristics of plasma HDL and brain HDL-like particles

Category	Plasma HDL	Brain HDL-like particles
Main apolipoproteins	ApoA-I, ApoA-II, ApoC-III, ApoC-I, ApoE	ApoE, ApoJ, ApoA-I, ApoD, ApoA-IV, ApoH
Major lipid components	Predominantly CE and PL; low TG and SL	Predominantly FC and PL; extremely low TG; abundant SL and 24S-OHC
Particle size range	8–12 nm (HDL3 smaller, HDL2 larger)	8–22 nm
Localization	Circulating in peripheral blood	Present in cerebrospinal fluid
Concentration	HDL-C typically 1.0–2.0 mmol/L in adult plasma	Approximately 1–10% of plasma HDL level
Structural type	Similar morphology; heterogeneous subtypes (HDL2, HDL3); ApoA-I as scaffold	Similar morphology; ApoE as major scaffold; more heterogeneity in structure and maturation
Primary functions	Reverse cholesterol transport; anti-atherosclerosis	Redistribution of brain cholesterol; Aβ clearance; neuroprotection
Enzymatic systems	Highly expressed LCAT, CETP, PLTP; also LPL and EL	LCAT in ~5%; CETP unclear; PLTP ~15%; LPL and EL nearly absent

Abbreviations: CE: cholesteryl ester; PL: phospholipid; FC: free cholesterol; TG: triglyceride; SL: sphingolipid; 24S-OHC: 24S-hydroxycholesterol; LCAT: lecithin–cholesterol acyltransferase; CETP: cholesteryl ester transfer protein; PLTP: phospholipid transfer protein; LPL: lipoprotein lipase; EL: endothelial lipase; CSF: cerebrospinal fluid

Synthesis and sources of cholesterol in the nervous system

The nervous system is the organ with the highest lipid content, with lipids constituting approximately 50–60% of its dry weight, of which 20–25% is cholesterol [87]. Cholesterol synthesis within the nervous system exhibits significant cell specificity [88], mainly carried out by astrocytes and oligodendrocytes, each with distinct functions [89]. Astrocytes serve as the main source of cholesterol in the adult brain [90]. From the perinatal period to adolescence, oligodendrocytes exhibit the highest cholesterol synthesis activity, with most of the cholesterol produced being used for myelin sheath formation [91]. Myelin is a cholesterol-rich membrane, with lipids making up about 70–85% of its dry weight, and cholesterol accounting for a significant proportion [92].

Although the BBB prevents the free transport of cholesterol molecules and lipoproteins between the central and peripheral systems, cholesterol metabolism in the brain is not completely isolated. This happens through oxysterols, which can partially cross the BBB, especially centrally derived 24S-hydroxycholesterol (24S-OHC, often abbreviated as 24-OHC in the literature) and peripherally derived 27S-hydroxycholesterol (27S-OHC, often abbreviated as 27-OHC) [93]. 24S-OHC is primarily synthesized by cholesterol 24-hydroxylase (CYP46A1) expressed in neurons within the central nervous system [94]. 24S-OHC freely diffuses across an intact BBB into the circulation, acting as the primary pathway for cholesterol clearance from the brain. Studies show that in healthy adults, approximately 6–7 mg of cholesterol is excreted daily from the brain as 24S-OHC. The concentration of intracerebral 24S-OHC is significantly higher than that in plasma, creating a concentration gradient that drives its efflux across the BBB, establishing 24S-OHC as a key metabolic pathway for maintaining brain cholesterol homeostasis [95].

Although neurons and glial cells also express small amounts of CYP27A1, its levels are extremely low, thus contributing minimally to brain synthesis. 27-OHC is mainly generated from cholesterol by sterol 27-hydroxylase (CYP27A1) in peripheral tissues and is the most abundant oxygenated sterol in circulation [96]. Research shows that 27-OHC, as a lipophilic side-chain oxidized cholesterol, can passively diffuse across an intact BBB into brain tissue via concentration gradients [95, 97]. However, in the presence of an intact BBB, 27-OHC entry efficiency remains low, resulting in limited brain levels [93]. The reverse flux of 27-OHC from the brain to circulation may regulate numerous key enzymes within the brain [98].

Cholesterol cannot directly cross the BBB due to its very low water solubility. However, its metabolites, such as 24S-OHC and 27-OHC, are believed to primarily cross the BBB via passive diffusion driven by concentration gradients. This is because they retain lipophilicity while introducing a side-chain hydroxyl group, enhancing their amphiphilicity [99, 100]. Some researchers have suggested that specific transporters may be involved in this process, though no direct evidence has been found yet [101, 102]. As a result, brain tissue depends on internal cholesterol production and efficient transport between cells to maintain functions like membrane stability, synaptic plasticity, and myelin repair [91]. Furthermore, in an in vitro BBB model using porcine brain microvascular endothelial cells, studies have observed SR-BI involvement in the uptake and transendothelial transport of vitamin E carried by peripheral HDL [103–105]. This finding suggests that SR-BI may play a role in the passage of vitamin E across the BBB. However, due to insufficient evidence from human brain microvascular endothelial cells, its specific mechanism in humans requires further validation. Vitamin E, especially α-tocopherol, is a lipid-soluble antioxidant. Extensive literature has documented its role in neutralizing free radicals and protecting neuronal membrane lipids from oxidative damage; further elaboration is omitted here [106–109].

Mechanisms of intercellular cholesterol transport in the nervous system

In the central nervous system, ApoE produced by astrocytes is initially transported through the ER-Golgi pathway and released outside the cell via constitutive secretion [110], the degree of lipidation of ApoE in the endoplasmic reticulum and Golgi apparatus remains unclear [111]. However, even without ABCA1, small and poorly lipidated ApoE particles can still be secreted. These particles later undergo lipidation on the outer leaflet of the plasma membrane, coupled with ABCA1-mediated cholesterol and phospholipid efflux, and form disc-like HDL particles [111]. Evidence shows that ApoE secretion is coupled with endocytosis, recycling, and resecretion. Short-term in vitro experiments show that ABCA1 deficiency significantly decreases ApoE levels in the culture medium and its secretion rate [37]. Subsequently, ATP-Binding Cassette Subfamily G Member 1 (ABCG1) becomes predominant in astrocytes, while ABCG4 is more prominent in neurons [112]. These cells efflux cholesterol, enriching HDL-like particles and enhancing cholesterol transport [113]. Newly formed disc-shaped HDL-like particles bind to apolipoproteins, free cholesterol, and phospholipids. With the involvement of LCAT and PLTP, they gradually mature into fully developed HDL-like particles [114]. During this maturation process, cholesterol metabolites like 24-OHC can act as endogenous agonists to activate Liver X Receptors (LXRs). LXRs heterodimerize with Retinoid X Receptors (RXRs), bind to LXR response elements on target gene promoters, and induce the expression of key genes, including ABCA1, ABCG1, and ApoE, thereby boosting cholesterol efflux capacity [115]. ABCA1 activity is also regulated by a cAMP-dependent protein kinase A (PKA) pathway. PKA enhances ABCA1 gene transcription and phosphorylates the ABCA1 protein, increasing its ability to efflux cholesterol [116]. HDL-like particles locally diffuse through the CSF to adjacent cells [117]. Mature HDL-like particles deliver cholesterol to target brain cells through interactions between ApoE and specific lipoprotein receptors [118]. These receptors include the Low-density Lipoprotein Receptor (LDLR), LRP1, Very Low-density Lipoprotein Receptor (VLDLR), and Apolipoprotein E Receptor 2 (ApoER2) [119]. Although LDLR and LRP1 are expressed in both neurons and astrocytes [120], LDLR expression is higher in glial cells. In contrast, LRP1 expression is higher in neurons [31]. Intracellular HDL-like particles undergo endosomal and lysosomal processing, where ApoE dissociates from lipids for recycling. Simultaneously, cholesterol is transported to cell membranes and organelles to support membrane structure and cellular function, forming local cholesterol transport pools [121]. SR-BI also mediates cholesterol transfer between cholesterol esters and neurons or astrocytes, providing a selective cholesterol exchange mechanism independent of HDL-like endocytosis [122], as illustrated in Fig. 1.

Fig. 1.

Cholesterol transport in the central nervous system. Astrocytes secrete ApoE through the ER-Golgi pathway, with initially low lipidation of the particles. These particles are subsequently lipidated with the help of ATP-binding cassette transporters, forming disc-like HDL-like particles. In the extracellular matrix, these particles gradually mature into spherical HDL particles under the action of enzymes such as lecithin-cholesterol acyltransferase. They then bind to receptors on brain cell surfaces (such as the low-density lipoprotein receptor, low-density lipoprotein receptor-related protein 1, etc.), facilitating cholesterol transport

Cholesterol clearance and homeostasis mechanisms in the nervous system

Cholesterol clearance and homeostasis in the nervous system primarily rely on three mechanisms: ACAT1 (Acyl-CoA: Cholesterol Acyltransferase 1) mediates the esterification of free cholesterol and its storage as cholesterol esters [123], the ABCA1/ApoE pathway mediates cholesterol efflux [124], neuron-specific cholesterol 24-hydroxylase (CYP46A1) converts cholesterol into 24S-OHC, which then crosses the BBB into peripheral circulation [125]. These mechanisms work synergistically to effectively maintain cholesterol homeostasis in the nervous system and prevent toxic accumulation.

Mechanisms of HDL-BBB interaction

Blood-brain barrier

The BBB is a critical structure maintaining homeostasis within the central nervous system. It is composed of the Neurovascular Unit (NVU), primarily consisting of cerebral microvascular endothelial cells, pericytes, tight junctions, basement membranes, and astrocytic foot processes [126], as illustrated in Fig. 2.

Fig. 2.

The structure of the BBB

BBB dysfunction is a common and critical pathogenic mechanism in sepsis-associated encephalopathy, stroke, and neurodegenerative diseases [127–129]. HDL, due to its multiple biological properties, including regulation of lipid metabolism, modulation of inflammatory responses, antioxidant effects, and promotion of endothelial function, is increasingly recognized for its potential protective role in the central nervous system [130]. Numerous studies indicate that elevating HDL levels or function in stroke and neurodegenerative diseases can reduce neuroinflammation while protecting cerebral vascular integrity and maintaining neural tissue structure [131, 132]. Accumulating evidence suggests HDL/HDL-like particles’ benefits extend beyond the cardiovascular system to modulate inflammation and vascular function within the nervous system.

Regulatory mechanisms of HDL-like particles on the blood-brain barrier during abnormal cholesterol loading

Based on the above understanding, we first explore the regulatory mechanisms of HDL-like particles on BBB function under conditions of cholesterol metabolism imbalance. Abnormal cholesterol metabolism can damage the BBB through multiple pathways. Under conditions of ischemia-hypoxia, inflammation, and oxidative stress, excessive cholesterol accumulation can induce metabolic reprogramming in microglia, accompanied by enhanced signaling of PLIN2 and Sterol Regulatory Element-Binding Proteins (SREBPs, primarily SREBP2 in ischemia; SREBP1 is upregulated in certain inflammatory/neurodegenerative models) [133, 134], as well as altered cholesterol esterification mediated by ACAT1 [135]. Notably, SREBP activation exhibits time-window and model-dependent patterns: SREBP2 upregulation is common in adult ischemia models [136], whereas transient downregulation of SREBP1/2 proteins is observed in early neonatal mouse ischemia-hypoxia [137]. This change promotes inducible nitric oxide synthase and mitochondrial Reactive Oxygen Species (mtROS) production, exacerbating cellular oxidative stress and driving microglia toward a pro-inflammatory phenotype [138, 139]. Furthermore, cellular stress from oxidative stress, ischemia, and Damage-Associated Molecular Patterns (DAMPs) induces NLRP3 inflammasome assembly in microglia and brain microvascular endothelial cells [140, 141], activated Caspase-1 cleaves Gasdermin D (GSDMD), forming membrane pores that trigger pyroptosis, leading to endothelial cell loss and disruption of the BBB [142]. Concurrently, Caspase-1 activation promotes the maturation and release of IL-1β and IL-18, thereby amplifying the local inflammatory response [143]. These cytokines and pyroptosis-released products also stimulate the MMP-9, which degrades tight junction proteins ZO-1 and Claudin-5. This significantly increases BBB permeability, leading to cerebral edema and triggering inflammatory cascades [144]. Notably, the abnormal accumulation of cholesterol in endothelial cell membranes can remodel membrane lipid rafts, thereby altering the physical properties of the membrane. This process provides priming signals for NLRP3 inflammasome assembly by activating TLR4/NF-κB, thereby influencing the inflammatory state of microglia and astrocytes [145], forming a metabolic disorder–inflammation positive feedback loop [146–148], which further impairs BBB function [149].

High expression of Trigger Receptor 2 (TREM2) on microglia promotes lipoprotein uptake/processing by binding to HDL-like particles, thereby reducing the lipid droplet burden and indirectly enhancing cholesterol handling capacity [150, 151], its downstream signaling primarily regulates microglial state via the DAP12-SYK-PI3K/AKT-mTOR axis [152, 153]. When ApoE- or TREM2-mediated cholesterol transport pathways are impaired (e.g., ApoEε4 or TREM2 deficiency) [154], transcriptional enrichment of inflammatory pathways like JAK/STAT is commonly observed, correlating with pro-inflammatory tendencies. This is often accompanied by worsened BBB permeability in some models [155, 156]. BBB disruption affects the relative homeostasis of cholesterol within the brain [157], allowing peripheral 27-OHC to more easily enter the brain while increasing the efflux of brain-derived 24S-OHC [158]. Since both oxidized cholesterol species are potent inhibitors of cholesterol synthesis, their abnormal flux exerts negative feedback regulation on intracerebral cholesterol synthesis [159, 160], disrupting central cholesterol homeostasis. Additionally, functional HDL-like particles facilitate lipid transport and redistribution within the brain through cholesterol transport proteins and receptors, helping to maintain cerebral lipid balance [161]. In animal models, ApoA-I/mimetic peptide intervention is linked to upregulation of ABCA1, reduced lipid droplets, decreased inflammation, and improved BBB structure and function [162]. However, these mechanisms and effects need more thorough validation in humans. Given the diverse roles of HDL-like particles, we will now further explore their impact on the BBB.

Effects of different ApoE subtypes on blood-brain barrier function

After elucidating the mechanisms by which HDL-like particles counteract cholesterol metabolism disorders, we shift our focus to ApoE, the key apolipoprotein of HDL-like particles, to explore how different ApoE subtypes influence BBB function. Since the discovery of the association between APOEε4 and Alzheimer’s disease (AD) in the 1990s [163], numerous studies have yielded exciting advances: Compared to the most common ApoEε3, ApoEε4 represents a significant risk factor for Alzheimer’s disease, while ApoEε2 exhibits relative protective effects [164–166]. However, evidence from the past decade indicates that ApoE’s influence extends beyond classical amyloid pathology, being closely associated with neuroinflammation, energy metabolism disorders, myelin loss, and altered BBB permeability [167, 168]. Imaging and biomarker studies show that ApoE-related BBB dysfunction can exist statistically independently of Aβ and tau markers [169–171], Multiple observational cohort studies in imaging research indicate that ApoEε4 carriers exhibit BBB dysfunction even in cognitively normal individuals without overt Aβ pathology [172]. Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) studies further reveal elevated cortical gray matter BBB permeability in cognitively normal, amyloid-negative elderly ApoEε4 carriers, suggesting ApoE4-related BBB disruption may occur before cognitive decline or amyloid pathology [173].

Early pathological observations suggest that AD patients exhibit thinning and discontinuity of the microvascular basement membranes, accompanied by leakage of plasma proteins (e.g., prothrombin). BBB leakage is more prevalent in individuals carrying at least one ApoEε4 allele [174]. Mechanistically, in animal and in vitro models, functional ApoE2/3 (including endogenous mouse ApoE) is primarily secreted by astrocytes and binds to LRP1 receptors on the surface of brain microvascular endothelium and pericytes, activating downstream signaling to maintain BBB integrity [175]. This pathway blocks the Cyclophilin A (CypA)-NF-κB-MMP-9 inflammatory cascade, thereby preventing excessive activation of MMP-9) and its potential damage to cerebral microvessels [176]. MMP-9 hydrolyzes major components of the capillary basement membrane (e.g., type IV collagen) and degrades tight junction proteins (ZO-1, Occludin, Claudin-5). Its excessive activation significantly increases BBB permeability, leading to cerebral edema and a cascade of inflammatory responses [176]. Without normal ApoE signaling inhibition, the ApoEε4 background may impair LRP1 expression or function, causing failure of CypA-NF-κB-MMP-9 pathway suppression. This leads to sustained overactivation of the pathway in pericytes and endothelial cells, thereby exacerbating breakdown and dysfunction of the BBB [177]. Research by Montagne et al. shows that ApoE4 induces BBB damage by activating the CypA–MMP-9 pathway, and elevated soluble Platelet-derived Growth Factor Receptor β (sPDGFRβ) in CSF reflects the severity of pericellular injury, correlating closely with BBB leakage and cognitive decline [170]. Storck et al. further demonstrated, using a Slco1c1-CreERT2-induced LRP1 brain endothelial cell knockout model, that LRP1 deficiency leads to enhanced MMP-9 activity, a significant reduction in tight junction proteins such as Claudin-5 and P-gp, accompanied by increased CSF IgG levels and cognitive decline [53]. This indicates that the ApoE/LRP1 signaling axis plays a critical role in BBB protection by inhibiting the CypA-MMP-9 pathway and maintaining P-gp levels.

In mouse models, ApoE4-mediated BBB impairment exhibits an astrocyte-specific “harmful acquired function” [178]. Jackson et al. established humanized ApoEε4 mice, revealing that astrocytic expression of ApoE4 alone is sufficient to upregulate intracerebral MMP-9 levels, disrupt tight junctions, and reduce astrocytic terminal coverage of blood vessels—directly increasing BBB permeability. Selective knockout of ApoEε4 in astrocytes reverses these impairments [179]. This finding is primarily based on animal models and requires further validation in human populations. Recent studies suggest ApoE4 may induce abnormal deposition of perivascular fibronectin (Fibronectin 1, FN1), disrupting astrocyte-endothelial cell intercellular communication; these findings stem mainly from early genetic clues and model studies, awaiting further confirmation [180]. This suggests that ApoE4 not only affects individual cells but also weakens neurovascular unit coupling by altering the composition of the intercellular matrix.

At the metabolic level, ApoE4 significantly disrupts astrocytic lipid homeostasis: ApoE4 astrocytes exhibit excessive and abnormally large unsaturated lipid droplets, showing disordered triglyceride composition and size, and increased susceptibility to lipid peroxidation [181, 182]. Concurrently, ApoE4 induces cerebral cholesterol accumulation, activating the C/EBPβ-δ-secretase (also known as δ-secretase or AEP) pathway, which triggers oxidative stress and neuroinflammation. Current research suggests that this pathway may indirectly affect BBB stability through neuroinflammation and neuronal injury [183].

In endothelial cells, the ApoE4 background induces a pro-inflammatory transcriptional profile in the brain endothelium, accompanied by an enhanced propensity for leukocyte adhesion/transmigration, and impaired tight junction regulation [184]. ApoE allele differences are also involved in the regulation of the PKCθ–occludin pathway, this effect synergizes with chronic inflammation to further compromise BBB integrity [185].

Regarding macromolecular transport, ApoE alleles influence the crossing of metabolic waste and harmful proteins across the BBB. Clinical and animal studies indicate reduced Aβ efflux efficiency in ApoEε4 carriers, with potential increased activity of brain entry pathways such as the Receptor for Advanced Glycation End products (RAGE). However, these findings are significantly affected by variations in models and assay methods. However, these findings are significantly influenced by model and assay variations [186]. Human-derived iPSC-BBB models provide direct evidence: compared to other alleles, exogenous ApoE4 significantly impairs Aβ40 transport from the cerebral to the blood side, while ApoE2 enhances Aβ42 clearance. This mechanism involves the differential function of endothelial LRP1 and P-gp [187]. Regarding lipid transport across the BBB, existing evidence indicates inconsistent relationships between ApoE4 and brain lipid uptake or metabolism across species, molecular forms, and methodologies: Small sample cohort studies observed higher baseline CSF free fatty acids in ApoEε4 carriers, who also exhibited distinct responses to intravenous triglyceride loading compared to non-ApoEε4 carriers [188]. Perfusion studies in humanized mice found that ApoE4 reduces the passage of polyunsaturated fatty acids, such as Docosahexaenoic Acid (DHA) [189]. These differences suggest that ApoE4 may further influence the BBB by altering blood lipid levels and lipid transport pathways; however, confirmation requires large-scale studies with standardized methodologies [190]. In vitro evidence suggests that ApoE4 is associated with decreased early endosomal markers (Rab5/EEA1) and reduced uptake efficiency of molecules such as epidermal growth factor and transferrin [191], potentially indirectly affecting BBB function by influencing endocytosis and membrane protein renewal in neurons and glial cells.

In summary, ApoE influences the BBB through multiple pathways, including the activation of inflammatory enzyme pathways that lead to vascular unit structural damage, the disruption of astrocyte-endothelial communication, and the induction of cerebral lipid metabolism disorders. To systematically review the mechanisms of ApoE subtype-mediated BBB damage and their experimental basis, this paper summarizes recent relevant studies (see Table 2). However, it is worth noting that not all studies reach consistent conclusions. For instance, studies have indicated that in various dementia and aging populations, increased BBB permeability shows no significant association with amyloid pathology or ApoE genotype. Instead, it correlates with vascular metabolic factors, such as diabetes, and vascular function indicators, including Vascular Endothelial Growth Factor (VEGF) and adhesion molecules [203, 204]. This suggests that findings linking ApoE4 to BBB impairment may vary across studies due to methodological and population differences, necessitating further mechanistic research to clarify and validate these findings.

Table 2.

Summary of studies on ApoE isoform-mediated mechanisms of blood–brain barrier dysfunction

Year	study model (Type + System)	Mechanistic pathway/target tested	Experimental manipulation/intervention	Key findings	Ref.
2011	[In vitro, mouse] Triple co-culture BBB (WT mouse brain microvascular endothelial cells + pericytes + astrocytes from APOE3-KI or APOE4-KI mice); TEER readout. [In vivo, mouse] Evans blue leakage (ApoE4-KI > ApoE3-KI).	Target Tested: LRP1 activates PKCδ, leading to threonine phosphorylation of occludin; tyrosine phosphorylation of occludin was not detected; tight junction protein levels (occludin, claudin-3, claudin-5) were unchanged.	RAP reduced PKCδ; anti-LRP1 reduced PKCδ and threonine phosphorylation of occludin, and decreased TEER. (No PKC inhibitor was used.)	ApoE4-BBB showed lower TEER than ApoE3-BBB; LRP1 blockade in the ApoE3 setting recapitulated decreases in PKCδ/occludin-Thr and TEER; in vivo, Evans blue leakage was higher in ApoE4-KI mice.	[184]
2012	Human clinical observational cohort, mild–moderate AD; CSF + plasma; BBB by CSF albumin index (≥9 impaired).	Association: higher plasma TG (per +10 mg/dL → albumin index +0.13); HDL lower at group level but not significant in multivariable models; TC/LDL not significant; no APOE4–lipid interaction was tested.	None (observational)	BBB-impaired group: TG higher, HDL lower; the TG association remained significant after adjustment (age, sex, APOE4, blood pressure, statin use), while HDL was not significant after adjustment.	[192]
2012	in vivo, mouse: TR-APOEε2/ε3/ε4 (expressing human ApoE2/3/4), Apoe−/−, GFAP-APOE; BBB leakage (dextran/IgG, cadaverine), TJ/ECM, microvasculature & CBF.	Pericyte CypA→NF-κB→MMP-9; ApoE3 protein (from APOEε3) restrains the pathway via LRP1, whereas ApoE4 protein (from APOEε4) fails.	CypA deletion/cyclosporine A; NF-κB blockade (PDTC, siRelA); MMP-9 blockade (SB-3CT, siMMP-9); siLrp1 (hippocampus).	APOEε4 (ApoE4) and Apoe−/− → leaky BBB with microvascular/CBF deficits; APOEε2/ε2 and ε3/ε3 → intact BBB. Blocking the CypA–NF-κB–MMP-9 axis restores barrier (and vascular/CBF measures); LRP1 knockdown in APOEε3 mice mimics an APOEε4–like vascular phenotype.	[175]
2015	In vitro, human brain vascular pericytes (APOEε3/ε3).	Isoform-dependent suppression of pericyte migration by ApoE.	Exogenous ApoE3 vs ApoE4 particles (± APOE siRNA background; e.g., 5 ng/mL).	ApoE3, not ApoE4, reduces migration and rescues the hyper-migratory phenotype after APOE knockdown; proliferation unchanged.	[193]
2015	Same as above.	ApoE–LRP1 signaling restrains pericyte motility.	APOE siRNA; LRP1 siRNA or RAP (LRP1 antagonist).	APOE knockdown ↑ migration; LRP1 knockdown/RAP abolishes ApoE’s restraint → LRP1-dependent effect. (No BBB readout.)
2015	Same pericyte culture.	RhoA–actin cytoskeleton regulation of migration/adhesion.	apoE siRNA; RhoA inhibitor.	APOE knockdown ↑ active RhoA (~2–3×) and ↑ migration/adhesion with actin spreading; RhoA inhibition normalizes migration. (Integrin-α2 ↓, fibronectin ↑
2015	Human, in vivo—NCI vs MCI; DCE-MRI (hippocampal Ktrans) + CSF biomarkers (incl. sPDGFRβ).	CSF sPDGFRβ ↑ (pericyte-injury marker) correlates with hippocampal Ktrans; CSF/plasma albumin ratio ↑.	None (observational).	Age-related hippocampal BBB breakdown (NCI), worse in MCI—notably CA1 & DG; sPDGFRβ ↑ and positively associated with Ktrans →candidate BBB biomarker	[149]
2016	Human, ex vivo—postmortem frontal cortex (BA9/10) from AD (APOE3 vs APOE4) and age-matched non-AD controls; quantitative confocal IF (PDGFRβ+ pericytes, lectin+ endothelium, fibrin/IgG, LRP1, CypA, MMP-9).	LRP1–CypA–MMP-9 axis in pericytes/endothelium; LRP1↓ in AD (no APOE genotype difference); CypA/MMP-9↑ (APOE4 > APOE3); pericyte coverage & capillary length quantified.	None (observational, postmortem analysis).	AD APOE4 shows greater pericyte loss, reduced microvascular length, extravascular fibrin/IgG↑ and CypA/MMP-9↑ vs AD APOE3; LRP1↓ in AD across genotypes; BBB leakage correlates with pericyte loss.	[194]
2017	Clinical, cohort analysis—QAlb (CSF/serum albumin ratio) as an indirect BBB-permeability marker.	QAlb↑; correlates with CSF markers of microvascular endothelial dysfunction/angiogenesis (VEGF, ICAM-1, VCAM-1); no association with APOE genotype or amyloid pathology.	None (observational).	QAlb is higher in dementia (all-cause) and in type 2 diabetes; associations point to BBB dysfunction but are independent of APOE genotype and amyloid pathology.	[195]
2018	Animal, in vivo—WT C57BL/6 mice; controlled cortical impact (TBI); cortical microvessels isolated; BBB assessed by Evans blue and MRI.	CypA–MMP-9 axis; pericytes (PDGFRβ), MMP-9, tight-junction proteins (ZO-1, occludin, claudin-5).	Cyclosporin A (CypA antagonist; 20 mg/kg, i.p.) in WT mice only.	TBI→pericyte loss, MMP-9↑and TJ↓; CsA attenuates MMP-9 and accelerates BBB repair.	[196]
2018	Animal, in vivo—humanized APOEε3 vs ε4 knock-in mice (expressing ApoE3 vs ApoE4 proteins) with TBI; cortical microvessels analyzed.	Effect of ApoE protein isoform on BBB repair via CypA–MMP-9/TJ recovery; PDGFRβ dynamics (pericyte repopulation).	None for genotype comparison (no drug in APOE groups).	APOEε4 mice show delayed BBB closure, sustained microvascular MMP-9↑, impaired recovery of ZO-1/occludin/claudin-5, and prolonged PDGFRβ↓ vs APOEε3.
2021	Preclinical (in vivo + in vitro)—cortical Aβ co-infused with native APOE3 vs APOE4 lipoproteins in mice; two-photon imaging, flow cytometry, bulk & single-cell RNA-seq; primary-microglia Aβ-uptake assays.	APOE-isoform–specific lipoprotein phospholipids shape microglial activation and Aβ uptake; APOE–TREM2 interaction (Trem2 deficiency modulates E4 effects).	Cortical infusion of Aβ + native APOE3 or APOE4 lipoproteins; WT vs Trem2-knockout comparisons; no pharmacologic inhibitor.	APOE3 lipoproteins drive faster microglial migration and higher Aβ uptake (≤24 h), a stronger DAM-like transcriptional response, and better protection in cognition than APOE4; Trem2 deficiency reduces Aβ uptake only with E4.	[197]
2021	Animal, in vivo—EC-specific Lrp1^f/f; Tie2-Cre; BBB by IgG/fibrin, tracer leakage & MRI Ktrans.	EC-LRP1↓→CypA–NF-κB–MMP-9↑; TJ/collagen IV↓; LRP1 rescue normalizes pathway.	CsA/Debio-025 (CypA blockade); AAV2/BR1–LRP1 endothelium-targeted rescue.	EC-Lrp1 KO→ progressive BBB leak and neurodegeneration; CypA blockade or LRP1 rescue reverses BBB defects.	[177]
2021	In vitro—human astrocytes & SH-SY5Y monocultures; recombinant ApoE2/ApoE3/ApoE4 proteins (20 nM, 24 h).	ApoE protein isoforms modulate cytokines (IL-1β/IL-6/TNFα, E4 > E3≈E2) and BDNF forms (ApoE2→mature BDNF↑; ApoE3/E4→proBDNF↑).	Exogenous ApoE protein isoforms (no inhibitor).	ApoE2 protein increases mature BDNF and neuronal viability; ApoE4 protein elevates cytokines and reduces neuronal viability.	[198]
2021	In vitro—human HMC3 microglia; recombinant ApoE2/ApoE3/ApoE4 proteins (20 nM, 24 h).	ApoE protein isoforms tune microglial state (TNFα↑ with E4; IL-6↓ with E4; TREM2/Clec7a↑ with E4).	Exogenous ApoE protein isoforms (no inhibitor).	ApoE4 protein drives a more pro-inflammatory/DAM-like phenotype; ApoE2/3 proteins are comparatively less activating.
2021	Animal, in vivo—MAPT/AEP mice; AAV-mediated expression of ApoE3 vs ApoE4 proteins.	Effect of ApoE protein isoform on AEP-mediated tau cleavage, synaptotoxicity, and cognitive readouts.	AAV-ApoE3 or AAV-ApoE4 (viral expression of ApoE protein isoforms).	ApoE3 protein preserves tau integrity and neuronal structure, reduces synaptotoxicity, and supports cognitive measures vs ApoE4 protein.
2021	Preclinical (in vitro + in vivo)—GLUT1-targeted (mannose) + CPP-modified (RVG/penetratin) liposomes loading pApoE2; in-vitro BBB (bEnd.3–astrocyte coculture) and C57BL/6 mice (i.v.).	GLUT1 targeting + CPP translocation → BBB crossing and brain expression of the ApoE2 protein isoform (neurons/astrocytes/endothelium).	RVGMAN/PenMAN dual-modified liposomes + pApoE2–chitosan; single i.v. dose; transfection across BBB model.	Dual-modified liposomes increase ApoE2 expression (≈2× vs control in vitro; ~36–38 ng ApoE2/mg brain protein in vivo) with no overt toxicity; single-ligand/unmodified carriers are inferior.	[199]
2021	In vitro—neuronal/glial cultures exposed to ApoE protein isoforms ±27-hydroxycholesterol (27-OHC).	ApoE4 + 27-OHC → C/EBPβ–δ-secretase (AEP) pathway activation, leading to APP/tau cleavage and neurotoxic signaling.	None (comparative exposure)—ApoE2/3/4 proteins ±27-OHC; no pharmacologic inhibitor.	ApoE4 + 27-OHC synergistically enhances C/EBPβ–AEP signaling, increases neurotoxic processing of APP/tau and worsens neuronal viability	[200]
2021	Clinical, observational—DCE-MRI Ktrans in cognitively unimpaired APOEε4 carriers (ε3/ε4, ε4/ε4) vs APOEε3/ε3; CSF biomarkers ± Aβ/tau PET.	Observational—BBB breakdown localized to hippocampus/parahippocampal gyrus, independent of Aβ/tau; CSF sPDGFRβ↑ correlates with Ktrans and predicts cognitive decline; CSF CypA–MMP-9↑ in ε4.	None (imaging + biofluids).	Unimpaired APOEε4 carriers already show medial-temporal BBB leakage; severity increases with impairment; baseline sPDGFRβ predicts future decline; CSF CypA/MMP-9 associates with pericyte injury.	[170]
2024	In vitro—human iPSC-derived BMEC-like + pericyte-like cells from isogenic APOE2/3/4/KO lines; Transwell Aβ40/42 transcytosis.	LRP1 (abluminal) and P-gp (luminal) mediate brain-to-blood Aβ efflux; ApoE protein isoforms and APOE genotypes differentially modulate clearance vs deposition.	Aβ40/42±recombinant ApoE2/ApoE3/ApoE4; RAP (LRP1 blocker) and cyclosporin A (P-gp inhibitor) used as pathway probes.	Barrier tight across genotypes (TEER > 3,000 Ω × cm2); APOE2 BMEC-like cells show Aβ42 transport↑. In APOE KO, rhApoE3→Aβ40↑, rhApoE2→Aβ42↑, rhApoE4→minimal. APOE4 pericyte-like cells show extracellular Aβ42↑/uptake↓, indicating weaker clearance with E4.	[187]
2024	Clinical, observational—DCE-MRI to estimate Ktrans plus multi-shell diffusion MRI (RSI); participants from cognitively unimpaired to mildly impaired; groups: APOEε4 carriers vs APOEε3/ε3 non-carriers.	Observational—relationship between BBB permeability (Ktrans) and microstructure (RSI); modulation by Aβ status and APOEε4.	None (observational imaging ± biofluids).	APOEε4 carriers show cortical Ktrans↑—strongest in CU Aβ−; entorhinal Ktrans tracks injury (free water↑, neurite density↓) ⇒ early BBB dysfunction.	[173]
2025	Preclinical, multi-system—human postmortem brains; human iPSC-derived 3D astrocyte & cerebrovascular models; APOEε4 KI and APP-KI mice; zebrafish—focusing on the astrocyte–endothelial BBB interface.	ApoE4 protein/APOEε4 → FN1 (fibronectin)↑ in astrocytes & endothelium; integrin–FAK signaling suppresses VEGFA (astrocyte) → HBEGF (endothelium) → IGF1 (astrocyte) cascade, disrupting gliovascular communication and tight junctions.	Pathway probes—Aβ42 or TNFα challenge; VEGFR inhibitors; FAK inhibitor PF-573228; HBEGF/IGF1 add-back; Torin2 (mTOR inhibitor); camptothecin (Topo-I inhibitor); zebrafish fn1b−/− rescue.	ApoE4/APOEε4 drives FN1 deposition at the BBB, co-aggregating with ApoE4 and impairing astrocyte–endothelial signaling and barrier integrity; FAK inhibition or HBEGF/IGF1 add-back restores the VEGF axis & tight junctions → FN1 is a tractable target.	[201]
2025	Clinical, observational—non-diabetic, biomarker-confirmed AD cohort; BBB markers: QAlb and free light chains (FLCs); insulin resistance by TyG index (tertiles); APOE genotypes (ε4 dose 0/1/2).	Observational—association of insulin resistance (TyG) and its interaction with APOEε4 with BBB permeability (QAlb, λ-FLC).	None (biofluids plus multivariable statistical modeling).	Higher TyG → BBB leak↑ (QAlb↑, λ-FLC↑); permeability is highest in APOEε4/ε4 with high TyG → ε4 adds/modifies IR-related BBB dysfunction.	[202]

Multilayered mechanisms of apolipoprotein A-I in blood-brain barrier regulation

Following the previous discussion on how different ApoE subtypes affect BBB function, we now turn to apolipoprotein ApoA-I in HDL to explore its complex mechanisms in regulating the BBB. ApoA-I is primarily expressed in the liver and small intestine, with extremely low or undetectable expression in normal brain parenchyma. Therefore, ApoA-I in the brain is most likely mainly derived from the peripheral system, and it may enter the brain through receptor-mediated endocytosis across the BBB. At the same time, the Blood-CSF Barrier (BCSFB) may also be involved in the process of peripheral ApoA-I entering the CSF [205]. In vitro studies using the human Cerebral Microvascular Endothelial Cell (hCMEC/D3) model suggest that the transmembrane uptake of ApoA-I is saturable and exhibits cholesterol-dependent, rather than clathrin-dependent, endocytic characteristics. Based on previous studies, this endocytosis is likely primarily mediated by the scavenger receptor SR-BI, and a small amount of transcytosis may occur from the luminal to the basal side of the endothelial monolayer [48]. In vivo radiotracer studies demonstrate extremely low brain penetration of intravenously injected ApoA-I in rodents, comparable in magnitude to plasma proteins like albumin [48]; In contrast, ApoA-I mimetic peptides (e.g., 4F) exhibit significantly higher brain permeability [206], and can be studied as functional substitutes, though they do not represent the barrier-crossing capacity of intact ApoA-I.

Effects on BBB endothelial integrity and permeability: In mouse Middle Cerebral Artery Occlusion (MCAO) models and in vitro Oxygen-glucose Deprivation (OGD) models, exogenous HDL/ApoA-I reduced plasma IgG and albumin extravasation while maintaining tight junction proteins and Trans-epithelial Electrical Resistance (TEER), suggesting protective effects on the barrier [207]. At the population level, in neuroimmunological disorders such as multiple sclerosis, higher plasma concentrations of HDL-C and ApoA-I correlate with lower CSF total protein, albumin ratio, and immunoglobulin levels, suggesting reduced BBB permeability. This is accompanied by a significant decrease in the number of activated immune cells (e.g., CD80+CD19+ B cells) entering the central nervous system [208]. However, these findings are correlational observations and cannot establish causality, requiring further experimental validation.

Anti-inflammatory and Antioxidant Effects of ApoA-I: ApoA-I exhibits significant anti-inflammatory regulatory effects [209], reducing inflammatory activation in BBB endothelial cells. In vitro experiments, pre-treatment of human microvascular endothelial cells with ApoA-I or HDL significantly inhibits palmitic acid-induced NF-κB activation, decreases the expression of the cell adhesion molecule ICAM-1, and reduces the release of the pro-inflammatory cytokine IL-6 [210]. This mechanism may relate to ApoA-I-mediated membrane cholesterol efflux: cholesterol depletion disrupts the lipid raft structure of endothelial cell membranes, preventing the aggregation of Toll-like Receptor 4 (TLR4) at lipid raft microdomains—a critical step in the initiation of inflammatory signaling—and thereby attenuating NF-κB pathway activation at its source [211]. It should be noted that most evidence for this effect originates from non-brain-derived human microvascular endothelial cells, and its equivalence in brain endothelium requires further validation. ApoA-I exhibits antioxidant effects [212], It binds to Amyloid-β (Aβ) peptides and inhibits their aggregation, thereby mitigating Aβ-induced neurotoxicity and oxidative damage [213, 214]. ApoA-I maintains BBB function at the molecular level, including antioxidant defense and vascular regulation by carrying the PON1 enzyme and activating the endothelial nitric oxide synthase signaling pathway. PON1 is an antioxidant enzyme associated with HDL [215] with no endogenous expression in the brain, may facilitate the passage of ApoA-I particles across the BBB and into brain tissue [216, 217]. In Alzheimer’s disease mouse models, studies have detected PON1 and PON3 proteins aggregated within glial cells surrounding cerebral amyloid plaques, which locally degrade lipid peroxides and mitigate oxidative stress damage [218]. However, direct evidence for their entry into the brain via ApoA-I across the BBB remains lacking, making this a point of speculation and debate. On the other hand, the interaction between ApoA-I/HDL and endothelial SR-BI rapidly activates Endothelial Nitric Oxide Synthase (eNOS), enhancing Nitric Oxide (NO) production [219]. The HDL-SR-BI binding process occurs within microvesicular structures on the endothelial cell membrane [220]. This promotes the dissociation of caveolin-1, which inhibits eNOS, from the eNOS complex, thereby triggering eNOS activation and the release of NO [221]. NO not only contributes to cerebral vasodilation and microcirculatory perfusion [222], but also improves endothelial function and neurovascular unit coupling [223]. However, much evidence for this pathway originates from the systemic/coronary endothelium. Considering the scarcity of caveolae in brain endothelium and the MFSD2A-mediated transcellular inhibition, its conservation and effect magnitude in brain endothelium require dedicated validation (See Table 3).

Table 3.

Mechanistic insights from experimental and clinical studies on ApoA-I/HDL and the blood–brain barrier

Year	Study Model (Type + System)	Mechanistic Pathway/Target Tested	Experimental Manipulation/Intervention	Key Findings	Ref.
2015	Clinical observational cohort study in 154 patients with clinically isolated syndrome (CIS) suggestive of MS	Associations of serum HDL-C and ApoA-I levels with CSF-derived markers of BBB integrity and immune cell infiltration	Measurement of serum lipids and CSF biomarkers before treatment (interferon-β1a); correlation analyses with MRI and immunological CSF variables	Higher HDL-C and ApoA-I levels were associated with lower CSF albumin, total protein, IgG, and albumin quotient; also correlated with reduced CD80+ and CD80+CD19+ cell frequencies in CSF, suggesting HDL may protect against BBB injury and central nervous systems immune infiltration	[208]
2019	In vivo: Wild-type rats (125I-ApoA-I injection); In vitro: hCMEC/D3 human brain endothelial monolayers (AF647-ApoA-I)	ApoA-I crosses the BBB via clathrin-independent and cholesterol-mediated endocytosis and is transcytosed across endothelium	IV injection of 125I-ApoA-I and quantification of PS values in six brain regions; In vitro use of AF647-ApoA-I + inhibitors (MbCD, nystatin) and siRNA knockdown of clathrin; confocal microscopy and Western blot to track transcytosis	ApoA-I detected on abluminal side in vitro; Cholesterol depletion significantly reduced uptake; Clathrin knockdown did not affect ApoA-I uptake; PS values indicate brain uptake via BBB in addition to BCSFB	[48]
2020	Clinical cohort (stroke patients with favorable vs. unfavorable outcome at 3 months) and in vitro BBB model (hCMEC/D3)	HDL particle size and protein composition (PON1/AAT) affect anti-inflammatory and antioxidant capacity on BBB endothelial cells	HDL isolated from stroke patients or healthy donors; (i) anti-inflammatory effect on TNFα-stimulated hCMEC/D3 (qPCR: SELE, MCP1, IL8, CLDN1); (ii) antioxidant function (inhibition of Cu2±induced LDL oxidation); (iii) Western blot: PON1 and AAT in HDL	Stroke patients with unfavorable outcomes had fewer small HDL particles, lower HDL-PON1 content, and higher HDL-AAT; these HDLs were less effective at inhibiting TNFα-induced SELE/IL8/MCP1 expression and LDL oxidation, indicating impaired BBB-protective functions	[216]
2021	In vivo: Mouse model of acute ischemic stroke (transient MCAO with reperfusion); In vitro: hCMEC/D3 brain endothelial cells under OGD/reperfusion conditions	HDL exerts BBB-protective and neuroprotective effects via endothelial SR-BI (Scavenger Receptor Class B Type I)	HDL infusion (80 mg/kg apoA-I) at reperfusion in MCAO mice; Comparison between endothelial SR-BI-deficient (Tie2-Cre) and myeloid SR-BI-deficient (LysM-Cre) mice; OGD/reoxygenation in hCMEC/D3 cells with or without HDL + SR-BI inhibitor BLT-1	HDL reduced infarct size and BBB leakage in control but not endothelial SR-BI-deficient mice; In vitro, HDL preserved barrier integrity, which was impaired by SR-BI inhibition; Protective effects were independent of myeloid SR-BI	[207]
2021	In vivo: Tg2576 transgenic AD mice vs. wild-type (WT) controls; Tissue-based analysis using immunohistochemistry	PON1 and PON3 delivered by HDL across BBB accumulate near Aβ plaques and may reduce oxidative stress in glial cells	Histological and immunofluorescence analysis in five brain regions; Double and triple staining for PON1/3 and glial markers (GFAP, Iba1, Nestin, NeuN); comparison with WT	PON1/PON3 highly expressed around Aβ plaques in star-shaped glial-like cells (astrocytes, microglia, oligodendrocytes); PON3 partially colocalized with microglia; supports HDL-mediated delivery of antioxidant enzymes into central nervous systems; no local PON1/3 gene expression detected	[218]
2022	Clinical observational study (71 anti-NMDAR encephalitis patients vs. 71 controls)	Association of serum ApoB/ApoA-I ratio with systemic inflammation and disease severity	Retrospective analysis of ApoB/ApoA-I ratio, CRP, ESR, ICU stay, mRS scores; subgroup analysis based on high vs. low ApoB/ApoA-I	Elevated ApoB/ApoA-I ratio was significantly associated with ICU admission, longer hospital stay, increased CRP/ESR, and mRS ≥3; ApoA-I levels were reduced in patients; findings support a pro-inflammatory role of altered lipoprotein balance	[224]
2025	In silico: Multiscale molecular dynamics (MD & T-REMD) simulations of apoA-I ‘LN’ peptide and Aβ42CC in aqueous phase and HDL3C particles	In silico: Multiscale molecular dynamics simulations of apoA-I ‘LN’ peptide and Aβ42CC in aqueous phase and HDL3C particles	Conventional MD and T-REMD simulations of Aβ42CC with the ‘LN’ region of apoA-I via hydrophobic contacts and H-bonds; HDL3C particles stabilize Aβ at the protein-surface, potentially preventing aggregation; Asp48 of apoA-I may anchor the complex to HDL surface lipid headgroups	Aβ42CC forms stable, energetically favorable complexes with the ‘LN’ region of apoA-I via hydrophobic contacts and H-bonds; HDL3C particles stabilize Aβ at the protein-surface, potentially preventing aggregation; Asp48 of apoA-I may anchor the complex to HDL surface lipid headgroups	[214]

The role of the ApoM/S1P signaling axis in blood-brain barrier homeostasis regulation

After elucidating the multilayered mechanisms of ApoA-I, we will now focus on the regulatory role of the ApoM/S1P signaling axis carried by HDL-like particles in maintaining BBB homeostasis. HDL apolipoprotein M (ApoM) is a 26 kDa protein discovered in 1999 by Xu and Dahlbäck et al. Belonging to the lipid transport protein family, ApoM possesses hydrophobic pockets capable of binding small-molecule lipids [225]. Initial reports indicated that ApoM is primarily synthesized by the liver and proximal tubular epithelial cells in the kidneys [226]. Subsequently, ApoM expression was also detected in brain microvascular endothelial cells and colon tissue [227, 228]. Epidemiological studies have revealed that plasma ApoM levels negatively correlate with waist circumference, the severity of insulin resistance, and CRP [229], suggesting that ApoM may play a role in regulating metabolic and inflammatory processes. Most plasma ApoM binds to HDL [230]. One of ApoM’s most important physiological functions is binding and transporting S1P [231]. Approximately 60–70% of plasma S1P binds to ApoM carried by HDL, with the remainder primarily binding to carriers such as albumin [232], ApoM is regarded as the primary partner protein for S1P.

The ApoM-S1P axis is a crucial basis for HDL-like particles-mediated protection of the vascular endothelium. For example, ApoM deficiency exacerbates increased endothelial permeability during inflammation, while S1P supplementation partially reverses this effect [233]. S1P is an amphipathic bioactive lipid generated by sphingosine phosphorylation [234]. Compared to peripheral tissues, S1P concentrations are higher within the central nervous system, with particularly abundant levels in the spinal cord and brainstem, followed by the cerebellum; cortical concentrations are relatively low [235]. S1P molecules are composed of a hydrophilic phosphate head group and an 18-carbon unsaturated fatty alcohol backbone. The phosphate group enables it to activate specific receptors, while the hydrophobic long chain allows it to embed into cell membranes or bind to lipid particles such as HDL-like particles [236]. S1P is primarily synthesized and released into the circulation by erythrocytes, platelets, vascular endothelium, and immune cells [237], with erythrocytes play the most important role in maintaining the plasma S1P homeostatic concentration [238]. Endothelial and immune cells dynamically regulate S1P signaling in local microenvironments. The body has five G protein-coupled S1P receptors (S1PR1–5) [239], among which S1PR1 is highly expressed in brain endothelial cells and is a key receptor in maintaining the integrity of the vascular barrier [240]. S1PR1 knockout mice exhibit embryonic lethality due to vascular leakage and developmental defects, underscoring the importance of this receptor for vascular stability [239] S1P signaling drugs are used in the treatment of neurological disorders, such as S1P receptor modulators (e.g., fingolimod, seholimod, ozanimod, ponimod), which are applied in multiple sclerosis [241]. These drugs interfere with S1P-mediated lymphocyte extramigration by activating and internalizing S1PR1 on lymphocytes, thereby reducing the number of autoimmune lymphocytes entering the central nervous system and controlling inflammatory damage [242]. Furthermore, S1P signaling plays a crucial role in various central nervous system disorders, including stroke, neuroinflammation, neurodegeneration, and brain tumors, as extensively demonstrated by numerous studies [243, 244]. For instance, in ischemic stroke models and neurodegenerative diseases like Alzheimer’s disease, abnormal S1P signaling is closely associated with BBB dysfunction and neuronal injury [245, 246]. Conversely, within the microenvironment of brain tumors, such as gliomas, S1P has been demonstrated to participate in tumor-associated vascular permeability and immune regulation [247]. In studies of spontaneously hypertensive stroke rats, chronic cerebral hypoperfusion for 4 weeks resulted in significantly downregulated S1PR1 expression in the brain. This was accompanied by reduced expression of brain microvascular tight junction proteins (claudin-5, ZO-1, and occludin), increased BBB permeability, and the accumulation of phosphorylated tau protein in neurons [248].

Given that ApoM is the primary carrier of plasma S1P, the ApoM/S1P axis is considered to play a crucial role in regulating BBB function. S1P transported by ApoM can activate the S1PR1 receptor on endothelial cell surfaces more efficiently, thereby promoting barrier stability [249]. In ApoM knockout mice, BBB permeability to small-molecule tracers significantly increases, and brain endothelial cells exhibit markedly elevated transport of macromolecules like albumin via pinocytosis. Application of S1PR1-selective agonists partially corrects BBB leakage and excessive albumin transendothelial transport in ApoM-deficient mice [250], ApoM-delivered S1P primarily maintains the BBBs low permeability by activating endothelial S1PR1. Mechanistic studies further indicate that ApoM-bound S1P preferentially induces S1PR1 signaling pathways, which are beneficial for maintaining the integrity of the barrier function. In the absence of ApoM, free or carrier-bound S1P (e.g., albumin-bound) may more readily act on S1PR2/S1PR3 receptors [251]. However, direct causal evidence currently stems primarily from rodent models and in vitro experiments with cerebral microvascular endothelial cells. S1PR1 promotes actin ring formation and tight junction stabilization in the cortex via the Gi/PI3K/Tiam1-Rac1 pathway, thereby maintaining endothelial barrier integrity [252]. Conversely, S1PR2 and S1PR3 promote stress fiber formation and endothelial contraction in endothelial cells through the G12/13-RhoA/ROCK pathway [253], hereby increasing BBB permeability. Pathological S1PR2 upregulation has been observed to affect the BBB in traumatic brain injury models adversely: excessively activated S1PR2 in cerebral microvascular endothelium induces matrix metalloproteinase-9 expression via the JNK/c-Jun pathway, accompanied by tight junction protein degradation, increased BBB permeability, and exacerbated neuronal injury [254]. However, this mechanism primarily originates from rodent models, and its applicability to humans requires further validation.

ApoM-bound S1P maintains BBB integrity and stability by preferentially activating endothelial S1PR1/Rac1 signaling, whereas excessive S1PR2-mediated signaling impairs barrier function. Regulating the ApoM/S1P axis is crucial for BBB protection. On one hand, enhancing the ApoM and S1PR1-mediated signaling pathway is expected to strengthen BBB protection; on the other hand, specifically inhibiting harmful S1PR2-mediated pathways is also considered to have therapeutic potential [255]. However, the direct impact and clinical significance of the ApoM/S1P axis on the human BBB still need to be validated through more prospective and interventional studies.

Apolipoprotein J (Clusterin) and blood-brain barrier function

Beyond ApoA-I and the ApoM/S1P axis, another key HDL-like particles component, ApoJ (also known as Clusterin), plays a crucial role in maintaining the integrity of the BBB. Clusterin is a secreted glycoprotein first identified in ram epididymal fluid during the 1980s, named for its significant upregulation during cell clustering processes [256]. This protein possesses multiple aliases, including ApoJ, TRPM-2, and SP-40,40. It is widely distributed in plasma and CSF, participating in lipid metabolism, regulating inflammation, and controlling protein aggregation [257]. ApoJ gene polymorphisms are closely associated with Alzheimer’s disease; genome-wide association studies have identified ApoJ as one of the most significant risk genes for AD, ranking second only to ApoE [258]. ApoJ binds to Aβ, inhibits its aggregation, and promotes its clearance across the BBB [259]. These properties position ApoJ as a crucial protective factor within the nervous system, with its expression often significantly altered during aging and neurodegenerative processes.

ApoJ provides a pathway for Aβ to cross the BBB, enabling its transport to the bloodstream or CSF for further metabolic clearance. When ApoJ is absent, this clearance pathway is impaired. Excess peptides retained in the brain tend to deposit on vascular walls, forming amyloid deposits. This mechanism has been validated in in vivo models: transgenic mice lacking ApoJ exhibit significantly increased cerebral vascular Aβ deposition and amyloid angiopathy [260]. Thus, ApoJ’s role in promoting Aβ clearance across the BBB and protecting cerebral blood vessels from Aβ toxicity is not to be underestimated.

ApoJ also maintains the immune homeostasis of the BBB through multiple pathways. On the one hand, ApoJ can bind to the terminal products of the complement cascade, preventing the membrane attack complex from attacking cells and thereby mitigating inflammation-mediated damage to the vascular endothelium and neurons [261, 262]. On the other hand, ApoJ can regulate cellular inflammatory responses within the brain. For example, in the APP23 transgenic AD model, exogenous administration of recombinant ApoJ enhances microglial phagocytic clearance of Aβ while significantly reducing pro-inflammatory cytokines (e.g., IL-17) [263]. This indicates ApoJ helps limit inflammation-mediated BBB damage and promotes Aβ clearance. Furthermore, ApoJ influences BBB-related signaling pathways. Aβ oligomers can induce ApoJ accumulation within brain microvascular endothelial cells while reducing its secretion, thereby triggering the release of Dickkopf-1 (DKK1) and inhibiting the Wnt/β-catenin signaling pathway. This signaling suppression downregulates P-gp expression on the endothelial cell surface, weakening the efflux capacity of Aβ across the BBB and forming a vicious cycle [263]. Consequently, in pathological conditions like AD, alterations in ApoJ represent both a response to Aβ toxicity and a potential contributor to further Aβ accumulation and BBB dysfunction.

The effects of ApoJ on BBB function are closely connected to its regulation of amyloid precursor protein metabolism and cholesterol homeostasis. Within BBB endothelial cells, upregulation of ApoJ encourages non-amyloid pathway processing of amyloid precursor protein and decreases intracellular Aβ accumulation [264]. However, some models indicate that ApoJ deficiency paradoxically reduces cerebral plaques [265], suggesting that its effects may be influenced by brain region, Aβ levels, and ApoE interactions.

It must be noted, however, that the role of ApoJ in regulating BBB function remains incompletely understood. Research indicates that ApoJ may influence Aβ clearance across the BBB by affecting cholesterol homeostasis and amyloid precursor protein metabolism. Nevertheless, its specific mechanism of action within BBB endothelial cells requires further investigation.

Interactions between HDL-like particles and pericytes in regulating blood-brain barrier function

The integrity of the BBB relies on the participation of supporting cells such as pericytes. HDL-like particles also play a crucial role in maintaining BBB homeostasis by regulating the function of pericytes. Pericytes are cells embedded in the capillary basement membrane, tightly enveloping the outer surface of endothelial cells. Through interactions with endothelial cells, astrocytes, and neurons, they collectively form neurovascular units that play a crucial role in maintaining the integrity and function of the BBB. In the peripheral vascular system, HDL and its major components are known to exert multiple protective effects on blood vessels. Increasing evidence suggests that HDL-like particles and their constituents also play a role in maintaining the structure and function of the BBB by regulating pericytes.

Structurally and functionally, pericyte cells form tight structural connections with endothelial cells by secreting and remodeling basement membrane components (e.g., type IV collagen) and adhesion molecules (e.g., N-cadherin, VE-cadherin), jointly maintaining the structural integrity of the BBB [266, 267]. Yamazaki et al.’s study found that in an in vitro co-culture model, pericytes expressing ApoE4, compared to ApoE3, significantly weakened the ability to induce endothelial cell basement membrane proteins (such as collagen IV) and extracellular matrix proteins, leading to impaired endothelial lumen formation and reduced barrier function. This phenomenon was further validated in ApoEε4 transgenic mice. With aging, these mice exhibited markedly reduced collagen IV expression in cerebral microvessels and significantly increased BBB permeability to plasma proteins [69]. Compared to ApoEε3 carriers and non-Alzheimer’s disease controls, the number and vascular coverage of PDGFRβ⁺ pericytes in the cortical microvasculature of ApoEε4 carriers are significantly reduced, suggesting accelerated pericytic degeneration in ApoE4-associated Alzheimer’s disease [194]. Current perspectives suggest this may be linked to ApoE4-mediated downregulation of LRP1. Defective LRP1 signaling weakens the protective function of this synergistic network [53], subsequently triggering pericyte detachment and basement membrane structural disruption, thereby exacerbating BBB damage.

Additionally, S1P carried by HDL-like particles plays a crucial role in maintaining vascular structure. It has been suggested that pericytes themselves may have the ability to synthesize and secrete S1P, but this mechanism has not been definitively confirmed [268]. By activating S1P1 receptors on adjacent endothelial cells, HDL promotes the membrane localization and expression of N-cadherin and VE-cadherin. This enhances pericytes’ ability to envelop capillaries and their adhesion to endothelium, thereby reinforcing the stable structure of the BBB [268, 269]. Thus, at the structural level, key functional components of HDL-like particles—including ApoE and S1P—may directly influence BBB stability by regulating the expression of basement membrane and adhesion molecules by pericytes.

In terms of signaling pathway regulation, HDL-like particles and their primary functional components modulate the interactions between pericytes and endothelial cells through multiple signaling pathways, thereby maintaining the functional stability of the BBB. S1P activates various intracellular signaling pathways by binding to S1P1 receptors on the surface of endothelial cells or pericytes, thereby playing a crucial role in maintaining vascular stability. In retinal microvascular models, the S1P/S1PR1 signaling axis upregulates adhesion molecules (e.g., N-cadherin) expressed by pericytes and endothelial cells, thereby enhancing structural interactions between them and preserving vascular barrier function [270]. Similar findings were observed in brain-derived experiments: S1P secreted by vascular endothelial cells promotes N-cadherin transport to the endothelial cell membrane surface by activating S1P1 receptors, thereby enhancing adhesion with pericytes and maintaining vascular structural stability [271]. Furthermore, HDL-like particles activate protective signaling pathways in endothelial cells or pericytes by binding their receptor SR-BI, including the activation of eNOS, as well as anti-apoptotic, antioxidant, and anti-inflammatory responses, while inhibiting pro-inflammatory pathways such as NF-κB activation, thereby further enhancing the structural stability of neurovascular units [122]. Regarding cholesterol metabolism and the clearance of toxic proteins, the interaction between ApoE and the LRP1 plays a crucial role in pericyte function. Ma et al. discovered that in APP transgenic mice and AD patient brain tissues, blood-brain barrier-associated pericytes can uptake and clear aggregated Aβ proteins via an LRP1-dependent mechanism. ApoE subtypes regulate this process: exogenous ApoE3 restores clearance function in ApoE-deficient pericytes, whereas ApoE4 fails to achieve the same effect [272].

HDL-like particles and their major components influence the function of perivascular cells through multiple mechanisms. HDL-like particles, carrying S1P, ApoA-I, and ApoE, form a synergistic regulatory network between perivascular cells and endothelial cells via receptor-mediated signaling pathways, including SR-BI, S1PR1, and LRP1. This network collectively maintains the functional and structural integrity of the BBB across multiple dimensions, including adhesion, metabolism, inflammation, and clearance. Future research should further elucidate the cell-specific and temporally regulated mechanisms of HDL-like particles’ action within neurovascular units, as well as their interactive networks with pericytes in signal transduction, metabolic regulation, and protein clearance. This will provide a reliable theoretical foundation and translational pathway for developing neurovascular protective strategies targeting the HDL-like particles–pericyte pathway.

Effects of HDL-like particles and glymphatic system interactions on the BBB

Conversely, HDL-like particles alleviate BBB burden by promoting glymphatic clearance of metabolic waste from the brain. The glymphatic system constitutes a convection-dominated exchange pathway between CSF and brain interstitial fluid, facilitated by Aquaporin 4 (AQP4) on astrocytic foot processes [273]. CSF enters the brain parenchyma via periarterial spaces, mixes with interstitial fluid, and exits through perivenular spaces before ultimately draining into peripheral lymph nodes via meningeal lymphatics [274]. Utilizing arterial pulsations to drive convective exchange between CSF and cerebral interstitial fluid effectively clears neurotoxic products like Aβ [275]. When glymphatic system flow is normal, it significantly reduces the exposure of cerebral vascular endothelial cells to harmful substances, helping maintain BBB integrity. On the other hand, glymphatic system dysfunction leads to the retention of proteins such as Aβ and tau, along with other metabolic waste products within the brain, thereby exacerbating the toxic burden in the neuronal microenvironment [276]. When the glymphatic system is impaired, the reactivity of microglia and astrocytes increases by secreting pro-inflammatory cytokines, which exacerbate microvascular damage [277]. The accumulation of metabolic waste and glial cell activation triggers inflammatory pathways, including the NF-κB and NLRP3 pathways. These inflammatory factors and proteases can downregulate the expression of tight junction proteins in endothelial cells (e.g., Occludin, Claudin-5), leading to increased BBB permeability.

The glymphatic system transports choroid plexus-derived ApoE into brain parenchyma, distributing it to neurons via periarterial channels with ApoE subtype specificity (ApoE2 > E3 > E4) [278]. Interestingly, as previously discussed, normally functioning HDL-like particles can mitigate sustained BBB damage by promoting intracerebral cholesterol redistribution, reducing glial lipid burden, and shifting microglia toward an anti-inflammatory phenotype. Conversely, dysfunctional HDL-like particles (e.g., ApoE4) impair this protective mechanism, exacerbating lipid deposition and inflammation in the brain while weakening the BBB barrier function. However, current research on potential synergistic interactions among HDL-like particles, the glymphatic system, and the BBB to jointly maintain brain homeostasis remains scarce. Direct experimental evidence remains scarce, warranting deeper exploration of the molecular mechanisms by which HDL-like particles regulate the glymphatic system. This includes investigating the effects of HDL components on AQP4 polarization or glial cell metabolism and evaluating the potential of HDL-like particle-related therapies targeting the glymphatic system (such as enhancing ApoE function or boosting glymphatic flow) for protecting the BBB. Uncovering the specific pathways of HDL-like particles-glymphatic system-BBB interactions could provide novel insights and strategies for preventing and treating neurodegenerative diseases and vascular pathologies.

Therapeutic potential of HDL/HDL-like particles in neurological disorders

The role of HDL and its analogues in various neurological disorders is gaining increasing attention, including Alzheimer’s disease/cerebral amyloid angiopathy (AD/CAA), stroke, spinal muscular atrophy, and post-traumatic stress disorder. In animal studies using Tg-SwDI mice (a model of cerebral amyloid angiopathy/Alzheimer’s disease), recombinant HDL or HDL mimetic peptides were found to reduce the brain pathological burden in AD/CAA mice and partially improve microglial activation [279]. Injection of apoA-I Milano or HDL mimetic peptides also reduced ischemic infarct volume and protected the BBB [280]. Recent multi-omics analysis further revealed that the quantity and function of HDL-like particles in CSF are significantly abnormal in patients with Spinal Muscular Atrophy (SMA). Nusinersen treatment partially corrects this dysfunction, enhancing the CSF cholesterol transport capacity [281]. In chronic stress-related neuropsychiatric disorders such as Post-traumatic Stress Disorder (PTSD), HDL levels are often markedly reduced, frequently accompanied by elevated pro-inflammatory factors and metabolic dysregulation. Epidemiological and small cohort studies have demonstrated that low HDL is closely associated with PTSD symptom severity and cognitive decline [282]. Early studies found that small HDL plasma particles can enter the brain via SR-BI-mediated uptake and trans-endocytosis [103]. The Lee team’s research further substantiated this view [47], HDL first binds to SR-BI on the endothelial apical membrane, then it forms endosomes via dynamin- and cholesterol-dependent invagination. These endosomes are transported intracellularly to the side of the basement membrane, where they release HDL into the brain tissue. This process is independent of clathrin, caveolin, or PDZ Domain Containing 1 (PDZK1) and exhibits BBB specificity. Kai Chen et al. discovered that certain small, lipid-poor HDL particles and their surface apolipoproteins (e.g., ApoA-I, ApoE, ApoJ) can still cross the BBB or Blood–CSF Barrier (BCSFB) under specific conditions [283]. Based on our previous discussion, it remains unclear whether HDL particles can fully cross the BBB. The findings mentioned above may be more related to the ability of certain apolipoproteins on HDL to cross the BBB. These findings suggest that enhancing or restoring HDL-like particles’ function holds potential therapeutic and protective value, laying the groundwork for subsequent clinical applications and further research. However, large-scale clinical trial data for HDL-like particles therapy in neurodegenerative diseases remain limited. Many studies are still restricted to animal models and early clinical phases, with major issues such as neurological indications, safe dosage determination, and functional outcome measures yet to be addressed. The path ahead is still long and challenging.

Limitations

Despite HDL/HDL-like particles’ significant potential in regulating BBB function, current research exhibits limitations. First, the mechanism of HDL crossing the BBB remains incompletely elucidated, particularly the precise regulatory pathways of SR-BI-mediated trans-endocytosis, which require further investigation. Current research predominantly relies on in vitro and animal studies, with limited investigations into its specific functions and mechanisms in humans. This makes it difficult to fully reflect the complexity of the human BBB and disease heterogeneity. Furthermore, HDL/HDL-like particles exhibit high structural and functional heterogeneity, and the variations in HDL/HDL-like particles heterogeneity and their functional changes under different physiological and pathological states remain poorly understood. Research on the interaction between HDL-like particles and the glymphatic system is scarce, with theoretical speculation predominating over direct experimental evidence. Furthermore, technical limitations hinder the precise quantification and analysis of subtypes of HDL/HDL-like particles and their functions, restricting in-depth functional studies and hindering the development of biomarkers based on HDL/HDL-like particles. Therefore, integrating multi-omics strategies with clinical research to overcome translational barriers is urgently needed.

Conclusions

HDL and HDL-like particles within the brain play multiple critical roles in maintaining BBB homeostasis and protecting central nervous system health. In summary, HDL-like particles not only participate in the precise transport and metabolic balance of cholesterol within the brain but also contribute to maintaining the structural and functional integrity of the BBB through various mechanisms. These include anti-inflammatory and antioxidant effects, as well as regulation of endothelial tight junctions and glial cell polarization. Key functional components such as ApoA-I, ApoE, ApoM/S1P, and ApoJ synergistically contribute to this process. Specifically, ApoA-I significantly enhances endothelial stability through SR-BI-mediated cholesterol efflux and the activation of eNOS. At the same time, ApoM-bound S1P preferentially activates the S1PR1 signaling pathway, reinforcing the barrier function of intercellular junctions. Conversely, individuals harboring the ApoEε4 allele face heightened risks of BBB structural damage. Key mechanisms involve the abnormal activation of inflammatory cascades (e.g., the CypA-NF-κB-MMP-9 pathway), lipid metabolism disorders, and disrupted glia-endothelial cell interactions, thereby accelerating the progression of neurodegenerative diseases.

The dynamic equilibrium between protective and destructive signaling pathways maintains the integrity of the BBB. The S1P–S1PR1 pathway enhances tight junctions and stabilizes endothelium; the SR-BI–eNOS pathway may contribute to BBB protection, though direct evidence in brain endothelium remains limited. Conversely, the CypA-MMP-9 and RhoA/ROCK pathways promote inflammation and cytoskeletal contraction, driving barrier disruption. Multiple studies confirm this antagonistic relationship: S1PR1 activation upregulates adhesion molecules in endothelium and pericytes, strengthening endothelial tight/adhesion junctions to reduce BBB permeability and limit inflammatory cell infiltration [232, 284]. Conversely, excessive RhoA/ROCK signaling induces actin reorganization and desquamation of tight junctions, markedly increasing BBB leakage [255]. Within this dynamic equilibrium, ApoE/LRP1 signaling occupies a pivotal hub position, simultaneously inhibiting the CypA-MMP-9 cascade and restricting RhoA activity. This coordination of protective and destructive signals maintains BBB homeostasis [193, 194, 285]. Notably, this regulatory hub fails in the presence of ApoE4 or LRP1 deficiency, allowing abnormal activation of both the CypA-MMP-9 and RhoA pathways. This imbalance between barrier protection and destruction mechanisms accelerates BBB dysfunction and associated pathological progression [50, 170].

Although the mechanisms of action for the aforementioned pathways have been extensively validated in cellular and mouse models, several limitations warrant careful consideration. First, numerous conclusions rely on in vitro systems and rodent models; yet, significant differences exist in the structure and function of the BBB across species, which limits the validity of extrapolating model results to humans. Second, the current lack of direct human studies and clinical data targeting these pathways limits our comprehensive understanding of their actual roles in disease progression and therapeutic potential. Third, related drug intervention strategies face challenges in clinical translation, particularly in terms of feasibility and safety. For example, S1P receptor modulators may induce systemic immunosuppression [284], RhoA/ROCK inhibitors may cause hypotension and impair physiological vascular function [255], broad-spectrum MMP-9 inhibition may disrupt normal repair processes [194], and strategies targeting ApoE/LRP1 require careful consideration of functional differences among APOE alleles [193]. Finally, these pathways are intricately intertwined in BBB regulation, and single-target interventions may trigger unintended cascading effects and side effects. Therefore, there is an urgent need to develop BBB models that more closely resemble human physiological conditions and conduct longitudinal clinical studies further to clarify the proper roles of these pathways in humans. This will also assess the feasibility and risks of combined targeted interventions, providing a more robust foundation for precision treatment of BBB-related diseases.

Despite these limitations, the signaling mechanisms mediated by these HDL/HDL-like particles demonstrate significant therapeutic potential in neurological disorders such as neurodegenerative diseases and stroke. Enhancing protective pathways or inhibiting destructive pathways holds promise for improving BBB function and mitigating neural injury. For instance, S1P receptor modulators are already used in the treatment of conditions such as multiple sclerosis. One representative drug, FTY720 (fingolimod), structurally resembles S1P. After conversion in vivo, it acts on the S1PR1 receptor, inhibiting the migration of immune cells out of lymph nodes, reducing central inflammatory cell infiltration, and strengthening tight junctions in the cerebral microvascular endothelium. This reduces damage to the endothelium by inflammatory mediators and the crossing of white blood cells through the BBB, thereby exerting neuroprotective effects [286–288]. Additionally, intervention strategies mimicking HDL function have garnered significant attention. Recombinant HDL preparations or ApoA-I mimetic peptides (e.g., 4F) can partially replicate the vascular protective effects of native HDL. Animal models demonstrate that exogenous HDL reduces BBB permeability and cerebral edema after cerebral ischemia, thereby lowering early mortality and neurological deficits post-stroke [127, 206, 207]. Similarly, interventions targeting the ApoE/LRP1 pathway have yielded preliminary results—an ApoE receptor mimetic peptide activates LRP1 in cerebral hemorrhage models, effectively inhibiting the harmful CypA-MMP-9 cascade, reducing increased BBB permeability around hematomas, and mitigating secondary damage [176]. Therapies targeting the SR-BI pathway are also under investigation, such as enhancing SR-BI expression or function to boost eNOS/NO signaling. This aims to protect cerebral microvessels by promoting endothelial repair and vasodilation [47, 289]. Overall, these L-BBB pathway-based intervention strategies involving HDL/HDL-like particles are predominantly in preclinical or early clinical stages. While specific efficacy requires further validation, they demonstrate promising trends in reducing neuroinflammation and protecting the BBB. This suggests feasibility for developing novel therapies in neurodegenerative diseases like Alzheimer’s and Parkinson’s, as well as acute cerebrovascular events such as ischemic stroke and intracerebral hemorrhage. The aforementioned strategies offer promising therapeutic approaches for maintaining BBB integrity and interrupting the vicious cycle of neuropathology.

The interaction between HDL-like particles and pericytes highlights its crucial role in maintaining the stability of the neurovascular unit. HDL-like particles components such as S1P and ApoE can regulate pericytes-endothelial cell adhesion and basement membrane remodeling, promoting the clearance of neurotoxic products like Aβ-, thereby alleviating pathological burden on the BBB [284, 285, 290]. Emerging evidence suggests that HDL-like particles may synergize with the glymphatic system to clear metabolic waste from the brain, although this hypothesis requires experimental validation [273, 291].

To overcome these challenges, future efforts should focus on resolving HDL/HDL-like particles–BBB interaction under near-physiological conditions and advancing translational research: developing more realistic human-derived models to precisely characterize pathways and drug efficacy, such as iPSC-derived NVU/BBB organoids and “BBB-on-a-chip” microfluidic platforms [292, 293]. Address key mechanistic steps, such as elucidating the decisive role and druggability of SR-BI-mediated transcytosis in the crossing of apolipoproteins across the BBB [47]. Explore combined multi-pathway interventions and precision stratification: enhance protective axes (S1P–S1PR1) while limiting the cascading effects of destructive axes (CypA-MMP-9), tailoring interventions based on disease types and ApoE genotypes; detect and potentially mobilize the brain lymphatic drainage pathway to synergistically enhance metabolic clearance [291]. Overall, advancing along the “model-mechanism-stratification-combination” strategy pathway holds promise for translating HDL/HDL-like particles-mediated barrier protection into verifiable clinical protocols targeting BBB dysfunction-related diseases such as AD and stroke.

Author contributions

Manuscript design: YW; writing original draft: PZ; data collection: PZ, LX; figure preparation: PZ, LX; supervision, review and editing: LX, YW. All approved the final version of the manuscript.

Funding

This work was supported by Yunnan Clinical Center for Emergency traumatic disease, 2024YNLCYXZX0003 (to YW); National Natural Science Foundation of China, 82460370 (to YW); The Innovative Team of Yunnan Province, 202305AS350019 (to HY).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.