Vascular Cell Senescence Contributes To Blood-Brain Barrier Breakdown

Abstract

Background and Purpose

Age-related changes in the cerebrovasculature, including blood-brain barrier (BBB) disruption, are emerging as potential risks for diverse neurological conditions. Since the accumulation of senescent cells in tissues is increasingly recognized as a critical step leading to age-related organ dysfunction, we evaluated whether senescent vascular cells are associated with compromised BBB integrity.

Methods

Effects of vascular cell senescence on tight junction (TJ) and barrier integrity were studied using an in vitro BBB model, composed of endothelial cells (ECs), pericytes (PCs) and astrocytes. Additionally, TJ coverage in microvessels and BBB integrity in BubR1 hypomorphic (BubR1^H/H) mice, which display senescence cell-dependent phenotypes, were examined.

Results

When an in vitro BBB model was constructed with senescent ECs and PCs, TJ structure and barrier integrity (evaluated by transendothelial electrical resistance and tracer efflux assay using sodium fluorescein (NaF) and Evans blue (EB)-albumin) were significantly impaired. ECs and PCs from BubR1^H/H mice had increased SA-β-gal activity and p16^INK4a expression, demonstrating an exacerbation of senescence. The coverage by TJ proteins in the cortical microvessels were reduced in BubR1^H/H mice, consistent with a compromised BBB integrity from permeability assays. Importantly, the coverage of microvessels by end-feet of aquaporin 4 (AQP4)-immunoreactive astrocytes was not altered in the cortex of the BubR1^H/H mice.

Conclusions

Our results indicate that accumulation of senescent vascular cells is associated with compromised BBB integrity, providing insights into the mechanism of BBB disruption related to biological aging.

Introduction

The blood–brain barrier (BBB) is a diffusion barrier¹ that maintains the homeostasis of the central nervous system by regulating the entry of circulating molecules and peripheral cells into the brain.² In cerebrovasculature, the central nervous system (CNS) endothelial cells (ECs) form tight junctions to limit paracellular transport.³ Proper regulation and maintenance of this EC barrier integrity inside the vessels is an essential feature of the BBB.³ In addition, it is increasingly recognized that the ECs are covered by pericytes (PCs), astrocyte end-feet, and the capillary basement membrane, all of which must also be intact for proper BBB stability and function.¹ Disruption of this barrier system has been implicated in various neurological conditions including ischemic stroke⁴ and Alzheimer’s disease (AD).⁵ BBB dysfunction exacerbates these disorders by allowing an aberrant influx of potentially harmful plasma constituents and cells, initiating a cascade of events including neurotoxicity, neuro-inflammation and neurodegeneration. Importantly, BBB integrity is compromised during aging,^{6, 7} which likely precedes the onset of mild cognitive impairment and AD.^{2, 6} However the underlying mechanism of how aging disrupts the BBB integrity remains unclear.

While aging is a highly complex biological process mediated by multiple mechanisms, it is increasingly recognized that senescent cells accumulate in organs, where they may play a role in tissue aging.^{8, 9} Senescence limits the regeneration potential of tissues. Furthermore, the accumulation of senescent cells in organs changes the surrounding microenvironment and compromises tissue repair/renewal by releasing a variety of matrix metalloproteinases and inflammatory cytokines, which are referred to as the senescence-associated secretory phenotype.^{8, 10} Although increased numbers of senescent vascular smooth muscle cells (VSMCs) and ECs were detected in aged peripheral vessels and atherosclerotic lesions,^11-13 our knowledge on the relationship between cell senescence and the cerebrovascular system including the BBB is limited. Therefore, we utilized an in vitro BBB model composed of ECs and PCs with senescence phenotypes and performed a series of assessments to evaluate whether the accumulation of senescent vascular cells is associated with deteriorated BBB integrity. Furthermore, we analyzed senescent effects on BBB integrity in vivo using the accelerated aging BubR1 hypomorphic (BubR1^H/H) mouse model, in which cellular senescence is causally implicated in age-related features.¹⁴

Materials and Methods

Primary Cultures of Mouse Endothelial Cells and Pericytes

Mouse brain ECs and PCs were isolated from the brains of 5 to 8-week-old C57BL/6 WT, BubR1^+/+ and BubR1^H/H mice or 40 to 50-week-old C57BL/6 WT mice by enzymatic digestions as described previously^{15, 16} with slight modifications. The purity of P1 ECs and P4 PCs when analyzed by CD31 (1:100, abcam) or platelet-derived growth factor receptor-β (PDGFR-β) (1:50, R&D Systems) immunofluorescence staining was > 95%, respectively. Detail methods are available as supplementary materials.

In vitro BBB Model

In vitro BBB model was constructed as described.^{15, 17} Briefly, PCs were seeded on the bottom side of the collagen I-coated polycarbonate membrane of a Transwell insert (0.4-μm pore size; Costar, Corning) at a density of 1.5 × 10⁴ cells/cm². PCs were allowed to attach firmly for 1 day and ECs were seeded on the upper side of the insert at a density of 1.5 × 10⁵ cells/cm² on the next day. Three days before the ECs were seeded onto the membrane, astrocytes had been seeded (1.0 × 10⁵ cells/cm²) on poly-D-lysine-coated plate and maintained in astrocyte culture medium. Finally, the Transwell inserts with ECs and PCs were transferred into the 6-well (for western blot or immunofluorescence analysis) or 24-well plates (for evaluation of barrier integrity) containing astrocytes. The cells were refed with the EC medium (mBEC medium II) in the upper chamber and the pericyte medium in the lower chamber. Day 0 was defined by EC plating day on membrane and the medium in both of the chambers was changed to fresh media at day 3. Main analysis was performed on day 5.

Results

Preparation and Characterization of Senescent Vascular Cells

Senescent cells commonly have two distinguishing characteristics; 1) induction of senescent-associated β-galactosidase (SA-β-gal) activity which reflects the enlarged lysosomal compartment of senescent cells and 2) increased cyclin-dependent kinase inhibitors, p21 and p16^INK4a, which are thought to mediate permanent cell cycle arrest.^{18, 19} Pulse treatment with H₂O₂ is a commonly used method for stress-induced cell senescence.²⁰ Consistently, P1 ECs treated with 50 μM H₂O₂ showed decreased cell growth compared to controls (Supplemental Figure IA and IB). However, H₂O₂ treatment failed to consistently increase the levels of p21 and p16^INK4a in ECs, while P1 ECs treated with H₂O₂ showed increased p21 levels at day 5; rather significant increased levels of p16^INK4a were detected in P1 ECs compared to P0 ECs even in the absence of H₂O₂ treatment (Supplemental Figure IC). On the other hand, 100 μM H₂O₂-treated P4 PCs showed increased p16^INK4a levels accompanying a decreased growth rate, confirming their senescence (Supplemental Figure II). These results indicate that the efficacy of H₂O₂ to induce senescence phenotype differs depending on cell types and that passaging procedures significantly affect the cell cycle regulation, particularly for ECs.

We next compared the phenotypes of primary cells from young (5 to 8-week-old) and middle-aged mice (40 to 50-week-old). Using the same validation methods for H₂O₂-treated cells, we found that both ECs and PCs from middle-aged mice showed significant increases in SA-β-gal activity (Figures 1A and 1B) and p16^INK4a and p21 levels (Figure 1C), together with decreased cell growth (Figure 1D). The expression levels of cell specific markers for ECs (CD31) and PCs (PDGFR-β) did not differ between the primary cells from young and middle-aged mice (Supplemental Figure III). Given that ECs and PCs from middle-aged mice exhibited features reminiscent of senescence, these cells were used to examine how accumulation of senescent vascular cells affects the barrier integrity in an in vitro BBB model. ECs and PCs from aged mice (>60-week-old) were less viable/proliferative and highly vulnerable to passaging, suggesting that they had an even further exacerbation of the senescent phenotypes, however they were not suitable for the in vitro studies (data not shown).

Figure 1. Characterization of cerebrovascular endothelial cells and pericytes from young and middle-aged mice.

A, Senescence-associated β-galactosidase (SA-β-gal) activities were detected in primary ECs and PCs from young (5 to 8-week-old) and middle-aged (40 to 50-week-old) C57BL/6 WT mice. B, Number of SA-β-gal-positive cells was quantified comparing ECs and PCs from young and middle-aged mice. C, The mRNA levels of p16^INK4a and p21 in the cells were quantified by RTPCR. D, Number of cells was counted at 5 and 10 days after plating of the cells. Bar represents 200 μm. Data are plotted as mean ± S.E.M. (n = 3). *, p < 0.05, **, p < 0.01, n.s., not significant. Student’s t test was employed for the statistical analysis.

Vascular Cell Senescence Impairs the Barrier Integrity in an in vitro BBB Model

To investigate the effects of cellular senescence on BBB function, we examined the barrier integrity of an in vitro BBB model composed of ECs/PCs from middle-aged (senescent BBB model) or young mice (young BBB model). The Transwell insert where ECs and PCs had been plated was co-cultured with primary astrocytes to mimic neurovascular unit conditions (Figure 2A). Under these co-culture conditions, transendothelial electrical resistance (TEER) reflecting the paracellular barrier function increased in both senescent and young BBB models depending on the length in culture. The values reached a peak at day 5 in co-cultures and started decreasing afterwards in young BBB models; while in senescent BBB models they increased and plateaued around day 5 to 7 in co-cultures (Supplemental Figure IV). The peak TEER values, recorded on day 5 in co-cultures, were significantly lower in the senescent BBB model (Figure 2B). Additionally, permeability coefficients of NaF and EB-albumin were significantly higher in the senescent BBB model on day 5 in co-cultures, confirming the lower barrier integrity (Figure 2B). Together, these results indicate that functional changes associated with vascular cell senescence lead to impaired BBB integrity in an in vitro BBB model.

Figure 2. Cellular senescence attenuates barrier integrity and endothelial TJ in an in vitro BBB model.

A, Left panels: Schematic representation of the triple co-culture for in vitro BBB model. ECs were cultured on semipermeable filter inserts and PCs were plated on the bottom side of the filters, while astrocytes were cultured into the well of culture plate. Right panels: Representative ECs, PCs and astrocytes cultures stained for CD31 (endothelial cell marker; red), PDGFR-β (pericyte maker; green) and GFAP (astrocyte marker; green), respectively. Nuclei were counterstained with DAPI (blue). B, In vitro BBB models composed of ECs/PCs from middle-aged mice (senescent BBB model) or young mice (young BBB model) were prepared. Barrier integrity of standard and senescent BBB models was evaluated by TEER measurement (leftl), permeability coefficient for NaF (middle) and EB-albumin (right) at 5 days after plating for co-culture, respectively. C, ECs in standard and senescent BBB models were subjected to staining for ZO-1, occludin or Claudin-5 (green) at day 5 of culture. White arrows indicate the tight junctions with frayed appearance. Dotted white arrows indicate the abnormal cells with the smaller cytoplasm highly immune-reactive for TJ proteins. D, Percentages of ECs with frayed borders as demonstrated by staining for TJ proteins were calculated in standard and senescent BBB models. Data are plotted as mean ± S.E.M. (n = 3-4). *p < 0.05. Student’s t test was employed for the statistical analysis.

Vascular Cell Senescence Alters TJ Structures

To investigate the mechanisms for the lower barrier integrity in senescent BBB models, TJ structure patterns and distributions of TJ proteins were examined in both BBB models, because these components have been shown to impact on TEER in ECs.^{21, 22} When ECs in BBB models were immunostained for TJ proteins (ZO-1, occludin and Claudin-5) at day 5 of co-culture, TJ borders were smooth and predominately localized at the cell–cell junctions in young BBB models (Figure 2C). In contrast, they were disorganized in the senescent BBB model (Figure 2C, white arrows and 2D). The distributions of occludin could hardly be detected in cell borders, but was diffusely distributed in the cytoplasm in the senescent BBB model (Figure 2C). Furthermore, there was an increase in the number of abnormal cells with the smaller cytoplasm highly immune-reactive for TJ proteins, likely representing less viable or apoptotic ECs (Figure 2C, dotted white arrows). When analyzed by western blotting, the expression levels of TJ proteins did not differ between young and senescent BBB models (Supplemental Figure V). Together, these results indicate that reduced barrier integrity in the senescent BBB model is associated with altered TJ structure and distribution in the EC layer, while the total expression levels of TJ proteins did not differ significantly.

Impaired BBB Integrity and Reduced TJ Protein Coverage in BubR1^H/H Mice

Growing evidence suggests that lack of genomic integrity contributes to the aging process in both humans and mice.²³ Since BubR1 is an essential component of the spindle assembly checkpoint proteins during mitosis, BubR1^H/H mice, in which BubR1 expression is reduced to 10% of the normal level, exhibit progressive aneuploidy and development of specific early aging-associated phenotypes, including short life span, cachectic dwarfism, lordokyphosis, cataracts, loss of subcutaneous fat, and impaired wound healing.²⁴ Thus, to investigate the association between vascular cell senescence and BBB integrity in vivo, we analyzed brain microvessels (MVs) in BubR1^H/H mice. Since it has already been shown that senescent cells accumulate in several tissues in BubR1^H/H mice,²⁴ first we sought to establish whether primary cultures of ECs and PCs from 5 to 8 week-old BubR1^H/H mice display increased indicators of senescence. As expected, both ECs and PCs from BubR1^H/H mice demonstrated significantly increased SA-β-gal activity compared to those from BubR1^+/+ littermates (Figures 3A and 3B). Furthermore, mRNA levels of p16^INK4a were up-regulated in the ECs and PCs from BubR1^H/H mice (Figure 3C). Together, these results confirm that BubR1 insufficiency accelerates senescence of ECs and PCs in mice.

Figure 3. Increased brain vascular cell senescence in BubR1^H/H mice.

A, SA-β-gal activities were detected in ECs and PCs from BubR1^+/+ mice and BubR1^H/H mice (5 to 8-week-old). B, Number of SA-β-gal-positive cells was quantified comparing ECs and PCs from BubR1^+/+ mice and BubR1^H/H mice. C, The mRNA levels of p16^INK4a and p21 in the cells were quantified by RT-PCR. Bar represents 200 μm. Data are plotted as mean ± S.E.M. (n = 4). *, p < 0.05, **, p < 0.01, n.s., not significant. Student’s t test was employed for the statistical analysis.

To investigate features associated with cerebrovascular senescence in vivo, we then compared the BBB function in BubR1^H/H and BubR1^+/+ mice. When BBB permeability was evaluated using an EB technique, higher leakage of the dye was observed around the MVs of 5-month-old BubR1^H/H mice compared with BubR1^+/+ littermates (Figure 4A). Consistently, brain homogenates from 5-month-old BubR1^H/H mice had significantly higher BBB permeability to both peripherally administered NaF (Supplemental Figure VI) and EB (Figure 4A, Right panel) compared to BubR1^+/+ littermates.

Figure 4. Impaired BBB integrity and reduced TJ protein coverage in BubR1^H/H mice.

A, Left panels (upper): Representative confocal images and fluorescence histograms showing perivascular leakage of EB dye in the cortex of 5-month-old BubR1^+/+ littermates and BubR1^H/H mice. EB dye (2%; 4 mL/kg of body weight) was injected intraperitoneally, and EB leakage in the brains was analyzed after 6 hours of injection. Left panels (lower): Fluorescence intensities across the section (white lines) including MVs stained with anti-collagen type IV antibody (red) and EB dye (green) were quantified to confirm the EB leakage around MVs. Right panel: Quantifications of extravasated EB dye in the brains of 1- and 5-month-old BubR1^+/+ littermates and BubR1^H/H mice. B, C, D, Frozen sections of cortex and hippocampus from 5-month-old BubR1^+/+ littermates and BubR1^H/H mice were co-stained for TJ proteins (green; ZO-1, occludin or Claudin-5) and an endothelial marker Isolectin-B4 (red). TJ coverage was quantified against EC layers as determined by Isolectin-B4 immunoreactivity in different brain regions. Bar represents 100 μm. Data are plotted as mean ± S.E.M. (n = 4-5). *, p < 0.05, **, p < 0.01, n.s., not significant. Student’s t test was employed for the statistical analysis.

Finally, to evaluate TJ formations, we performed immunostaining for TJ proteins including ZO-1, occludin and Claudin-5 in brain sections from BubR1^H/H mice and BubR1^+/+ littermates (Figure 4B, 4C and 4D). While most of the MVs were accompanied with strong, continuous and linear staining patterns of TJ proteins in BubR1^+/+ littermates, BubR1^H/H mice exhibited a higher incidence of punctate or fragmented staining of ZO-1 and occludin in MVs at 5 months of age (Figure 4C and 4D). When ZO-1 coverage area was normalized by the length of Isolectin B4-positive ECs, an 18% reduction in ZO-1 coverage was detected in BubR1^H/H mice (Figure 4D). Similarly, occludin and Claudin-5 coverage in BubR1^H/H mice were 15% and 8% lower than those of BubR1^+/+ littermates, respectively (Figures 4B and 4C). These differences in TJ protein coverage were predominantly observed in cortex and hippocampus while no differences were detected in TJ protein coverage in brainstem or cerebellum (Figures 4B, 4C and D and Supplemental Figure VII). Additionally, isolated brain MVs from cortices of BubR1^H/H mice and BubR1^+/+ littermates also confirmed the reduced TJ protein coverage in BubR1^H/H mice (Supplemental Figure VIII). Together, these results indicate that cerebrovascular senescence induced in BubR1^H/H mice is associated with impaired BBB integrity and reduced TJ protein coverage.

Accelerated Gliosis and Increased Perivascular Astrocytes Do Not Alter AQP4-positive Astrocytic End-feet Coverage of Microvessels in BubR1^H/H Mice

In addition to ECs and PCs, astrocytes also contribute to BBB maintenance. Therefore, we asked if there are differences in astrocyte-vascular interaction between BubR1^H/H mice and controls. To evaluate astrogliosis in BubR1^H/H mice, we first examined the expression levels of an activated astrocyte marker, glial fibrillary acidic protein (GFAP), and an activated microglia marker, ionized calcium-binding adapter molecule 1 (Iba1). Immunostaining revealed that GFAP-positive astrocytes and Iba1-positive microglia in the cortex were significantly increased in BubR1^H/H mice (Figures 5A and 5B). Western blotting also showed an increase of GFAP in BubR1^H/H mice at 1 and 5 months of age (Figure 5C), while no difference was detected in pre-synaptic vesicle protein synaptophysin and the post-synaptic density protein 95 (PSD-95) levels.

Figure 5. Accelerated activation of astrocytes and microglia in BubR1^H/H mice.

A, GFAP and Iba1 were immunostained in cortices of BubR1^+/+ littermates and BubR1^H/H mice at 5 months of age. Bar represents 1 mm. B, Slides were subjected to digital analysis using the Aperio ImageScope software and the expression levels of GFAP and Iba1 were quantified. C, Expression levels of GFAP, synaptophysin (Syp) and PSD-95 were analyzed by Western blotting in the brains of littermate controls and BubR1^H/H mice at 1 and 5 months of age. Data are plotted as mean ± S.E.M. (n = 4-5). *, p < 0.05. n.s., not significant. Student’s t test was employed for the statistical analysis.

When astrocytes in the perivascular regions were histologically investigated, BubR1^H/H mice exhibited strong GFAP immunoreactivity and hypertrophic morphology with thick proximal processes at the ages of 1 and 5 months (Figure. 6A), which are characteristics of reactive astrocytes.²⁵ Furthermore, astrocytes from BubR1^H/H mice had multiple processes surrounding the adjacent MVs that were not evident in littermate controls (Figure. 6A). When the percentages of collagen type IV-positive MVs associated with reactive astrocytes were quantified, they were significantly increased in BubR1^H/H mice in an age-dependent manner (Figure 6B). However, there was no difference in AQP4-positive astrocytic end-feet coverage of Isolectin-B4-positive MVs/ECs between BubR1^H/H mice and BubR1^+/+ littermates at 5 month (Figure 6C and 6D). Together, these results indicate that accelerated gliosis and increased perivascular astrocytes do not affect astrocytic end-feet coverages of MVs/ECs in BubR1^H/H mice.

Figure 6. Increased association of MVs with reactive astrocytes in BubR1^H/H mice without affecting MV coverage by aquaporin 4-positive astrocytic end-feet.

A, Representative confocal images of cortices from 1- and 5-month-old BubR1^+/+ littermates and BubR1^H/H mice co-stained with anti-GFAP (green) and anti-collagen type IV antibody (red). B, Quantifications of MVs associated with reactive astrocytes in the cortices. Reactive astrocytes were identified by strong GFAP immunoreactivities and hypertrophic morphologies with thick proximal processes. The percentage of MVs associated with reactive astrocytes were calculated by dividing the numbers of collagen type IV-positive MVs that overlap with reactive astrocytes by those of total collagen type IV-positive MVs. C, Representative confocal images of cortices from 1- and 5-month-old BubR1^+/+ littermates and BubR1^H/H mice co-stained with anti-AQP4 (green) and Isolectin B4 (red). D, Quantifications of MVs covered by AQP4-positive astrocytic end-feet in the cortices. Bar represents 100 μm. Data are plotted as mean ± S.E.M. (n = 4-5). **, p < 0.01. n.s., not significant. Student’s t test was employed for the statistical analysis.

Discussion

Recently, a concept which connects cellular senescence and age-related tissue dysfunction has emerged.^{8, 9, 26} Indeed, in the central nervous system, involvement of senescence in neurons, astrocytes and microglia is increasingly recognized in the development of brain aging and/or age-related neurological disorders.^27-30 However, whether cellular senescence impacts BBB integrity has not been established. In this study, we show that accumulation of senescent vascular cells results in impaired barrier integrity and altered TJ structure using an in vitro BBB model. The use of primary cells rather than cell lines to replicate the in vivo neurovascular unit is crucial, especially when investigating the physiological BBB functions. However, these cells are fragile and the phenotypic changes during in vitro culture and/or passage could potentially affect the outcome measures. Therefore, we sought to prepare senescent cells with shorter culture periods. To this end, we have utilized ECs/PCs from middle-aged mice to construct in vitro senescent BBB models. These in vitro results were reinforced by our in vivo observations in BubR1^H/H mice; the presence of vascular cell senescence accompanied impaired BBB integrity and reduced brain microvessel TJ coverage.

In an in vitro BBB model, the accumulation of senescent vascular cells was accompanied by lower barrier integrities together with altered TJ structures and the shift of intracellular localization of occludin from the membrane to cytoplasm. Since the frayed TJ structure has been shown to result in lower barrier integrity compared to continuous TJ,²¹ it is likely that increased frayed TJ formations in ECs was, at least in part, causally implicated in lower TEER values in senescent BBB models. On the other hand, it is still unclear whether the changes in intracellular localization of occludin affected its assembly into the tight junctions in senescent ECs. Given phosphorylation of occludin correlates with occludin localization and integrity of TJs,³¹ it would be interesting to examine if the observed increase in cytoplasmic occludin in senescent ECs is associated with phosphorylation of occludin and how this process is affected by senescence. In addition, since expression levels of adhesion molecule 1 (VCAM-1) and intercellular adhesion molecule 1 (ICAM-1) are declined in a senescence-dependent manner in ECs,³² this senescence-induced suppression of adhesion molecules may contribute to lower barrier integrities by influencing TJ formation. EC senescence also facilitates the production of endothelin-1³³ and decreases endothelial NO³⁴, which is potentially harmful to the BBB.^{35, 36} Moreover, increased activation of NF-kB p65, and increased susceptibility to apoptosis were observed in senescent ECs.³² Thus, senescence in ECs likely disturbs BBB functions both through direct effects on EC phenotype and secondarily by inducing susceptible environments to stress due to their pro-inflammatory status.

Senescent VSMCs contribute to the pro-inflammatory environment through an upregulation of interleukin-6 (IL-6), chemokines and innate immune receptors³⁷. Since both PCs and VSMCs are vascular mural cells belonging to the same cell lineage³⁸, it is conceivable that the senescence-associated secretory phenotype of PCs also exacerbates the senescence-related dysfunctions of BBB through their interaction with ECs. Interestingly, a positive correlation between a degree of PC injury as measured by soluble PDGFR-β levels in the cerebrospinal fluid (CSF) and the BBB permeability in the hippocampus were detected in mice and humans⁶. Therefore, the senescence of PCs may be one of the critical steps to induce age-dependent BBB breakdown, although further studies are needed.

Markedly decreased BubR1 expression in multiple tissues has been observed in wild-type mice during aging²⁴. Importantly, in BubR1^H/H mice, removal of p16^INK4a-positive senescent cells has been shown to prevent and/or delay dysfunction in several tissues such as the eye, adipose tissue, skeletal muscle,¹⁴ suggesting that these mice can be used as models to explore the potential association between cell senescence and age-related phenotypes. In BubR1^H/H mice, we found impaired BBB integrity and reduced brain microvessel TJ coverage, where primary cultures of cerebrovascular cells from these mice displayed senescence-associated phenotypes. Consistently, BubR1^H/H mice show age-related vascular phenotypes in peripheral larger arteries.³⁹ Since there was no significant difference in BBB permeability at 1 month of age (Figure 4A and Supplemental Figure VI), the disturbed BBB integrity in adult BubR1^H/H mice likely depends on their accelerated aging phenotype rather than defects in brain vascular development during embryogenesis or the early postnatal period. Thus, together with in vitro findings, our results indicate that vascular cell senescence per se is likely a mediator of BBB breakdown in these mice. On the other hand, it should also be considered that functional changes of other cell types affect the BBB, because the neurovascular unit is comprised of endothelial cells, astrocytes, pericytes, neurons, microglia and other cells.^{2, 40} In this regard, while we observed significantly activated perivascular astrocytes in BubR1^H/H mice (Figure 6), the coverage of MVs by end-feet of AQP4-immunoreactive astrocytes was not altered in the cortex of BubR1^H/H mice. Thus, structural alterations in astrocytic end-feet do not seem to be critically involved in BBB disruption of these mice. It is possible, however, that the senescence of astrocytes may indirectly contribute to the loss of BBB integrity by reducing the ability to provide a glial-derived barrier-inducing factor⁴¹ as well as increasing the production of reactive oxygen species (ROS)⁴² and inflammatory cytokines. ²⁷

There are several limitations to this study. First, we cannot completely exclude the possibility that another mechanism which was not mediated by cell senescence contributed to the observed BBB disruption, since we used certain biological stress (organismal aging, genomic instability due to BubR1 hypomorphism) to induce cell senescence both in vitro and in vivo which may affect other physiological processes. In this regard, we may take advantage of BubR1^H/H mice where removal of p16^INK4a-positive senescent cells has been shown to delay tissue dysfunction in several peripheral tissues.¹⁴ In vivo “rescue” experiment to examine if removal of senescent cells could reverse the impaired BBB integrity would further interrogate the importance of functional changes associated with vascular cellular senescence on BBB integrity. Second, the use of murine primary ECs might be less potent as a model to detect the TJ protein dynamics by senescence since the expression of TJ molecules in murine ECs dramatically decreases during the isolation process and under culture conditions compared to acutely isolated ECs. Indeed, while we observed significant changes in intracellular occludin distribution by immunocytochemistry, we could not detect any difference in biochemical properties of occludin (e.g. occludin oligomeric assemblies reported in rats under stressed condition⁴³) by SDS-PAGE (Supplemental Figure V), which might be because of the lower expression levels of TJ molecules in our murine ECs compared to the other cell/species used. Therefore further studies are needed to better understand the relationship between cell senescence and TJ protein dynamics in ECs. Third, due to the structural nature of this BBB model, there was a technical limitation to maintain three cell types (i.e., ECs, PCs, and astrocytes) in three different culture media suitable for each cell population. ECs in the upper chamber were cultured in the EC medium, while PCs and astrocytes in the lower chamber were maintained in the pericyte medium. Thus, a potential mixture of these media during co-cultures may impact phenotypes of these cells in an unexpected way. Since PCs need to be maintained in the medium containing lower FBS concentration (2%) to avoid their differentiation,⁴⁴ we used the pericyte medium to feed PCs and astrocytes in the lower chamber, which may also affect astrocyte conditions. While this triple co-culture setting is widely used as the method of choice for reproducing the neurovascular unit in a dish^{15, 16}, the results obtained through this model should be interpreted with these caveats, and compensated by in vivo experiments.

In summary, our study reveals that vascular cell senescence accompanies impaired BBB integrity in vitro and in vivo, providing implications regarding the mechanism of BBB disruption induced by aging. Since aging-related BBB dysfunction likely triggers the pathogenic pathways for cerebrovascular diseases as well as exacerbates the neuronal damages caused by these disease conditions, our findings also warrant the exploration of cerebrovascular senescence as a potential target for interventions and risk stratifications toward prevention or treatment of neurological conditions related to BBB disruption, such as stroke and AD. Finally, pharmacological approaches or gene therapy to suppress and/or restore vascular cell senescence at BBB may have potential to be novel therapeutic strategies for these diseases.