Melatonin protects blood-brain barrier integrity and permeability by inhibiting matrix metalloproteinase-9 via the NOTCH3/NF-κB pathway

The pathophysiological mechanism of white matter hyperintensities of cerebral small vessel disease (CSVD) includes an impaired blood-brain barrier (BBB) with increased permeability. Neuroinflammation likely contributes to the disruption of the BBB in CSVD. Therefore, understanding the molecular mechanism of how neuroinflammation causes BBB damage is essential to preventing BBB disruption in CSVD. Matrix metalloproteinase 9 (MMP-9) contributes to BBB damage in neuroinflammatory diseases. In this study, we observed that interleukin-1β (IL-1β)-induced MMP-9 secretion in pericytes increased BBB permeability to sodium fluorescein (Na-F) by damaging the disruption of VE-cadherin, occludin, claudin-5, and zonula occludin-1 (ZO-1). Melatonin reduced BBB permeability to Na-F and inhibited the disruption of the adherens and tight junction proteins. Melatonin also downregulated MMP-9 and upregulated tissue inhibitor of metalloproteinases 1 (TIMP-1) gene expression, which decreased the MMP-9/TIMP-1 ratio. In addition, nuclear translocation of NF-κB/p65 induced by IL-1β in pericytes upregulated MMP-9 expression, which was inhibited by the NF-κB inhibitor PDTC. However, the NOTCH3 inhibitor DAPT significantly inhibited NF-κB/p65 translocation to the nucleus, while melatonin in combination with DAPT significantly prevented NF-κB/p65 translocation than DAPT alone. Our results suggest that melatonin reduced MMP-9-induced permeability of the BBB. Melatonin reduced MMP-9 expression and activity, which was induced by IL-1β through the regulation of the NOTCH3/NF-κB signaling pathway in pericytes, suggesting that pericytes regulate BBB integrity and function.

matter hyperintensities (WMH) are common magnetic resonance imaging (MRI) feature of CSVD [3]. The pathophysiological mechanism of WMH is the breakdown and increased permeability of the blood-brain barrier (BBB) [4,5].
The BBB is composed of endothelial cells, astrocytes, and pericytes in the central nervous system (CNS) [2]. Pericytes are chiefly involved in the regulation of blood flow, maintenance of BBB functional integrity, and stabilization of vessel architecture [6][7][8][9][10][11]. Disruption of the BBB is a key feature associated with neuroinflammatory conditions in CSVD [1]. Matrix metalloproteinases (MMPs), which regulate the structure and function of extracellular matrix molecules under both normal and pathophysiological conditions, are an important mediator of BBB damage [12].
MMPs are a family of zinc-dependent endopeptidases that degrade the basement membranes and extracellular matrix of surrounding cells [13][14][15]. In brain tissues, MMP-2 and MMP-9 are major contributors to BBB damage. More MMP-9 is secreted from brain pericytes than from the BBB endothelial cells and astrocytes, suggesting that pericytes are the major source of MMP-9 in the CNS [12,13,16].
Pericytes are an important cell type in the pathophysiological processes of BBB disruption, including neuronal degenerative changes during agedependent cognitive impairment disorders [6], such as Alzheimer's disease (AD) [17,18], Parkinson's disease (PD) [19,20], multiple sclerosis (MS) [21], and cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) [22][23][24]. Pericytes function similarly in all of these disorders, which include BBB dysfunction, and may be affected by the presence of neuroinflammatory molecules [25]. In addition, brain microvascular pericytes exhibit multipotent stem cell characteristics and can differentiate into astrocytes, neurons, and oligodendrocytes in the presence of bFGF [26,27]. In the neuroinflammatory response that occurs in many neurological diseases, MMP-9 degrades components of the basal lamina, leading to the disruption of the BBB [12]. Pericytes have also attracted attention as neuroinflammatory mediators in the brain [28,29]. Moreover, IL-1β has been shown to play a vital role in the process of neuroinflammation in the neurological pathologies [30]. IL-1β may be a critical mediator of MMP-9 activity and subsequent disruption of the tight junction in intracerebral hemorrhagic models [31]. Many studies have indicated that IL-1β-induced MMP-9 is secreted from pericytes via the NF-κB signaling pathway [28,[32][33][34]. Moreover, IL-1β is involved in tongue squamous cell carcinoma and human dystrophic myogenesis via NOTCH signaling pathways [35,36]. Therefore, we hypothesized that BBB permeability was disrupted by IL-1β inducing MMP-9 secretion from pericytes.
The NOTCH pathway is an evolutionarily conserved signaling system involved in cellular physiology processes of developing and differentiated tissues [37,38]. There are currently four known NOTCH receptors (NOTCH 1-4) in the vertebrates [39]. Each NOTCH receptor plays a different role in physiological processes; Relevant to CSVD, NOTCH3 mutations are the critical causative gene defect in CADASIL, which is an autosomal dominant stroke disorder that results in vascular dementia [40] and is the most frequently inherited type of CSVD [41]. CADASIL presents with small arterial vascular smooth muscle cells and brain capillary pericytes [42,43] that are enriched with NOTCH3 receptors compared with endothelial cells and astrocytes [44]. However, the pathophysiological role of NOTCH3 in brain pericytes remains largely unknown. NOTCH3 regulates multiple NF-κB activation pathways and enhances the release of NF-κB transcription factors from inhibitory components as well as the transacting activity upon downstream promoters [45,46]. NF-κB is a family of transcription factors that includes p50, p52, p65, RelB, and c-Rel [47]. A heterodimer of p65 and p50 subunits was the first NF-κB molecule described and is inhibited by the IκBα protein in unstimulated cells [48]. NF-κB is activated by TNF-α in human dermal fibroblasts, which in turn can activate pro-MMP-2 [49].
Melatonin is a hormone secreted from the pineal gland and released into the cerebrospinal fluid and circulation. In the brain, melatonin binds to melatonin receptors where it exerts various important biological effects [50][51][52]. Melatonin demonstrates neuroprotective effects in models of cerebrovascular diseases through either antioxidant function, anti-inflammatory effects, or inhibition of MMPs. Melatonin plays a pivotal role in oxidative and apoptotic signaling, mitochondrial homeostasis, and calcium excitotoxicity in neurodegeneration. Melatonin may affect the NOTCH signaling pathway through multiple targets involved in the pathogenesis of familial and idiopathic PD and regulate leucine-rich repeat kinase 2 (LRRK2), which is a key molecule involved in PD [53]. The melatonin signaling pathways that protect the integrity of the BBB against MMP-9 disruption are currently unknown. Here, we hypothesize that melatonin reduces the expression and activity of MMP-9 secretion in pericytes by regulating the NOTCH3/NF-κB signaling pathway in the BBB. Our hypothesis suggests that pericytes are involved in the regulation of BBB integrity and function and that melatonin is important for BBB protection.

Characterization of cells in the triple co-culture BBB model by immunofluorescence
Prior to the triple co-culture BBB model experiments, the human brain microvascular pericytes, endothelial cells, and astrocytes were characterized. The shape and size of human brain microvascular pericytes were different from the other two cell types of the BBB. Pericytes in culture spread large with irregular projections and were positive for pericyte-markers α-smooth muscle actin (α-SMA) and NG2 chondroitin sulfate proteoglycan and negative for von Willebrand factor (vWF) and glial fibrillary acidic protein (GFAP) staining ( Figure 1). Human brain microvascular endothelial cells grew in non-overlapping, continuous monolayers and were large, homogenous, closely opposed and polygonal cells with an oval, centrally located nucleus and indistinct cell borders, and were positive immunostaining for an endothelium marker von Willebrand factor (vWF) (Figure 1). Astrocytes were polygonal with long cell processes and characterized by GFAP immunostaining (Figure 1). These results indicated that the in vitro BBB model is composed of cells that can be differentiated by cell-type-specific morphology and protein expression.

Melatonin inhibited the expression levels of MMP-9 in pericytes
We investigated whether pericytes, versus astrocytes and endothelial cells, were the primary producer of MMP-9, (Supplementary Figure 1). The effect of melatonin on the expression of MMP-9, MMP-2, TIMP-1 and TIMP-2 in pericytes was examined by qRT-PCR. The results showed that MMP-9 mRNA was higher in the IL-1βtreated group relative to the control group (Figure 2A, 2E, 2I). Upon pretreatment with the MMP-9/-2 inhibitor SB-3CT, the enhanced MMP-9 expression induced by IL-1β treated was significantly reduced in the IL-1β + SB-3CT group compared with the IL-1β-treated group (Figure 2A). Upon pretreatment with melatonin, the enhanced MMP-9 expression in the IL-1β-treated group was significantly reduced compared with that in the IL-1β + SB-3CT and IL-1β-treated groups, respectively ( Figure 2A). These results suggest that melatonin decreases the IL-1β-induced expression of MMP-9 in pericytes. Figure 1. Characterization of primary cultures by immunofluorescence microscopy. Human brain microvascular pericytes showed positive immunostaining for α-SMA and NG2 but were negative for endothelial or astrocytic markers. Endothelial cells expressed the von Willebrand factor (vWF), while astrocytes were positive for GFAP. Scale bar = 50 μm. AGING The expression of MMPs could be regulated by NF-κB. When the NF-κB inhibitor PDTC was added, the IL-1βenhanced expression of MMP-9 in the IL-1β + PDTC group was significantly reduced relative to that in the IL-1β-treated group ( Figure 2E). Upon pretreatment with melatonin, MMP-9 expression was more significantly reduced in the IL-1β + PDTC + MEL group compared with the IL-1β + PDTC and IL-1β-treated groups, respectively ( Figure 2E). In addition, NOTCH3 signaling has been predicted to modulate NF-κB expression. When the NOTCH3 inhibitor DAPT was added, the IL-1βinduced expression of MMP-9 was significantly reduced relative to that in the IL-1β-treated group ( Figure 2I). Upon pretreatment with melatonin, the enhanced expression of MMP-9 was significantly more reduced in the IL-1β + DAPT + MEL group compared with that in the IL-1β + DAPT and IL-1β-treated groups, respectively ( Figure 2I).
We observed no difference in MMP-2 mRNA level in IL-1β-treated group relative to the control group ( Figure 2C, 2G, 2K), which suggested that the MMP-2 expression was unaffected by IL-1β. When the MMP-9/-2 inhibitor SB-3CT was added to the cell cultures, the expression of MMP-2 was significantly reduced in the IL-1β + SB-3CT group compared with the IL-1β-treated group ( Figure  2C). Upon pretreatment with melatonin, the expression of MMP-2 was not affected in the IL-1β + SB-3CT + MEL group compared with the IL-1β + SB-3CT group; however, its expression significantly decreased compared with that in the IL-1β-treated group ( Figure 2C), suggesting that melatonin did not affect the expression of MMP-2 in the pericytes.
Finally, we observed that DAPT decreased the effect of IL-1β on the MMP-9/TIMP-1 mRNA ratio, and melatonin significantly decreased the effect of IL-1β by in the IL-1β + DAPT + MEL group compared with that in the IL-1β + DAPT group ( Figure 2M). Melatonin alone or in combination with PDTC or SB-3CT significantly decreased IL-1β induction of the MMP-9/TIMP-1 mRNA ratio in the IL-1β + PDTC or WB-3CT + MEL group compared with those in the IL-1β + PDTC and SB-3CT groups ( Figure 2M). These results indicate that melatonin affects not only MMP-9 and TIMP-1 mRNA levels, but also the MMP-9/TIMP-1 mRNA ratio.
Secreted MMPs activity is highly modulated by TIMPs which are the endogenous inhibitors of MMPs. We AGING examined TIMP-1 and TIMP-2 protein levels in pericytes using western blot analysis. When compared with the control group, TIMP-1 was significantly increased in the IL-1β-treated group. However, melatonin pretreatment significantly increased the expression of TIMP-1 protein in the IL-1β + DAPT + Mel group even though DAPT decreased the expression of TIMP-1 in the IL-1β + DAPT group ( Figure 3D, 3F). Results are presented as the mean ± SD, n = 3. * p < 0.01 compared with the control; # p < 0.01 compared with the IL-1β-treated group; ## p < 0.01 compared with the IL-1β + DAPT group; $ p < 0.01 compared with the IL-1β + PDTC group; & p < 0.01 compared with the IL-1β + SB-3CT group. AGING Moreover, there was no difference in TIMP-1 protein level in the IL-1β + PDTC or IL-1β + SB-3CT groups compared with the IL-1β-treated group ( Figure 3D, 3F). Changes in the TIMP-2 protein level were only detected in the IL-1β + SB-3CT and IL-1β + SB-3CT + MEL treatment groups, while no changes were detected in the other six groups ( Figure 3D, 3H).

Melatonin inhibits MMP-9-induced BBB permeability
BBB permeability was determined using Na-F in the Transwell assay. The permeability of the BBB in the IL-1β-treated group obviously increased at different time points compared with that in the control group ( Figure  4A-4C). However, the permeability to Na-F was significantly reduced in the IL-1β + DAPT, PDTC, and SB-3CT groups relative to the IL-1β-treated group after 30 min of treatment ( Figure 4A-4C). Furthermore, the permeability significantly decreased in the IL-1β + DAPT, PDTC, and SB-3CT treatment groups pretreated with MEL relative to the IL-1β + DAPT, PDTC, and SB-3CT groups that did not receive pretreatment ( Figure 4A-4C). These results indicate that melatonin can reduce BBB permeability that is induced by IL-1β, which potentially occurs through the stimulation of MMP-9 activity. Results are presented as the mean ± SD, n = 3. * p < 0.01 compared with the control; # p < 0.01 compared with the IL-1β-treated group; ## p < 0.01 compared with the IL-1β + DAPT group; $ p < 0.01 compared with the IL-1β + PDTC group; & p < 0.01 compared with the IL-1β + SB-3CT group. AGING

Melatonin inhibits MMP-9 via the NOTCH3-NF-κB pathway
We observed NOTCH3 expression in pericytes and not in astrocytes or endothelial cells (Supplementary Figure 2). The qRT-PCR results showed that the IL-1β-treated group exhibited higher NOTCH3 expression relative to the control group ( Figure 7A-7C). When the NOTHC3 inhibitor DAPT was added, the results showed that NOTCH3 and p65 mRNAs were significantly reduced relative to the IL-1β-treated group ( Figure 7A, 7D). When melatonin was added to those growth conditions, NOTCH3 and p65 expression levels were significantly reduced compared with the IL-1β + DAPT group that did not receive the melatonin pretreatment ( Figure 7A, 7D). When the NF-κB inhibitor PDTC was added, p65 mRNA was significantly reduced relative to IL-1β-treated group ( Figure 7E); however, there was no difference in the expression of NOTCH3 mRNA in the IL-1β + PDTC group relative to IL-1β-treated group ( Figure 7B). We hypothesized that NOTCH3 participated in regulating the NF-κB pathway by inducing changes in NOTCH3 mRNA expression between the IL-1β + DAPT and IL-1β + PDTC groups. When melatonin was added, the results showed that NOTCH3 and p65 mRNA were significantly reduced in the IL-1β + PDTC + MEL group compared with the IL-1β + PDTC group ( Figure 7B, 7E), which suggested that melatonin could additionally inhibit NOTCH3 and p65 expression induced by IL-1β in pericytes. When the MMP-9/-2 inhibitor SB-3CT was added, there was no difference in the expression of NOTCH3 and p65 mRNA relative to the IL-1β-treated group ( Figure 7C, 7F).
NF-κB/p65 is a member of the NF-κB transcription factor family. These transcription factors translocate to the nucleus and bind to a specific sequence in the genome to regulate MMPs transcription when cells are stimulated with inflammatory factors. We observed that activated NF-κB/p65 increased in the IL-1β-treated group, while pretreatment with DAPT or PDTC inhibited the nuclear translocation of NF-κB/p65 relative to IL-1 β treatment alone ( Figure 8H, 8J). However, there was no difference in the IL-1β + SB-3CT group relative to the IL-1 β-treated group (Figure 8, 8H, 8J). Next, we also analyzed the cellular location changes of NF-κB/p65 after IL-1 β treatment, pretreated with DAPT, PDTC, and SB-3CT in combination with melatonin by immunostaining method. We observed that NF-κB/p65 was located in the cytoplasm in the control group ( Figure  9A-9C), and translocated into the nucleus in IL-1βtreated cells (Figure 9D-9F). Pretreatment with DAPT and PDTC inhibited the p65 translocation, indicated by changes in immunofluorescence patterns in cells from the IL-1β + DAPT and IL-1β + PDTC groups ( Figure 9G-9I, 9M-9O). This inhibition was not observed in the IL-1β + SB-3CT group ( Figure  9S-9U). Pretreatment with melatonin clearly suppressed the p65 translocation, as seen in the IL-1β + DAPT + MEL, IL-1β + PDTC + MEL and IL-1β + SB-3CT + MEL groups ( Figure 9J-9L, 9P-9R, 9V-9X). Taken together, these results demonstrated that melatonin protects BBB integrity and permeability via the NOTCH3/NF-κB pathway ( Figure 10).

DISCUSSION
The results of our study can be summarized as follows.
The expression levels of MMP-9 mRNA and protein were upregulated in pericytes upon IL-1β stimulation, and the mRNA and protein levels of the endogenous inhibitor TIMP-1 were also upregulated, which increased the MMP-9/TIMP-1 ratio. Conversely, changes in MMP-2 mRNA and protein levels in response to IL-1β stimulation were not observed. Second, melatonin pretreatment decreased the expression levels of MMP-9 mRNA and protein but increased TIMP-1 mRNA and protein levels. Third, increased MMP-9 activity disrupted adherens and tight junctions of the BBB, and increased endothelial cell permeability to Na-F. Fourth, melatonin decreased the MMP-9/TIMP-1 ratio and reversed the   (NICD) expression which activated NF-κB/p65 expression. NF-κB/p65 translocated to the nucleus to induce MMP-9 expression, and MMP-9 disrupted BBB integrity and permeability. When melatonin, DAPT, PDTC, and SB-3CT were added respectively, NOTCH3 expression was significantly inhibited, leading to decreased NF-κB/p65 and MMP-9 expression, which protected BBB integrity and permeability.
damaging effect of MMP-9 had on endothelial cell-cell junctions, providing a safeguard for BBB integrity. Finally, melatonin inhibited NOTCH3, prevented the nuclear translocation of p65, and further reduced MMP-9 gene expression. To our knowledge, these results are the first to demonstrate that IL-1β-induced MMP-9 expression and activity in pericytes limits the disruption of BBB integrity.
MMPs and endogenous TIMPs are important regulators and play key roles in many biological processes and diseases [50]. The role of MMP-9 secreted from pericytes in regulating BBB integrity is not yet well elucidated. In order to investigate this role, we analyzed MMP-9/-2 and TIMP-1/-2 expression in pericytes by qRT-PCR. The results suggested that MMP-9 mRNA level was increased in IL-1β-treated pericytes and was reduced with pretreatments of SB-3CT, PDTC, DAPT. In addition, in combination with melatonin, these inhibitors initiated a further reduction in MMP-9 mRNA and protein but increased TIMP-1 mRNA and protein levels.
Tai et al [54] demonstrated that melatonin decreased neurovascular damage by attenuating MMP-9 activation and increasing TIMP-1 in transient focal cerebral ischemia. We also analyzed MMP-9 and MMP-2 activities in pericytes by gelatin zymography, which indicated that MMP-9 increased in IL-1β-treated pericytes. When the inhibitors SB-3CT, PDTC, and DAPT were added to the cultures in the presence of IL-1β, MMP-9 activity significantly decreased to varying degrees. In addition, SB-3CT, or PDTC, or DAPT in combination with melatonin pretreatment significantly attenuated MMP-9 activity, suggesting that melatonin is a potential MMP-9 inhibitor [55].
The BBB contains an important endothelial cell barrier structure that plays an important role in the maintenance of CNS homeostasis [50,56]. The fundamental basis for barrier function mainly consists of adherens and tight junctions in endothelial cells, which seal off the paracellular pathway between adjacent brain capillary endothelial cells [57,58]. VE-cadherin is the main adherens junction in endothelial cells of all types of vessels [58]. The main tight junction proteins include occludin, claudins, and ZO-1 [56]. Therefore, maintaining and regulating BBB functions are important to assure the integrity of the VE-cadherin adherens junction protein and the tight junction proteins occludin, claudin-5, and ZO-1, which depend on the extent of the endothelial cells barrier permeability [50,59]. Pericytes are critical coordinators of BBB functions and CNS homeostasis, regulate cerebral blood flow, contribute to the secretion of ECM proteins, and provide structural support and angioarchitecture [59]. The BBB endothelial permeability changes involve the activation of MMPs and spatial redistribution of junction proteins [60]. In the BBB, MMP-9 is primarily secreted from pericytes and is associated with BBB disruption in several diseases including ischemic stroke, intracerebral hemorrhage, hemorrhagic transformation, brain edema, Alzheimer's disease and Parkinson's disease [16,61]. Furthermore, oxidants, inflammatory cytokines, and pharmacological agents can provoke MMPs activations that disrupts the adherens junction and the tight junction proteins [50].
Our previous studies suggested that IL-1β increased the permeability of the HUVEC monolayer by disrupting adherens and tight junctions and expanding the junction gap widths [50,62]. In this study, we showed that in IL-1β-induced pericytes, there was a significant increase in MMP-9, which enlarged the intercellular gap by destroying the cell junction elements VE-cadherin, occludin, claudin-5, and ZO-1; this was consistent with the observed increase in BBB permeability to Na-F. When the MMP-9/-2 inhibitor WB-3CT was added, the results showed that the BBB permeability was decreased. Moreover, when the NF-κB inhibitor PDTC was added, the BBB permeability decreased, as well as the expression of MMP-9 and the MMP-9/TIMP-1 ratio; however, no changes were observed in the expression of MMP-2 and NOTCH3. Furthermore, when the NOTCH inhibitor DAPT was added, the BBB permeability also decreased and MMP-9 expression, the MMP-9/TIMP-1 ratio, and p65 expression were reduced, while no changes occurred in MMP-2 expression.
The BBB permeability due to hyperglycemia during early reperfusion in stroke is involved in autophagy-and MMP-2/9-mediated reduction and redistribution of tight junction proteins [63]. The BBB properties were substantially induced by brain pericytes through the change of the expression of tight junction proteins and Na-F permeability. In addition, limited primary cell availability essentially eliminates the possibility of using patient-matched brain pericytes for disease modeling applications. Stebbins et al [27] demonstrated that hPSCderived brain pericyte-like cells in the BBB were indistinguishable from those induced by primary human brain pericytes. Pericyte involvement in CADASIL can result in increased capillary vessels permeability and disturbances in cerebral microcirculation, which can lead to white matter injury. Since capillary pericytes regulate vessel contractility, their degeneration can also cause defective vasomotor reactivity, a phenomenon observed very early in CADASIL, before the development of histopathological changes in vessel walls [23]. Interestingly, following white matter demyelination, pericytes promote differentiation of oligodendrocyte progenitors involved in CNS regeneration via the a2chain of laminin [64]. Moreover, pericytes are involved in the recruitment of neutrophils to the sub-endothelial layer and breaching of the vascular wall by secreting the chemoattractant IL-8 and the metalloproteinases MMP-2 and MMP-9, which break down endothelial cell-to-cell contacts [65]. Pericyte degeneration also causes the breakdown of the BBB, leading to blood-derived toxic substances passing into the central nervous system [6]. As a crucial component of the neurovascular unit, pericytes have diverse functions that regulate capillary blood flow and BBB permeability, as well as BBB stability [66]. Indeed, pericyte-deficient animals do not establish proper vessel boundaries and ultimately cause BBB breakdown and leakage [67]. BBB dysfunction results from tight junction changes, increased transcytosis, loss of trophic support from pericytes or altered hemodynamics [68].
Melatonin is mainly secreted by the pineal gland [51,52] and has powerful antioxidant effects, including strong free-radical scavenging that can regulate several molecular pathways under different pathological conditions, including oxidative stress [69], inflammation, apoptosis, and cell death [70]. Melatonin has direct and indirect protective functions through the detoxification of free radicals, mitochondrial protection, reduction of free radical formation, prevention of neuronal overexcitation and several anti-inflammatory actions in neurodegenerative diseases, such as Alzheimer's disease, Huntington' disease, Parkinson's disease, amyotrophic lateral sclerosis, spinocerebellar ataxia, and multiple sclerosis [70]. Melatonin plays a regulatory role in MMPmediated physiological and pathological processes and human diseases. Melatonin can inhibit the tumorigenesis of bladder cancer and renal cell carcinoma metastasis by downregulating MMP-9 expression through different signaling pathways [71,72]. In our study, melatonin pretreatment of the IL-1β + DAPT/PDTC/SB-3CT + MEL groups inhibited the increase of MMP-9, reversed the MMP-9-induced damage of VE-cadherin, occludin, claudin-5, and ZO-1, and reduced the increase in BBB permeability to Na-F. In addition, melatonin promotes the expression of the MMP-9 endogenous inhibitor TIMP-1 and decreases the MMP-9/TIMP-1 ratio compared with that in the IL-1β-treated and IL-1β + DAPT/PDTC/SB-3CT groups. These results suggest melatonin could counter the damaging effects of MMP-9 by modulating the regulatory balance between MMP-9 and TIMP-1.
Many recent studies suggested that inflammatory factors (TNF-α or IL-1β) or medicines can inhibit MMP-9 expression via the NF-κB signaling pathway in different types of cells in different diseases [73][74][75][76]. In our study, when pericytes cultures were pretreated with the NF-κB inhibitor PDTC, the expression of MMP-9 decreased by PDTC indicating MMP-9 expression is regulated by the NF-κB signaling pathway. Activation of NF-κB by NOTCH3 led to T-cell tumorigenesis and enhanced NF-κB-dependent anti-apoptotic and proliferative pathways [77]. NOTCH3 down-regulation by siRNA decreased the expression of MMP-9 in the hepatocellular carcinoma [78]. When pericytes cells from the BBB in vitro model were pretreated with the NF-κB inhibitor PDTC, NOTCH3 expression was unaffected. However, pretreatment with the NOTCH3 inhibitor DAPT significantly inhibited the expression of NF-κB/p65 and translocation of NF-κB/p65 to the nucleus, which suggests that the NF-κB signaling pathway is regulated by NOTCH3. If the expression of NOTCH3 was significantly increased, it could also increase the expression of MMP-9, which damages the intact of BBB. This interpretation is consistent with our experimental results. How pericytes are causally involved in neuroinflammation will improve our understanding of how infection alters the bidirectional crosstalk that modulates brain functions in the neurovascular unit. The NOTCH3 research field is still in its infancy and we need more evidence of the impact of NOTCH3 in pericytes on the vascular integrity of the BBB. Moreover, we are particularly interested in further investigations into the role of NOTCH3 in pericytes and how it contributes to the integrity and function of the BBB in animal models or clinical studies.
In conclusion, the results from our study demonstrate an aggregation-dependent effect of IL-1β on the increased expression of NOTCH3 in pericytes, which can induce MMP-9 secretion from pericytes and disrupt BBB integrity. Moreover, our study highlighted MMP-9 as one of the mediators of IL-1β-induced alteration of NOTCH3. These results are consistent with the hypothesis that MMP-9 is involved in BBB pathogenesis through potential NOTCH3/NF-κB signaling pathways. We also suggest that pericytes are potential participants in the events of the BBB pathogenesis in CSVD.

Construction of the in vitro BBB model [79]
The schematic drawing of the in vitro BBB model is presented in Supplementary Figure 3.  Table 1 shows the primers used for amplification of the genes that were analyzed.

Evaluation of blood-brain barrier permeability
The flux of sodium fluorescein (Na-F) across the endothelial monolayers in a Transwell system (0.4 µm pore size polyester membrane inserts, Corning, Union City, CA, USA) was determined as previously described [50,79]. When all cells reached confluency, the BBB model cells were treated with IL-1β (10 ng/mL) with or without DAPT (10 μmol/L), or PDTC (25 μmol/L), or SB-3CT (20 μmol/L), and with or without melatonin (MEL, 10 μmol/L) in serum-free medium for 24 h. Briefly, 10 μg/mL Na-F (MW: 376 Da) was added to the upper inserts of the Transwell system for 0, 5, 15, 30, 60, 90, and 120 min, at which time 10 μL medium was aspirated from the lower compartments and replaced with the same volume of fresh serum-free medium. The fluorescence value that passed through the inserts was determined by fluorescence multiwall plate reader Synergy4 (BioTek, USA) at an excitation wavelength of 485 nm, and an emission wavelength of 535 nm.

Gelatin zymography
To assay the enzymatic activity of MMP-2 and MMP-9, conditioned media from the BBB models were analyzed by gelatin zymography [50,80,81]. Serum-free conditioned medias were subjected to 10% sodium dodecyl sulfate-polyacrylamide gels (SDS-PAGE) containing 1mg/mL gelatin (Sigma-Aldrich, St. Louis, MO, USA) in non-reducing conditions. Protein samples were mixed with 1% (w/v) SDS sample buffer in nonreducing conditions and loaded onto the gels and ran at 125 V until the Bromophenol Blue dye ran off the end of the gels. Gels were washed twice with 2.5% (v/v) Triton X-100 and subsequently incubated in developing buffer containing 50mmol/L Tris/HCl (pH 7.8), 0.2 mol/L NaCl, 5 mmol/L CaCl2 and 0.02% sodium azide for 42 h at 37°C. Gels were stained with 0.1% Coomassie Blue R-250 followed by destaining in different concentrations of methanol/acetic acid (30% Gelatinolytic activity appeared as transparent bands on a blue background and was quantified by scanning the bands using an automatic chemiluminescence image analysis system (Tanon-5200, Shanghai, China). The band intensities were quantified using the ImageJ program (https://rsb.info.nih.gov/ij).

AUTHOR CONTRIBUTIONS
W.Q. and J.L. planned and carried out experiments, analyzed data, and wrote the paper; R.Z., S.G., J.F., and M.X helped with experiments and analyzed data. R.Z. and J.Z. planned experiments and supervised all the work described in this paper. All authors approved the submitted and published versions. AGING

Supplementary Figures
Supplementary Figure 1. Pericytes were the primary producer of MMP-9, rather than astrocytes and endothelial cells with the IL-1β induction. (A) qRT-PCR analysis of the expression levels of MMP-9 in pericytes, astrocytes and endothelial cells from BBB model cells. The data, normalized to GAPDH expression, are presented as the mean ± SD, n=3. **** p < 0.0001 compared with pericytes. (B) Western blot analysis of the expression levels of MMP-9 in pericytes, astrocytes and endothelial cells from BBB model cells. (C) MMP-9 protein level was quantified by densitometry analysis using ImageJ. The data, normalized to β-actin expression, are presented as the mean ± SD, n=3. **** p < 0.0001 compared with pericytes.