The role of high-mobility group box protein 1 in chronic cerebral hypoperfusion-induced vascular cognitive impairment animals

Background: The molecular mechanisms of vascular cognitive impairment (VCI) are diverse and still in puzzle. VCI could be attributed to chronic cerebral hypoperfusion (CCH). CCH may cause a cascade of reactions involved in ischemia and neuroinflammation which plays important roles in the pathophysiology of VCI. High-mobility group box protein 1 (HMGB1) is a non-histone protein that serves as a damage-associated molecular signal leading to cascades of inflammation. Increased level of HMGB1 has been established in the acute phase of CCH. However, the role of HMGB1 at the chronic phase of CCH remains elucidated. Methods: We performed modified bilateral common carotid artery occlusion (BCCAO) in C57BL/6 mice to induce CCH. We examined the cerebral blood flow (CBF) reduction by laser doppler flowmetry, the protein expression of HMGB1 and its pro-inflammatory cytokines (TNF-a, IL-1b, and IL-6) by western blotting and immunohistochemistry. The brain pathology was assessed by 7T-animal MRI and amyloid-b accumulation was assessed by amyloid-PET scanning. We further evaluated the effect of HMGB1 suppression by injecting CRISPR/Cas9 knock-out plasmid intra-hippocampus bilaterally. Results: There were reduction of CBF up to 50% which persisted three months after CCH. The modified-BCCAO animals developed significant cognitive decline. The 7T-MRI image showed hippocampal atrophy, although the amyloid-PET showed no significant amyloid-beta accumulation. Increased protein levels of HMGB1, TNF-a and IL-1b were found three months after BCCAO. HMGB1 suppression by CRISPR/Cas9 knock-out plasmid restored the CBF, IL-1B, TNF-alpha, IL-6, and attenuated hippocampal atrophy and cognitive decline. Conclusion: HMGB1 plays a pivotal role in the pathophysiology of the animal model of CCH and it might be a candidate as therapeutic targets of VCI.

4 as it is responsible for the neuronal damage and the worsening symptoms. There are also several studies about HMGB1 induced chronic inflammation and neurodegeneration in Alzheimer's disease (AD), amyotrophic lateral sclerosis, Parkinson disease and multiple sclerosis. Previous study revealed that HMGB1 could impair neurit outgrowth in mice model of AD and led to cognitive decline (13). In human study, serum HMGB1 increased significantly in AD patients and correlated well with the level of amyloid-b (14). The HMGB1 is also believed to cause disruption of blood-brain-barrier integrity (15).
Nonetheless, the detailed mechanisms of HMGB1 in mediating cognitive decline at the chronic phase after CCH have not been established yet. Previous experimental model of CCH only demonstrated that the administration of anti-HMGB1 neutralizing antibody at acute stage of ischemia could reserve hippocampal neuronal death and improve cognitive impairment (16). Although the positive effect of that antibody remains even in 12 weeks, the persistence of high HMGB1 at the chronic phase of CCH has not been investigated. In addition, it is still not clear whether the suppression of HMGB1 in the chronic phase would also be beneficial. In this study, for the first time, we used modified bilateral common carotid artery occlusion to induce CCH in mice and suppressed the HMGB1 level by CRISPR/Cas9 knock-out plasmid during the chronic stage. Hence, we aimed to figure out the role of HMGB1 in the chronic stage of CCH and to explore its potential application on vascular cognitive impairment.

Animals and study design
Male C57bl/6j mice (16-week-old, weighing 25 to 35 grams, Bio-Lasco Taiwan Co., Ltd) were used for all the experiments. They were placed under controlled temperature (22 ± 1°C) and humidity (55 ± 10%) with a 12 h light/dark cycle (lights on at 07:00 h). Food and water were given as libitum to all mice throughout the experiments. Animal care and 5 experimental procedures in this study had been approved and were in accordance with the guidelines for the Care and Use of Laboratory Animals from Ethics Committee of Taipei Medical University.
The mice groups were divided into two parts: part one was before HMGB1 CRISPR/Cas9 KO plasmid injection, and part two was after the injection. For part one, the mice were randomly grouped as sham-control (n=12) and BCCAO (n=12). For part two, the mice were randomly grouped as sham-CRISPR control, sham-CRISPR HMGB1 KO, BCCAO-CRISPR control, and BCCAO-CRIPSR HMGB1 KO. Each group in part two had six mice (Fig. 1). This study followed the ARRIVE guidelines.

Model establishment of CCH by modified bilateral common carotid artery occlusion (BCCAO) surgery
The modified-BCCAO was performed by ligating both the common carotid artery (CCA) but with some modification from the previous procedure (17,18). Due to mice were dead soon after both of the CCA were ligated simultaneously, hence one of the CCA was ligated transiently for 30 minutes. The animals were anesthetized with isoflurane 2%. Cerebral blood flow (CBF) was measured while doing the surgery. A sagital midline incision (~1 cm length) was performed to expose the parietal skull. The skin was carefully dissected and a fiberoptic probe of laser doppler flowmetry (LDF) was put directly to the skull 2 mm caudal and 5 mm lateral from bregma to measure the CBF. A cervical midline neck incision (~1 cm length) was made. Both salivatory glands were carefully separated and mobilized to visualize the underlying common carotid artery (CCA). Both CCA were carefully separated from the respective vagal nerves and accompanying veins without harming these structures. A tight, double 5-0 silk suture loop (proximal & distal) was made around right 6 CCA. Re-measured the CBF, there must be a reduction of CBF by 80-90% from the baseline. Finally, we closed the wound by suturing.
One week later, the procedure was repeated for left CCA. The steps were similar except there was transient ligation of left CCA. A small polyethylene tubing (diameter 0,58 mm) was inserted in between the left CCA and the silk sutures. This tubing was used as a splinting of the left CCA in order to avoid damaging of the arterial walls when the sutures were tightened. Left CCA was occluded for 30 minutes by tightening the silk sutures.
Then, we re-measured the CBF and there must be a reduction of CBF by 80-90% from the baseline. After 30 minutes, released the ligation of left CCA and removed the polyethylene tubing. After the whole procedure, mice were put in a heating pad for 30 minutes while waiting them awake. After awaking, they were put back into their cages. The procedure of sham-surgery was similar with BCCAO excluding the ligation of both CCA. control group, we injected a vehicle with the same volume and procedure. The injection was conducted at 1 month after BCCAO/sham surgery. Two months after injection, the mice had an MRI examination to evaluate the brain pathology and hippocampal atrophy as well as a novel object recognition (NOR) test.

Novel object recognition (NOR)
Mice tend to interact more with a novel object than with a familiar one. This tendency has been used by behavioral pharmacologists and neuroscientists to study learning and memory. A popular protocol for such research is the object recognition task (19). The procedure consisted of three phases: habituation phase, sample phase, and test phase.
We perform this procedure with modifications (20). During habituation phase, each mouse was allowed to explore the field in the absence of objects for 10 min for two consecutive days in order to make them become familiar with the field. On the third day during the sample phase, two objects were placed symmetrically onto the arena. The mice were placed at the mid-point of the wall opposite the sample objects with its body parallel to the side walls and its nose pointing away from the objects, allowed to freely explore for 5 min. Time spent exploring the objects was recorded. During the test phase, one of the two objects used in sample phase was randomly replaced by a novel one, then the mice were re-introduced to the arena for 5 min exploration after a 4-hours delay. Between each mouse, any feces were cleared, the arena and objects were cleaned with 70% ethyl 8 alcohol. The video tracking system was used to collect behavioral performances automatically. The time spent exploring both the novel and the familiar objects was recorded (TN, TF). Object discrimination was evaluated by the recognition index (RI): RI = TN/(TN + TF).

Motor function test
Beam walking test: Further subtle motor coordination and balance were assessed using a modified balance beam (beam walking) test. This procedure was based on modified protocol describe in Luong et al (2011) (21). The beam apparatus consisted of 1-meter beams with a round-rough surface of 12 mm (or 6 mm width) resting 50 cm above the table top on two poles. A black box was placed at the end of the beam as the finish point.
Food was placed in the black box to attract the mouse to the finish point. A lamp (with 60watt light bulb) was used to shine light above the start point and serves as an aversive stimulus. The time to cross the center 80 cm was measured manually: one at 0 cm that started a timer and one at 80 cm that stopped the timer. A video camera was set on a tripod to record the performance. This test took place over 3 consecutive days: 2 days of training and 1 day of testing. On training days, each mouse crossed the 12 mm (or 6 mm) beam 3 times. On the test day, time to cross each beam were recorded. Two successful trials in which the mouse did not stall on the beam were averaged. Video recordings could be used for finer analysis of slipping and other observable motor deficits.
Rotarod: Briefly, gross motor control was measured using the rotarod (IITC Life Science, CA, USA). This procedure was based on the protocol described in Tung et al. (2014) with some modification (22). For this test, each mouse was placed on a cylindrical dowel (69.5 mm in diameter) raised 27 cm above the floor of a landing platform. Mice were placed on the dowels for 5 min to allow them to acclimatize to the test apparatus. Once initiated, 9 the cylindrical dowels began rotating and accelerated from 5 rpm to a final speed of 44 rpm over 60 s. During this time, mice were required to walk in a forward direction on the rotating dowels for as long as possible. When the mice were no longer able to walk on the rotating dowels, they fell onto the landing platform below. This triggered the end of the trial for an animal and the measurements of time to fall (TTF) were collected. Passive rotations where mice clung to, and consequently rotated with the dowel were also used to define the end of the trial. Mice were then returned to their cages with access to food and water for 10 min. This procedure took 3 days; day 1 and 2 were for training (each day consisted of 2-3 trials), and day 3 was for testing (consisted of 2 trials). The trials from testing day was used for analysis.

MRI (7T-aMRI)
MRI experiments to monitor the brain condition (e.g. atrophy and white matter lesion) after BCCAO were performed with a 7T horizontal MRI scanner (Bruker PharmaScan 70/16, USA), with a 7T/40cm magnet (Biospect Bruker console) and a surface coil. The mice were initially anesthetized with 3.0% isoflurane (Escain, Mylan Japan, Tokyo, Japan) and then with 1.5% to 2.0% isoflurane and 1:5 oxygen/room-air mixture during the MRI experiments. Rectal temperature was continuously monitored using an optical thermometer (FOT-M, FISO, Quebec, QC, Canada) and maintained at 37.0 °C ± 0.5 °C using a heating pad (Temperature control unit, Rapid Biomedical), and warm air was provided by a homemade automatic heating system based on an electric temperature controller (E5CN, Omron, Kyoto, Japan) throughout the MRI experiments. During MRI scanning, the mice laid in a prone position on an MRI-compatible cradle and were held in place by handmade ear bars. The first imaging slices were carefully set at the rhinal fissure, with reference to a mouse brain atlas. The modality of MRI performed was T2-weighted spin echo.

Statistical analysis
One-way ANOVA followed by Sidak post-hoc test was used to analyze western blot results, IF, and CBF measurement. NOR results were analyzed by using two-way ANOVA. Student unpaired t-test was used to analyze motor function MRI results. All analyses were performed by using GraphPad Prism 7 software (GraphPad Software, La Jolla, CA). The statistically significance was considered when the p value was less than 0.05.

Survival rate and CBF values
All procedures of modified-BCCAO were accomplished within 15 minutes, except for waiting time of 30 minutes to transiently ligate the left common carotid artery. After 3 months of BCCAO surgery, all mice had survived.
In the control group (sham-surgery), the mean CBF at 1 month and 3 months after the sham surgery did not change much from baseline. In contrast, the CBF values decreased significantly in the BCCAO group. The CBF values in right cerebral cortex decreased to 57.5% + 13.6% and 73.4% + 15.5% (mean + SD) in left cerebral cortex at 1 month after surgery (Fig. 2a). At 3 months after surgery, the CBF values remained low in BCCAO mice.

CCH induced hippocampal atrophy and memory decline, but did not cause behavioral alterations in motor coordination
To investigate the effects of CCH on motor coordination, we performed rotarod test and beam-balance walking test at 3 months after surgery (Fig. 2b, c). We found that BCCAO did not alter motor coordination, indicating there was no injury in cortical or subcortical lesion.
To study whether CCH impair memory function, Novel Object Recognition (NOR) test was 14 conducted. This test can be used to assess non-spatial working memory which describes some parts of hippocampus function and its relationship with perirhinal cortex (24) as well as frontal subcortical circuits (25). The BCCAO mice showed worse performance to recognize the familiar object than sham mice at 1 and 3 months after surgery (Fig. 2d, e; F (2, 28) = 8.815, p<0.01). Compare with sham-mice, BCCAO mice spent less time exploring the novel object when tested at 3 hours after training phase. Hence, the results of recognition index (RI) were <50%, indicates a short-term memory deficit in BCCAO mice due to impairment of memory retention for novel object or more preference to the familiar object (26,27) .
Hippocampal atrophy is associated with cognitive decline in dementia cases (28)(29)(30). To delineate this, hippocampal volumes were measured in sham and BCCAO mice at 1 and 3 months after surgery using T2-weighted image. Hippocampal atrophy was found during the time course after the surgery in BCCAO mice, but not in sham mice. BCCAO mice showed significant reduction of total hippocampal volume than sham mice at 1 and 3 months after surgery (Fig. 2f, g: p <0.05).
CCH is expected to cause amyloid-beta (Ab) accumulation particularly in Alzheimer's disease (31)(32)(33)(34). We also investigated whether CCH induced by BCCAO could develop Ab accumulation at 3 months after surgery by doing amyloid-PET scanning. However, we found no Ab accumulation in BCCAO mice (Fig. 3).

CCH increased the expression of HMGB1 at 3 months
CCH will induce neuroinflammation (30,35). However, there are plenty cascades of inflammation involved in CCH. We found that CCH induced by BCCAO altered the level of HMGB1 in cortex, striatum, and hippocampus at 3 months (Fig. 4). At 1 month after BCCAO surgery, we did not find any alteration of HMGB1 expression either in cortex, striatum or hippocampus (data not shown).
The expression of HMGB1 at chronic phase of CCH was further confirmed by immunostaining (Fig. 5). We found consistent findings that HMGB1 was expressed more in BCCAO mice. The pro-inflammatory cytokines such as TNF-a, and IL-1b were also upregulated. In addition, IL-6 was likely to increase in BCCAO mice compare to sham. All of them serve as the downstream signaling pathway of HMGB1.
The source of increased level of HMGB1 was from neuronal cells, mostly concentrated in the cytosol and some were released extracellularly. This was revealed by the triplestaining of HMGB1/NeuN/DAPI as shown in Fig. 6.

Administration of CRISPR/Cas9-KO of HMGB1 in BCCAO mice decreases the expression of HMGB1, restores CBF, attenuates hippocampal atrophy, and improves cognitive decline
To further delineate the role of HMGB1 at the chronic phase of CCH, we injected the HMGB1 CRISPR/Cas9 knock-out (KO) plasmid for suppressing the HMGB1 expression. The injection was done at 1 month after CCH. Two months after injection, the BCCAO HMGB1-KO mice showed a restored CBF, decreased level of HMGB1 from their brain tissue, attenuated the hippocampal atrophy, and improved the cognitive decline (Fig. 7). The decreased expression of HMGB1 was further confirmed by immunostaining as it also demonstrated decreased expression of pro-inflammatory cytokines (TNF-a, IL-1b, and IL-6) (Fig. 8).

Discussion
This is the first study to explore the role of HMGB1 at the chronic phase of chronic cerebral hypoperfusion (CCH) in a mouse model of VCI induced by modified-BCCAO. This is also the first study to use HMG-1 CRISPR/Cas9 KO plasmid to suppress HMGB1 in CCH. We found increased HMGB1 in 3 months after CCH together with reduced cerebral blood flow (CBF) ratio from baseline, hippocampal neuronal loss or atrophy and declined memory function. The administration of HMG-1 CRISPR/Cas9 KO plasmid at 1 month after surgery could reverse those consequences of CCH (Fig. 9).
Animal models establishment of CCH caused by common carotid artery occlusion have been verified by previous study using bilateral common carotid artery (BCCAO) in rats (36,37), bilateral carotid artery stenosis (BCAS) in mice (5,30,38,39) or gerbil (40), and unilateral common carotid artery occlusion (UCCAO) in mice (4,41). The rat model is the most widely used among the other models of CCH. However, it has limitation as causing damage in the visual pathway by occluding the ophthalmic artery thus may influence the behavioral assessment (42). It also has quite high mortality rates, ranging between 7%-25% depends on the strains used (43). The mice BCAS model also has higher mortality rate with microcoils of 0.16 mm in diameter (75% mortality rate) and variably mortality rates with microcoils of 0.18 mm in diameter (15-19%), 0.20 mm (17%), and 0.22 mm (13%) (44). The mice UCCAO model has no mortality but only induces mild short-term memory loss due to very mild reduction of CBF (42). The present study demonstrated that CCH induced by modified-BCCAO surgery caused moderate hypoperfusion at 1 and 3 months after the surgery. There was no mortality of animals. The CBF ratio was persistently reduced to 50-70% from baseline even after 3 months. Previous study used quantitative measurement by laser speckle imaging revealed that in bilateral carotid artery stenosis (BCAS) model, the CBF decreased to 62.9% at 2 hours after surgery and gradually restored to 81.7% at 1 month, 83.2% at 2 months, and 85.0% at 3 months( 5).
Another model of VCI, UCCAO (unilateral common carotid artery occlusion), has been showed that decreased of CBF measured by laser Doppler flowmetry to about 20-37% at 28 days after surgery (4,28). In this present study we show, for the first time to our knowledge, that the CBF ratio remains low at 3 months after surgery (5,6,45,46). The results in the present study might be caused by the procedure of BCCAO consists of 2 steps which allows less risk of death and persistently moderate reduction of CBF until 3 months after surgery.
CCH model in the present study significantly impairs the non-spatial memory function as showed by the lower recognition index (less than 0.5) in the NOR test. The similar findings were also found in other studies of CCH whether BCAS, UCCAO, or BCCAO was used in rats.
NOR test could demonstrate the essential function of hippocampus (especially the dorsal part) and perirhinal cortex (47,48). These two structures are responsible for non-spatial and short-term memory function (24,25). This result was further supported by hippocampal atrophy found from MRI image. Hippocampus is the most vulnerable region after global cerebral ischemia or CCH, especially CA1 neurons. Therefore, hippocampal atrophy could be found as the consequence of neuronal loss after CCH (6, 49-51). We did not observe any motoric disturbance at 3 months after surgery, indicates that CCH does not alter motor cortex or other structures associated with motoric function. This is in agreement with prior studies (4,33,52).
In the present study, we did not find any Ab accumulation. It suggests that the process of CCH only might not adequate to induce Ab accumulation in C57BL/6 mice. It needs other factors than CCH to initiate Ab deposition. Previous studies reported that Ab accumulation found in AD transgenic mice (at 5 months of age) after 1-month CCH induced by BCAS (33,53). In another study, 3 months old of AD transgenic mouse demonstrated Ab overproduction, decreased α-secretase activity and expression, as well as increased βsecretase activity and expression after 1-month CCH induced by UCCAO (54).
Nevertheless, a study using rat with BCCAO showed that CCH could promote Ab pathogenesis by activating β/γ-secretases (55); and a study using C57BL/6 mice reported that CCH increased Ab levels and enhanced β/γ-secretase levels after 6 months of BCAS (31). The latest findings suggest that CCH in wild type mice could possibly induce Ab accumulation or promote Ab pathogenesis only if CCH was applied in longer time period.
However, this hypothesis still needs further investigation.
Previous study has demonstrated the increased HMGB1 release at the acute phase of hypoperfusion induced subsequently persistent neuroinflammation (16). However, in the present study we found remarkable increased of HMGB1 at the chronic phase of CCH, in terms of 3 months. In the present study, the increased HMGB1 were found in cytosols of neuronal cells which could then be released extracellularly. The increased level of HMGB1 in the cytosol might be a response of dying cells or surrounding cell death due to hypoxia and ischemia caused by CCH which eventually leads to neuronal loss (3,56,57). It is speculated that increased HMGB1 level at this chronic phase should not be released from necrotic neurons. It is possibly released by apoptotic neurons since apoptosis is the main cause of neuronal death found at the chronic phase of CCH (6,58). In contrary, necrosis is the main findings at the acute phase of CCH (6). Since the role of HMGB1 in chronic stage of CCH has never been explored, the increased HMGB1 level in the present study could be a novel finding to support the contribution of HMGB1 at the chronic phase of CCH. It could also serve as another piece of evidence that HMGB1 is possibly released by apoptotic neurons under CCH.
Once HMGB1 is released extracellularly, it will bind to TLR2, TLR4, or RAGE receptors, then activating NF-KB pathway to initiate the production of pro-inflammatory cytokines such as IL-1b, IL-6, TNF-a (11,12). As confirmed by our findings, TNF-a and IL-1b levels were found higher in BCCAO compared to sham mice at 3 months after CCH. However, we did not find 19 any increased level of TLR2/4, RAGE, or protein involved in NF-KB signaling pathway (data not shown). These findings implicated that HMGB1 triggered the downstream inflammation pathway without changes of levels of its receptors. That might highlight the important role of HMGB1 in the pathophysiology of VCI induced by CCH. The administration of HMGB1 CRISPR/Cas9 KO plasmid could suppress the HMGB1 levels as well as the pro-inflammatory cytokines (TNF-a, IL-1b, and IL-6) and restore CBF. Moreover, the knock-out of HMGB1 successfully attenuated hippocampal atrophy and improved cognitive impairment at the chronic phase of CCH. The therapeutic effect of HMGB1 KO plasmid by CRIPRS/Cas9 system demonstrated in this study shows that the suppression of HMGB1 resulted in not only with the inhibition of neuro-inflammation but also increase of CBF. Nevertheless, the restoration of CBF after HMGB1 suppression is contrary to previous studies (16,59,60).
Noteworthy that previous studies investigated the effect of HMGB1 neutralization on CBF changes at the acute phase of ischemia. Therefore, the findings from previous studies might have a different scenario from the present study.
The mechanisms related to CBF restoration after HMGB1 suppression is unknown. It could probably involve the vascular smooth muscle cells (VSMC) and cerebrovascular endothelial cells (CVEC) migration and remodeling (61,62). Both VSMC and CVEC affected by nitric oxide (NO) are very important to maintain the vascular tone and regulate CBF. Another possible mechanism could be related to iNOS or eNOS/NO signaling pathway. Previous studies showed that inhibition of HMGB1 could reduce the expression of iNOS (63,64). As well known, inducible nitric oxide synthase (iNOS) could lead to NO overproduction which triggers peroxidative process and released of free radical species (65,66). It could induce endothelial dysfunction caused by impairment of eNOS/NO signaling pathway which subsequently decrease the CBF (65). However, the direct effects of HMGB1 on eNOS/NO signaling pathway which is the most important mediator for vessel dilatation are still 20 unclear which should be further investigated to delineate the exact mechanisms.
The present study has some limitations. First, we only investigated the effects of HMGB1 suppression at one single dose and one period of time, in terms of 2 months after CRISPR injection. This is a concept-proving study, the effects of this suppression by different doses, as well as given for different periods of time should be warrant if clinical application is considered in the future. Second, we did not investigate the role of HMGB1 in glial cells. Previous studies showed that HMGB1 could also be released rapidly after cerebral ischemia by astrocytes, or as late as 1 week after ischemia by microglia (19,67).
Moreover, it has been established that HMGB1 would activate glial cells once it released by the dying neurons (12,64). Therefore, investigating the role of HMGB1 in glial cells after CCH might be beneficial in the future studies. Next, the underlying mechanisms by which HMGB1 exerts its pro-inflammatory action after CCH is not totally clear in this study. Previous studies have proved that HMGB1 would bind to TLR2/TLR4 and/or RAGE which did not change in this study to initiate downstream inflammatory pathway (68, 69).
The molecular mechanisms of how HMGB1 affect CBF in this CCH model also needs further investigations. Moreover, we only performed diffusion tensor imaging (DTI) and failed to detect white matter tract lesions (data not shown). The previous studies demonstrated that CCH induced white matter lesions (5,30,41,(70)(71)(72), Investigations with the histological examination and/or electrophysiological measurement may be considered for the future study.

Conclusion
This study demonstrated for the first time that HMGB1 significantly contributes to cognitive impairment at the chronic phase of CCH, in part by modulating pro-inflammatory cytokines (TNF-a, IL-1b, and IL-6) and CBF changes. The administration of HMG-1 CRISPR/Cas9 KO plasmid could reverse the cognitive impairment and restore the CBF.
Thus, the suppression of HMGB1 expression might be a promising therapeutic target for vascular cognitive impairment due to CCH.

Acknowledgments
We thank PET-scan imaging facilities at Linkou Chang-Gung Memorial Hospital for providing the amyloid-PET scanning; and the technical staffs at Small Animal-based Magnetic Resonance Imaging Service, Taipei Medical University for their technological assistance.

Authors' Contributions
ANV and CJH conceived and designed the study; ANV and HJY performed the experiments and analyzed the data; KJL performed PET scan imaging and analyzed the data imaging; ANV and CJH wrote the manuscript; YCF, IS and CJH contributed in critically reviewing the 22 manuscript.

Funding
This study was supported by Taiwan Ministry of Science and Technology (MOST 106-2314-B-038-038-MY2).

Availability of data and materials
The datasets used and/or analyzed in the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
All experiments were performed following the guidelines from the Care and Use of Laboratory Animals and approved by the Ethics Committee of Taipei Medical University (Approval numbers: LAC-2016-0434). All efforts were made to minimize the suffering of the animals and reduce the number of animals used.

Consent for publication
Not applicable    The triple-staining of HMGB1 with NeuN and DAPI from cortex and hippocampus of CCH mice. The source of increased HMGB1 were from neuronal cells, mainly located in the cytosols, and some were released extracellularly; scale bar= 50 m  Summary of the study (CCH=chronic cerebral hypoperfusion, CBF=cerebral blood flow; KO=knock-out)