Mitochondrial carbonic anhydrase–VB inhibition rescues brain endothelial stress and memory in Alzheimer’s disease models Authors: Nicole L. Lemon, Elisa Canepa, Rafael Vázquez-Torres, Rebecca Parodi-Rullán, Tina Hirt, Lisa M. Petersohn, Tashnuva Rifat, Maryam H. Abyaneh, Marc A. Ilies, Silvia Fossati Published: 2026 ABSTRACT Alzheimer’s disease (AD) is a devastating neurodegenerative disorder with no effective cure, characterized by the cerebral parenchymal and vascular accumulation of aggregated amyloid-β (Aβ) and hyperphosphorylated tau. Cerebrovascular and mitochondria dysfunction are early causal events in the progression of AD. Previous studies support that inhibiting carbonic anhydrases (CAs) may prevent mitochondrial and cerebrovascular dysfunction in AD models. Here, we selectively target the mitochondrial CA isoform CA-VB by pharmacological and genetic manipulation in human cerebral microvascular endothelial cells (hCMEC), and we confirm the protective effects of the CA-V inhibitor in AD mice. CA-V inhibition and CA-VB KO prevent Aβ induced mitochondria-mediated endothelial apoptosis, loss of barrier resistance, and hCMEC inflammatory activation. Strikingly, CA-V inhibition also mitigates caspase activation and endothelial cell activation in the brains of 3xTg AD mice, resulting in preserved memory function. Our results demonstrate that selective CA-V inhibition is an effective and promising strategy against AD-mediated cerebrovascular pathology, neuroinflammation and cognitive impairment. INTRODUCTION Alzheimer’s disease (AD), the most common form of dementia, is a devastating neurodegenerative disease, currently afflicting over 55 million people worldwide (1). The most common risk factor for AD and other forms of dementia is age (1). AD is characterized by the accumulation of extracellular amyloid-beta (Aβ) aggregates (or plaques) and intraneuronal hyperphosphorylated tau tangles in multiple brain areas, including the hippocampus and cortex (2). Recently, therapies employing antibodies targeting Aβ to facilitate its clearance have been approved by the Food and Drug Administration (FDA) for the treatment of AD. However, because of their concerning side effects, such as amyloid related imaging abnormalities (ARIAs) (3–6), and only partial efficacy in decreasing cognitive impairment, there is still no effective cure for AD. As the aging population is growing rapidly, it is evident that novel therapies against this devastating disorder are vital. The demise of the cerebral vessels, with decreased cerebral blood flow (CBF) and increased blood-brain barrier (BBB) permeability, is one of the earliest causal events in the pathogenesis of AD, which in turn affect other cells composing the neurovascular unit (NVU), such as glial and neuronal cells (6–10). The NVU is a multicellular structure composed of cerebral endothelial cells (CECs), pericytes/smooth muscle cells, astrocytes, microglia, and neurons, which all work together to execute many vital functions, including maintenance of the physical barrier between the blood and the brain, nutrient exchange, regulation of CBF, and immunological surveillance (2, 11). Cerebral amyloid angiopathy (CAA), the accumulation of aggregated Aβ around and within the brain vasculature, is observed in about 90% of patients with AD (12, 13). Parallel to NVU dysfunction is the rise of neuroinflammatory processes, which exacerbate a vicious cycle of microvascular stress, BBB permeability, neuroinflammation and glial cell activation, and eventually neurodegeneration (14). Aβ deposition at the cerebral vessels has been observed to both cause and worsen NVU dysfunction by affecting mechanisms such as mitochondrial metabolism, oxidative stress, cell death, and neuroinflammation (10, 14). Many studies in mice and humans have confirmed the presence of vascular abnormalities, both Aβ-dependent and accelerating Aβ deposition, very early in the disease (6, 8, 15, 16). Single-cell transcriptomics in human AD brains has recently established an up-regulation of apoptotic pathways in microvascular CECs (17). Thus, improving endothelial and cerebrovascular function is a critical target for AD therapies. In AD brains, Aβ40 is known to accumulate more on the vasculature, while Aβ42 deposits more in parenchymal plaques. Although most cases of CAA are sporadic, there are known familial mutations, such as the Dutch mutation, characterized by Aβ40 mutated in position 22 (AβQ22). The Dutch mutation, among others within the amyloid precursor protein (APP), clinically presents with severe Aβ deposition at the brain vasculature, causing early onset dementia, CAA, strokes, and cerebral hemorrhages (18). Although wild-type (WT) and Dutch Aβ aggregates induce similar CEC mitochondria-mediated apoptotic mechanisms and BBB abnormalities, as previously shown (19, 20), AβQ22 has a faster aggregation rate compared to WT-Aβ40 (20–22), correlating with its higher toxicity (19), and therefore, it is a useful tool to model CAA toxic effects in cell studies such as this. Our laboratory has demonstrated that the FDA approved nonselective, pan-carbonic anhydrase (CA) inhibitors, Acetazolamide (ATZ) and Methazolamide (MTZ) are effective in ameliorating mitochondria-mediated cell death and stress mechanisms in models of amyloidosis (23–26). In the TgSwDI model of CAA, ATZ and MTZ preserved vascular fitness and prevented gliosis (25). There is also evidence that CA inhibitors are protective in models of stroke (2), obesity (27), and diabetes (28). ATZ and MTZ prevent Aβ induced cell death, through both the extrinsic and intrinsic apoptotic pathways, in many cell types, including CEC, smooth muscle cells, glial cells, and neurons, in addition to diminishing mitochondrial oxidative stress (23, 24, 26). The intrinsic apoptotic pathway, also referred to as mitochondria-mediated apoptosis, is regulated through signaling molecules such as BIM, BAX and BAK, and BCL-2. When BAX and BAK translocate to the outer mitochondrial membrane they form pores, increasing mitochondrial membrane permeability and promoting loss of mitochondrial membrane potential. Mitochondria then release cytochrome C (CytC) into the cytosol, leading to the activation of caspase-9, which activates caspase-3, resulting in the execution of apoptosis, through DNA fragmentation. CEC are the physical barrier responsible for the inhibition of entry of peripheral molecules, immune cells, or foreign substances into the healthy brain. When needed, they have a direct role in the recruitment of peripheral immune cells to the brain, as well as in the activation of glial cells within the brain during inflammation. Tight junction proteins (TJPs), such as occludin and claudin-5, are important regulators of BBB integrity. Although there are mixed reports on the expression of TJPs in models of AD, it is clear that the BBB becomes leaky due to AD and CAA pathology. ECs can become activated through the Aβ-mediated accumulation of reactive oxygen species (ROS), or other mechanisms (14). When EC are activated, they release pro-inflammatory cytokines and increase the expression of immune cell adhesion proteins such as vascular cellular adhesion molecule-1 (VCAM-1), intercellular adhesion molecule-1 (ICAM-1) and E-selectin, which promote the extravasation of immune cells form the periphery into the brain. When this immune response becomes chronic, it is detrimental to the brain tissue (14, 29). CAs are zinc metalloenzymes that catalyze the hydration of carbon dioxide to produce bicarbonate and a proton. This reaction is vital in many processes, such as the regulation of pH, ion, and redox homeostasis, as well as metabolism (2, 30, 31). There are 15 different CA isoforms expressed by humans, each one with its’ own tissue expression pattern, activity level, and cellular localization (2, 31, 32), highlighting their different functions. CA isoforms are expressed on the plasma membrane of cells, in the cytosol, and two are expressed in the mitochondria (CA-VA and CA-VB). The two mitochondrial isoforms have a differential expression pattern, CA-VB being more abundant in the brain (2, 31, 33). We have recently observed that the expression of the mitochondrial isoform CA-VB is increased in CEC treated with Aβ, as well as in the brain of the TgSwDI mouse model of CAA and in human AD and AD + CAA brains, emphasizing the potential role of CA-VB in the development of AD pathology (25), specifically through modulation of cerebrovascular function. The nonselective CA inhibitors ATZ and MTZ are FDA approved for other indications, such as glaucoma and high-altitude sickness. Although pan-CA inhibitors are demonstrated to be protective in vitro and in vivo, chronic long-term use is a concern due to their lack of selectivity and potential undesired effects (2, 34). ATZ and MTZ belong to the sulfonamide drug class, with established structures, which have been modified and screened to determine their inhibitory activity on multiple CA isoforms (32). In a study evaluating the activity of benzene sulfonamide CA inhibitors, it was determined that the compound 4-phenylacetamidomethyl-benzenesulfonamide (4ITP) has strong selectivity for CA-V (35, 36). 4ITP is considered highly selective for CA-VA and CA-VB, with a Ki value of 8.6 and 8.3 nM, respectively. This is about 10-fold more effective than ATZ and MTZ. 4ITP was determined to be much less effective on cytosolic isoforms CA-I and CA-II and even less on plasma membrane isoforms CA-IX and CA-XII (35). Our previous findings demonstrate that pan-CA inhibitors prevent CEC death through mitochondria-specific mechanisms and that CA-VB is increased in AD models and human AD brains. Hence, we hypothesized that the selective inhibition of mitochondrial CA-V would protect from Aβ induced CEC mitochondria-mediated apoptosis and NVU stress, therefore reducing cognitive impairment while also decreasing the off-target undesired effects of pan-CAIs. Here, we aimed to investigate this hypothesis in CECs and the 3xTG mouse model of AD (37), which presents both Aβ and tau pathology, in addition to vascular dysfunction (38). This study establishes that a known selective CA-V inhibitor, 4ITP, is effective at reducing Aβ-induced mitochondria-mediated CEC death, BBB dysfunction, and cerebrovascular stress in vitro. We also confirmed the specific role of CA-VB in Aβ-mediated CEC death and BBB dysfunction, through CRISPR-Cas9 CA-VB knockout (KO) in human CEC. Moreover, we pioneered the employment of this CA-V inhibitor in an in vivo AD model, showing that it can prevent endothelial cell loss and BBB dysfunction, mitigate neuroinflammation, and ameliorate cognitive impairment in 3xTG mice, therefore shedding light on CA-V as a potential specific target for AD therapy. RESULTS CA-V inhibition reduces mitochondria-mediated Aβ-induced apoptosis in CECs To confirm that selective inhibition of mitochondrial CA-V prevents CEC death and mitochondria-mediated apoptotic mechanisms (23–26, 34), we treated hCMEC with 25 μM AβQ22 alone or in combination with the CA-V inhibitor 4ITP (1, 10, and 30 μM) for 24 hours (Fig. 1) or 16 hours (Fig. 2). These time points were selected on the basis of previous evidence demonstrating AβQ22’s effects on mitochondria-mediated cell death within hCMEC (24). After confirming that AβQ22 induced apoptosis, measured as DNA fragmentation (the last step of the apoptotic pathway), we determined that 10 and 30 μM 4ITP significantly protected hCMEC from AβQ22-induced apoptosis (Fig. 1A), nearly rescuing DNA fragmentation, as well as CEC loss of viability (Fig. 1B), to the control levels. Since both concentrations (10 and 30 μM) of 4ITP were nontoxic to hCMEC (fig. S1) and rescued both AβQ22-induced hCMEC DNA fragmentation (Fig. 1A) and viability loss (Fig. 1B), these concentrations were selected for subsequent experiments. To test whether CA-V inhibition could prevent Aβ-induced mitochondria-mediated apoptosis, we measured AβQ22-induced caspase-9 activation, which is directly promoted by mitochondrial CytC release. We demonstrated that 10 and 30 μM 4ITP were protective against AβQ22-induced caspase-9 activation (Fig. 1C), assessed as increased formation of its cleaved fragment by WB. CA-V inhibition also significantly reduced the percentage of active executioner caspases-3– and -7–positive cells (Fig. 1D), measured by the specific fluorescent marker Cell-Event caspase-3/-7 (Invitrogen). These results suggest that selective CA-V inhibition with 10 and 30 μM 4ITP has a protective effect against AβQ22-induced mitochondria-mediated CEC apoptosis. CA-V inhibition protects against Aβ40-Q22-induced apoptosis in hCMEC. Aβ40-Q22 (Q22; 25 μM) with or without 4ITP (1, 10, or 30 μM) for 24-hour treatment. (A) Measurement of DNA fragmentation in human CMEC, expressed in fold of change (FOC) of untreated control (Ctrl). (B) Cell viability triggered by 24-hour Q22 treatment, expressed as % of Ctrl. (C) WB analysis of CMEC lysates to measure cleaved (activated) caspase-9 (cCas9); cCas9 normalized to actin on the right. (D) Representative IF images of hCMEC, challenged with Q22 alone or in combination with 10 and 30 μM 4ITP. Active caspase-3/-7–positive cells (green) induced by Q22, as plotted on the right. Original magnification: ×10. These data represent the combination of at least three experiments, each with two replicates, four images per replicate, graphed as means + SEM. Statistical significance was evaluated by one-way analysis of variance (ANOVA), followed by Tukey post hoc test. ****P < 0.0001, ***P < 0.001, **P < 0.01, and *P < 0.05. Created in BioRender. Fossati, S. (2026; https://BioRender.com/zg3f90e). 4ITP protects against Aβ40-Q22 induced mitochondrial apoptotic mechanisms and oxidative stress in hCMEC. HCMEC treated for 16 h with 25 μM Aβ40-Q22 (Q22), in the presence or absence of 10 or 30 μM 4ITP. (A) IF images depict the loss of Mito-tracker signal (in red) observed in CMEC challenged with Q22 indicating a loss of mitochondrial membrane potential. The release of CytC (green) is visualized as a diffused, cytosolic signal (indicated with arrows) in Q22-treated cells compared to untreated cells (ctrl). Original magnification: ×100. (B) Representative WB images of mitochondrial apoptotic mediators BCL-2, BIM, BAX, and BAK, following Q22 treatment, and relative quantification below (loading control, actin). (C) Mitochondrial H2O2 production, measured by Amplex Red. (D) 4-HNE enzyme-linked immunosorbent assay (ELISA), standard curve quantification. Data represent the combination of at least three experiments each with two replicates, graphed as means + SEM. Statistical significance was evaluated by one-way ANOVA, followed by Tukey post hoc test. ***P < 0.001, **P < 0.01, and *P < 0.05. Created in BioRender. Fossati, S. (2026) ( https://BioRender.com/zg3f90e). Cotreatment with 10 μM 4ITP also prevented AβQ22-induced loss of mitochondrial membrane potential (ΔΨ), measured by the Mito-Tracker CMH2xROS dye, which enters only mitochondria with a healthy ΔΨ (Fig. 2A). The loss of mitochondrial membrane potential was observed in AβQ22-treated CEC as a decreased Mito-Tracker CMH2xROS signal. The release of CytC, measured via immunofluorescence (IF), was evident by a diffused signal in AβQ22-treated cells, indicating CytC release into the cytosol. Both loss of ΔΨ and release of CytC were rescued by 4ITP. Because of these observations, we sought to understand which proapoptotic mediators could drive the mitochondrial membrane permeability loss and if 4ITP could affect their expression. Many proteins may modulate the pore formation on the outer mitochondrial membrane, including proapoptotic BAX, BAK and Bim, and antiapoptotic protein BCL-2 (39, 40). We observed that the inhibition of CA-V with 4ITP significantly prevented AβQ22-mediated BAX and Bim increase in hCMEC (Fig. 2B), while there were no significant changes in Bcl-2 or proapoptotic protein BAK expression. One mechanism that can alter the expression and interactions of mitochondrial pore regulating proteins, as well as affecting ΔΨ, is the accumulation of ROS (34, 41). After 16-hour treatment, AβQ22 significantly induced production of H2O2 by isolated mitochondria compared to untreated hCMEC. As hypothesized, mitochondrial H2O2 production was reduced to levels similar to the control by both concentrations of 4ITP (Fig. 2C). When H2O2 is produced within the cell, it can react with lipids, forming toxic lipid peroxidation products such as 4-HNE, a known inducer of cell death and marker of oxidative stress, which has also been observed in the brains of patients with AD (42). We measured 4-HNE in CEC treated with AβQ22 and 4ITP. The increase in 4-HNE induced by AβQ22 was significantly prevented by 30 μM 4ITP (Fig. 2D). Overall, these results confirm our hypothesis that the inhibition of mitochondrial CA-V prevents Q22-induced mitochondrial ROS production and mitochondria-mediated apoptotic mechanisms. CA-V inhibition prevents Aβ-driven loss of endothelial barrier integrity After confirming the positive effects of the CA-V inhibitor on mitochondria-mediated apoptosis and oxidative stress triggered by AβQ22 in hCMEC, we aimed to determine whether CA-V inhibition could also prevent Aβ-mediated loss of endothelial barrier function. To do so, we used a sublethal concentration (10 μM) of AβQ22 to observe changes in vascular barrier integrity not due to cell death (20). The transendothelial electrical resistance (R) of the CEC barrier was measured through the ECIS Zθ technology, as an indicator of BBB integrity and health, which is decreased in the presence of different Aβ species (20, 21). Cotreatment with 10 and 30 μM 4ITP protected against AβQ22-induced loss in R overtime, improving R to levels higher than the control in a dose-dependent manner (Fig. 3A). We also observed that 4ITP alone increased the R compared to untreated control CEC. 4ITP prevents Aβ40-Q22 induced loss in endothelial barrier integrity. (A) hCMEC barrier resistance was assessed by ECIS Zθ to evaluate barrier integrity. Human CMEC were treated with 10 μM Aβ40-Q22 (Q22), alone, or in combination with 10 or 30 μM 4ITP. (B and C) Western blot analysis of occludin and oligomeric claudin-5 (normalized to actin and plotted as % change of Ctrl), following 24-hour (B) and 48-hour (C) treatment. Data represents three individual experiments with two replicates per group, graphed as means ± SEM. Statistical significance was evaluated by one-way ANOVA, followed by Tukey post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Created in BioRender. Fossati, S. (2026) (https://BioRender.com/zg3f90e). To further investigate whether these changes in BBB resistance are associated with modifications in TJP expression, we assessed if 4ITP modulated AβQ22 effects on the expression of TJPs occludin and claudin-5 by WB. At 6 hours, AβQ22 did not significantly affect the expression of occludin and claudin-5 (43) (fig. S2), as it did not at 24 hours (Fig. 3B). However, 4ITP treatment, alone or in combination with AβQ22, significantly increased claudin-5 expression after 6 hours (fig. S2) or 24 hours (Fig. 3B), in line with the increased R of CEC monolayers treated with the CA-V inhibitor. After 48 hours of treatment with 10 μM AβQ22, we observed a reduction in the expression of occludin compared to untreated CEC, which was no longer significant in the presence of 4ITP, confirming its protective effects on barrier properties. We also observed a significant increase in occludin expression in cells treated with 30 μM 4ITP alone compared to control CEC (Fig. 3C). Overall, these results indicate that the CA-V inhibitor 4ITP increases tight junction expression and prevents AβQ22-induced loss of barrier resistance, confirming its therapeutical potential against BBB permeability in cerebrovascular disease. CA-V inhibition prevents Aβ-initiated endothelial activation Another important function of hCMEC is their role as mediators of the cross-talk between the peripheral and central nervous system (CNS) immune systems (14). Because of their bidirectional influence on cerebral and peripheral immune mechanisms, CEC activation state and the production of endothelial inflammatory mediators are critical in AD and other cerebrovascular diseases. To evaluate endothelial activation, we measured the expression of vascular adhesion molecules, known to extravasate peripheral immune cells into the brain, VCAM-1, and ICAM-1, following 6, 24, and 48 hours of treatment. The AβQ22 increased the expression of both VCAM-1 and ICAM-1 at different time points: At early time points (6 hours; Fig. 4A), ICAM-1 was significantly increased compared to the control, while there were no significant changes in VCAM-1 expression; at later time points (24 and 48 hours; Fig. 4, B and C, respectively), Q22 treatment significantly increased VCAM-1 protein expression compared to the control hCMEC, while ICAM-1 expression was not significantly changed. The cotreatment with 4ITP reduced the AβQ22-mediated increase in ICAM-1 to levels nonsignificantly different from the control at 6 hours and significantly restored VCAM-1 levels at both 24 and 48 hours, in a dose-dependent manner. 4ITP mitigates Aβ40-Q22 induced brain endothelial activation. Human CMEC were treated with 10 μM Aβ40-Q22 (Q22), alone or in combination with 10 or 30 μM of 4ITP. Western blot analysis of VCAM-1 and ICAM-1 (normalized to actin, and plotted as % change of Ctrl), at (A) 6 hours , (B) 24 hours , and (C) 48 hours following treatment. (D and E) Cytokine release was evaluated after 6 and 48 hours Q22 treatment using a multiplex MSD proinflammatory panel. Data represent three individual experiments with two replicates per group, graphed as means ± SEM. Statistical significance was evaluated by one-way ANOVA followed by Tukey’s [(A) to (C)] or Dunnett’s T3 [(D) and (E)] post hoc test. *P < 0.05, **P < 0.01, and ***P < 0.001. Created in BioRender. Fossati, S. (2026) (https://BioRender.com/zg3f90e). FOC, fold of change. We also aimed to correlate the CEC inflammatory mediators released into the conditioned media (CM). After 6 hours of AβQ22 treatment (Fig. 4D), we observed a significant decrease in the release of anti-inflammatory interleukin-10 (IL-10), which was reverted by 4ITP. Conversely, we observed a significant increase in the release of pro-inflammatory cytokine IL-6, which was not affected by mitochondrial CA-V inhibition. There were no significant changes in IL-4 or IL-12p70 after 6 hours of treatment. After 48 hours (Fig. 4E), AβQ22 treatment lowered IL-10 levels (P = 0.0585) and increased the release of pro-inflammatory cytokines IL-6 and IL-12p70 and pleiotropic cytokine IL-4 compared to untreated hCMEC. The 4ITP significantly up-regulated anti-inflammatory IL-10 release compared to Q22 alone (significantly at 10 μM) and diminished IL-6, IL-4 (significantly at 30 μM), and IL12p70 levels. To obtain a broader picture of amyloid-induced EC activation, we measured mRNA expression of a number of relevant immune cell mediators and cytokines through a customized Taq-man quantitative polymerase chain reaction (qPCR) array after 24 hours treatment. Confirming the above WB data, we observed a significant Aβ-mediated increase in VCAM-1 mRNA expression that was significantly reduced by 4ITP treatment (fig. S3). We also observed a significant increase in CXCL2, CSF3/granulocyte colony-stimulating factor (G-CSF), IL-12a/b, IL-4, IL-6, and E-selectin in cells treated with AβQ22. CA-V inhibition significantly reduced amyloid-induced increases in CXCL2, CSF3, IL-12b, IL-4, and E-selectin. Oppositely, CA-V inhibition significantly enhanced mRNA expression of anti-inflammatory cytokine IL-10 in the presence of amyloid. Further, we also observed a significant decrease in CXCL1, CSF2/granulocyte-macrophage CSF (GM-CSF), and P-selectin with 4ITP treatment in the presence of amyloid (fig. S3). These results suggest that CA-V inhibition ameliorates endothelial inflammatory activation at the transcriptional level, highlighting the therapeutical potential of its inhibition in AD and other cerebrovascular indications. CA-VB KO prevents Aβ-induced CEC apoptosis and loss in barrier resistance We recently reported that the expression of mitochondrial CA-VB is up-regulated in multiple models of amyloidosis, including hCMEC challenged with AβQ22, TgSwDI mice, and human AD/CAA brains (25). We have also demonstrated that silencing CA-VB in hCMEC is protective against Aβ-induced apoptosis. In contrast, the expression of CA-VA, the other mitochondrial isoform, or CA-II, the main cytoplasmic CA present in the CNS, did not change. Because of the mounting evidence that CA-VB is likely the mitochondrial isoform mediating Aβ toxicity in AD and to further explore the role of this CA isoform in CEC dysfunction, we used hCMEC CA-VB CRISPR-CAS9 KO cells. After confirming almost complete KO of CA-VB (Fig. 5A) compared to control cells (electroporated with CAS9 alone), we also confirmed that other isoforms of CA, such as CA-VA and CA-II, were still expressed in these cells (fig. S4). Strikingly, after 24 hours of treatment with 25 μM Q22, CA-VB KO hCMEC did not show the significant increase in apoptosis, measured as DNA fragmentation, seen in the WT cells (Fig. 5B), suggesting that CA-VB is vital for AβQ22-induced apoptosis in CEC. To further investigate the impact of mitochondrial CA-VB on endothelial apoptosis, we also measured Q22-induced caspase-3/-7 activation in CA-VB KO cells (Fig. 5C). We found that AβQ22-induced caspase-3/-7 activation was completely prevented in CA-VB KO CEC. In addition, IF analysis demonstrated that, after 16 hours of AβQ22 treatment, AβQ22-mediated loss of mitochondrial membrane potential and CytC release were prevented in CA-VB KO CEC (Fig. 5D). Notably, we also observed that CA-VB KO protected from AβQ22-induced mitochondrial H2O2 production (Fig. 5E). Last, to determine whether CA-VB KO protected against barrier integrity loss, we measured endothelial barrier R overtime with ECIS ZΘ. As expected, 10 μM AβQ22 triggered loss of barrier R in WT CEC, while in the absence of CA-VB, barrier R in the presence of AβQ22 challenge was significantly preserved. A remaining slight loss of R in AβQ22-challenged CA-VB KO CEC monolayers was observed, which may be due to an incomplete KO of CA-VB, or to the involvement of other players, such as CA-VA, which is also inhibited by 4ITP, in Aβ-mediated endothelial barrier dysfunction (Fig. 5F). CA-VB KO in hCMEC protects from Aβ-induced apoptosis and BBB permeability. (A) WB confirmed CA-VB absence in KO cells. (B) DNA fragmentation, plotted as FOC of untreated control (ctrl) cells, measured by cell death ELISA. (C) Representative IF images of active caspase-3/-7 (green) and quantification to the right in CMEC treated with 25 μM Q22 for 24 hours. Original magnification: ×10. (D) IF of CA-VB KO versus WT hCMEC to detect mitochondrial membrane potential (Mito-Tracker, red) and CytC (green) in hCMEC challenged with Q22 for 16 hours. Zoom images of the merge signal at the bottom depict altered mitochondrial network with perinuclear mitochondria in Q22-treated hCMEC, but not in CA-VB KO cells. Original magnification: ×100. (E) Mitochondrial H2O2 measured with Amplex Red kit. (F) Barrier resistance was measured over time with ECIS-Zθ in WT and CA-VB KO hCMEC treated or not with 10 μM Q22. Data represents three individual experiments with two replicates each, graphed as means ± SEM. Statistical significance was evaluated by two-way ANOVA, followed by Tukey post hoc test. *P < 0.05, **P < 0.01, and ****P < 0.0001. Created in BioRender. Fossati, S. (2026) (https://BioRender.com/zg3f90e). Because of the complete rescue of amyloid-induced loss in resistance in WT cells treated with 4ITP and incomplete rescue in CA-VB KO cells, we tested whether 4ITP could improve barrier R also in the CA-VB KO cells (fig. S5). We observed that in CA-VB KO hCMEC treated with 4ITP, there was still an increase in R, although less than in WT cells. We observed that 4ITP did improve the resistance back to control levels in CA-VB KO cells treated with amyloid. The effect of 4ITP on CA-VB KO cells is expected, as this compound inhibits both mitochondrial isoforms, CA-VA and CA-VB. Overall, these results indicate that CA-VB is necessary for Aβ-induced mitochondrial apoptosis and that its KO significantly ameliorates barrier integrity in CEC monolayers. CA-V inhibition in vivo decreases apoptosis and loss of BBB integrity, preventing neuronal loss and cognitive decline To evaluate the impact of CA-V inhibition in vivo, we determined the effects of 4ITP in a mouse model of AD, the 3xTG mouse model, which presents with both Aβ and tau pathology in the hippocampus and cortex. It has been recently demonstrated that these mice also display cerebrovascular deficits and mitochondrial dysfunction (38, 44–47). Male and female mice (WT and 3xTG) were treated with 4ITP (20 mg/kg/day; delivered through their diet) starting at 6 months, with an endpoint of 16 months. At 6 months of age the 3xTG mouse model is beginning to experience cognitive decline, representing early stages of MCI in humans (48). The CA-V inhibitor was delivered in their diet. The body weight of treated and nontreated mice was checked weekly, and mice were observed daily to evaluate toxicity throughout the study. First, we observed that 4ITP treatment in 3xTG mice lessoned overall brain apoptosis, measured as cleaved caspase-3 expression, in both the hippocampus and cortex through immunohistochemistry and WB analysis (fig. S6, A to C). Further, we observed a significant increase in DAPI+ (4′,6-diamidino-2-phenylindole–positive) DNA fragmentation in 3xTG compared to WT mice, which was decreased in 4ITP-treated mice (fig. S6, D and E). We wanted to understand the effects of 4ITP on brain endothelial cell apoptotic mechanisms in vivo. Through costaining of active caspase-3 and CD31, we observed that in the hippocampus of 3xTG mice, the % of capase-3–positive vessels was significantly increased compared to WT, while the difference from WT was not significant in the presence of 4ITP treatment (Fig. 6A). We observed a similar trend in the cortex, although not significant (fig. S7A). We also observed a significant reduction of the % area of CD31 in the hippocampus and a similar trend in the cortex of 3xTG mice, suggesting endothelial cell loss. This CEC loss was prevented by 4ITP treatment (Fig. 6, B and C). In addition, we uncovered a significant decrease in blood vessel diameter in the hippocampus and cortex of 3xTG mice compared to WT. This vasoconstriction was also significantly reverted in the presence of 4ITP treatment (Fig. 6D). Overall, these data indicates a preservation in microvessel quantity and function in the presence of the CA-V inhibitor. 4ITP protects from endothelial cell apoptosis, preserves vascular and BBB integrity in 3xTG mice. (A) Staining of cCASP3 and CD31 colocalization in the hippocampus of WT, 3xTG, and 3xTG mice treated with 4ITP (as in Materials and Methods). CD31 is represented in red and cCASP3 is represented in green. The % area of colocalized cCAS3 and CD31 was quantified with HALO imaging software. Original magnification: ×60. Quantification on the right. N = 4 to 7 mice per group, mixed male and female. Data represents 5 to 10 images per mouse and area. Statistical significance was evaluated by one-way ANOVA, followed by Tukey post hoc test. *P < 0.05 and **P < 0.01. (B) Images (20×) of WT, 3xTG, and 3xTG + 4ITP cerebral vessels stained with CD31 in red and DAPI in blue. (C) Quantification of % area of CD31. (D) Median blood vessel diameter of WT, untreated 3xTG mice, and 3xTG mice treated with 4ITP. [(C) and (D)] N = 5 to 6 mice per group, mixed male and female. Data represents 5 to 10 images per mouse and area. Statistical significance was evaluated by one-way ANOVA, followed by Tukey post hoc test. *P < 0.05 and **P < 0.01 (E and F) Western blot analysis of TJP occludin and extravasation proteins VCAM-1, ICAM-1 in the (E) hippocampus and (F) cortex of WT, 3xTG, and 3xTG +4ITP. [(E) and (F)] Data represents N = 6 mice per group, two technical replicates per mouse. Statistical significance was evaluated by two-way ANOVA, followed by Tukey’s post hoc test (main column effect). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Male and female mice are represented by blue and pink dots, respectively. Created in BioRender. Fossati, S. (2026) (https://BioRender.com/zg3f90e). We also found that VCAM-1 and ICAM-1 protein expressions were significantly increased in the hippocampus (Fig. 6E) and cortex (Fig. 6F) of 3xTG mice, and 4ITP significantly reduced the expression of both these endothelial activation markers in the cortex and of VCAM in the hippocampus. Last, in the hippocampus, there was a significant loss in the expression of occludin in 3xTG mice compared to WT, partially mitigated in 4ITP-treated 3xTG mice, which showed no significant difference from the WT animals (Fig. 6E). We found no significant changes in the expression of occludin in the cortex between WT and 3xTG mice, although 4ITP induced a significant increase in occludin expression compared to both WT and untreated 3xTg mice (Fig. 6F). Our results display some intersubject variability among some of these markers, particularly ICAM, which appears partially related to sex differences, in line with studies showing high variability in vascular proteins including TJP occludin and extravasation proteins VCAM-1 and ICAM-1 in aging as well as in models of AD (49–52). To evaluate changes in BBB permeability, we qualitatively assessed albumin leakage in the hippocampus of 3xTG mice. Strikingly, we observed an increase in albumin in the 3xTG mice brains and, particularly, areas of albumin leakage into the parenchyma (indicated with arrows) (Fig. 7A). These albumin-rich areas were not present in 4ITP-treated mice. 4ITP reduces microhemorrhages, improves hippocampal neuronal cell survival, and decreases Aβ40 and P-tau accumulation. (A) Representative images of co-staining CD31 and blood protein Albumin (green) in the hippocampus of WT, 3xTG, and 3xTG + 4ITP mice. Microhemorrhages are observed in 3xTG mice (indicated with arrows) and prevented in 3xTG-treated mice. Original magnification: ×60. (B) Representative images of NeuN (green) and DAPI (blue) in the hippocampus of WT, 3xTG, and 3xTG mice treated with 4ITP. Original magnification: ×60. (C) Quantification of %area NeuN in the hippocampus. (D) Correlation between % area NeuN and % area CD31 in the hippocampus of WT, 3xTG, and 3xTG + 4ITP mice. 4ITP-treated 3xTG had significant improvement in %area of NeuN and the correlation between % area NeuN and % area is significant. (E) There is a significant decrease in the insoluble Aβ40 as well as pTau231 in the hippocampus of 3xTG-treated mice. In all graphs, Male and female mice are identified with blue and pink dots, respectively. (C and D) N = 4 to 6 per group (indicated with dots). Statistical significance was evaluated by one-way ANOVA, followed by Tukey’s multiple comparison test. (E) N = 10 female 3xTG and N = 6 female 3xTG treated. Statistical significance was evaluated with t test Welch’s correction between 3xTG and 3xTG-treated mice. *P < 0.05 and **P < 0.01. Created in BioRender. Fossati, S. (2026) (https://BioRender.com/zg3f90e). To have a better understanding of the effects on neuroinflammation in vivo and continue to validate our in vitro findings, we measured cytokines and chemokines in the cortex of WT, 3xTG, and 3xTG 4ITP-treated mice. Similarly to the in vitro data, we observed a trending decrease in anti-inflammatory cytokine IL-10, which was significantly prevented in the brains of 4ITP-treated mice. We did not observe changes in IL-6 or IL-12p70 in the cortex of these mice. Differently, we saw a decrease in pleiotropic cytokine IL-4 in brains of 3xTG mice, which was significantly prevented in 4ITP-treated mice (fig. S7B). We also observed trends in increases in KC-GRO/CXCL1 and MIP-2/CXCL2 in 3xTG cortex, which were mitigated in 4ITP-treated 3xTG mice. As neuroimmune cell activation may be associated to endothelial cell death and endothelial cell activation, we measured the extent of astrogliosis in 4ITP-treated 3xTG mice (fig. S8, A and B); in the hippocampus and cortex, the % area of glial fibrillary acidic protein (GFAP) was significantly reduced in mice treated with 4ITP compared to untreated 3xTG mice. We also measured the concentration of 4-hydroxynonenal (4-HNE), a highly reactive aldehyde produced during lipid peroxidation, which is considered a marker of oxidative stress, in the cortex of WT, 3xTG nontreated, and treated mice, but we did not detect any significant changes (fig. S8C). To understand if the prevention of endothelial cell loss and preservation of BBB integrity are associated to the deterrence of neuronal loss and progression of neurodegenerative disease, we measured the % area of NeuN, a neuronal cell marker. We observed a decrease in % area of NeuN in the hippocampus of 3xTG mice, and the % area of NeuN was significantly increased in 4ITP-treated mice compared to nontreated 3xTG mice (Fig. 7, B and C). There were no significant changes in NeuN % area in the cortex of these mice (fig. S9, A and B). When we measured the colocalization between cleaved caspase-3 and NeuN, we observed no significant changes in the hippocampus between WT and 3xTG mice (fig. S9C). In the cortex, there was a significant increase in % area of colocalized NeuN and cCasp3 in 3xTG mice compared to WT, which was decreased in mice treated with 4ITP (fig. S9D). To appreciate the relationship between the endothelial cell and neuronal cell loss in mice, we measured the correlation between the % area of NeuN+ to the % area of CD31+ cells in the hippocampus (Fig. 7D) and cortex (fig. S9E) of WT, untreated 3xTG, and treated 3xTG mice. Intriguingly, we revealed a positive correlation both in the hippocampus (R = 0.74, P = 0.0025) and in the cortex (R = 0.5429, P = 0.0552). Further, we measured the levels of amyloid and phosphorylated tau in the hippocampus and prefrontal cortex of 3xTG female mice [as male 3xTG mice are known to have minimal accumulation of amyloid or P-tau at this age (48, 53)]. In the hippocampus, soluble and insoluble vasculotropic Aβ40 was significantly reduced in 4ITP-treated mice compared to untreated, suggesting a reduction in CAA, as we have previously shown for pan-CAIs (25). Insoluble ptau-231 was also significantly reduced in the hippocampus of 4ITP-treated 3xTG mice (Fig. 7E). Although there was much less accumulation, the trends were similar in the prefrontal cortex, where we observed a significant reduction in soluble and insoluble Aβ42 and a trending reduction in soluble Aβ40, while insoluble Aβ40 was not detected. Last, there were no significant changes in ptau-231 in the prefrontal cortex of treated or untreated 3xTG mice (fig. S9F). We then asked whether these beneficial changes in the brains of 4ITP-treated 3xTG mice were correlated with an improvement in cognition. To achieve this goal, we evaluated the mice’s cognitive ability over a 6-day period using the Barnes Maze, a paradigm used to measure the ability to consolidate and retrieve spatial memories. As expected, 3xTG mice exhibited an increased latency to find the escape hole throughout the learning trials. We first tested two different treatment times (6 to 16 months and 12 to 16 months) in the Barnes Maze, and the long-term 4ITP treatment (6 to 16 months), akin to treating MCI subjects in human trials, was much more successful at improving spatial memory compared to the shorter treatment (fig. S10). Therefore, we focused on the longer (6 to 16 months) treatment throughout this study. A significant difference between treated (6 to 16 months) and untreated mice was observed on days 1 and 2 (training) of the maze (Fig. 8A). Moreover, 4ITP-treated 3xTG mice did not show significant differences from WT mice on any day. On probe day (day 6), untreated 3xTG mice demonstrated a significant increase in the time to find the escape hole (latency) and number of mistakes, which was significantly reduced to WT levels in 4ITP-treated 3xTG animals (Fig. 8B). In addition, we did not observe any spatial memory differences between WT and WT + 4ITP (6 to 16 months) mice, suggesting lack of undesired cognitive effects of 4ITP in heathy participants (fig. S11, A and B). 4ITP prevents cognitive decline in 3xTG mice. (WT, N = 20; 3XTG, N = 21; 3XTG + 4ITP, N = 20; males and females, 16 months) Spatial memory was assessed using Barnes Maze over a 6-day period (A) WT, 3xTG-treated, and nontreated mice learning trajectory (latency, s) to find the escape hole overtime. (B) On probe day, the number of mistakes (F2,58 = 10.86, P ≤ 0.001) and latency to find the escape hole (F2,58 = 7.173, P = 0.0017) in each group was recorded. (C) Cued fear-conditioning test is improved in 3xTG mice treated with 4ITP. (D) Open field total distance traveled (motor activity) displayed no significant differences. The % of time in the center showed a trend to increased time in the center (decreased anxiety) for 4ITP-treated 3xTG mice compared to untreated 3xTg. (E) Rotarod test displayed no locomotor differences between groups. (F) Novel object recognition test is impaired in 3xTG mice and then improved in mice treated with 4ITP. Male and female mice are identified with blue and pink dots, respectively. (A) Statistical significance was evaluated by two-way ANOVA, followed by Tukey’s multiple comparison test. [(B) to (F)] Statistical significance was evaluated by one-way ANOVA, followed by Tukey post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Created in BioRender. Fossati, S. (2026) (https://BioRender.com/zg3f90e). In a fear-conditioning paradigm, assessing cue-induced fear associative memory, we observed a memory impairment (reduced freezing time) in 3xTG mice compared to WT, which was significantly reverted in 4ITP-treated mice (Fig. 8C). We did not see an impairment in WT-treated mice, again indicating a lack of undesired side effects (fig. S11C). Through open field test and rotarod, we detected no significant locomotor changes between WT and 3xTG mice (Fig. 8, D to F) or WT and WT + 4ITP (fig. S11, D and E). In addition, in the open field test, 4ITP-treated 3xTG mice showed a trend to increased time in the center compared to untreated 3xTG animals, suggesting a reduction in anxiety behavior upon treatment. Last, we observed a preservation in object memory of 3xTG-treated mice. With the novel object recognition test, we observed a significant reduction in the % time exploring the novel object during the testing phase in 3xTG mice compared to WT, indicating a loss in working memory. The time exploring the novel object was significantly increased in 4ITP-treated 3xTG mice, indicating a preservation of working memory (Fig. 8F). When separating the mice by sex we did observe some sex differences, although not significant (fig. S12). Notably, 3xTG male mice were larger and moved slower than female mice in the Barnes maze, as indicated by their similar trends but different degrees of significance in the latency to find the escape hole (fig. S12, A and B and D and E), as previously reported (51, 53). However, the latency to find the escape hole on probe day was improved by 4ITP in both males and females. Both males and female 3xTG mice made a significantly higher number of mistakes compared to WT mice on probe day, but 4ITP treatment resulted in a more significant improvement in the number of mistakes in female mice (fig. S12, B and E). In addition, we observed a rescue in cue-associated fear memory in both females and males (fig. S12, C and G). In the novel object recognition test, we observed an increase in %time exploring the novel object in both female and male 4ITP-treated mice; however, this increase was significant in male treated mice (fig. S12, D and H). We did not observe any significant differences in the open field test when analyzing the % of time spent in the center or the total distance traveled (fig. S12, I and J). There was a significant improvement in motor function (rotarod) in male 4ITP-treated 3xTG mice compared to untreated 3xTG mice (fig. S12K). Overall, these data confirms that CA-V inhibition—by preventing apoptosis, cerebrovascular dysfunction, and neuroinflammation in the brain of 3xTG AD mice—significantly rescues cognitive impairment. DISCUSSION This study demonstrates that mitochondrial CA-V is a promising target for the treatment of AD, CAA, and related cerebrovascular pathologies. We showed that the selective CA-V inhibitor 4ITP, more lipophilic and selective on CA-V than acetazolamide and methazolamide, prevents endothelial AβQ22-induced mitochondria-mediated apoptosis, caspase-9 activation and CytC release, loss of mitochondrial membrane potential, mitochondrial ROS production, loss of hCMEC barrier resistance, and endothelial inflammatory activation. In addition, through CA-VB CRISPR-KO cells, we confirmed mechanistically that the mitochondrial isoform CA-VB is necessary to promote AβQ22-induced endothelial mitochondria-mediated apoptosis, oxidative stress, and loss of EC barrier integrity. We anticipate the CA-V inhibitor and CA-VB KO would be also protective against WT Aβ40-induced mitochondria-mediated apoptosis and CEC activation, as Aβ40 has been demonstrated to induce similar mechanisms, when in oligomeric or proto-fibrillar state, in hCMEC, although further studies are warranted (19). Strikingly, the same CA-V inhibitor was protective in the 3xTG mouse model through the prevention of apoptosis, cerebrovascular dysfunction, and inflammatory activation, resulting in a reduction in AD pathology and a potent improvement in memory. Cerebrovascular and mitochondrial dysfunction are early and causative events observed in many diseases associated with aging (7, 9, 10, 14, 20, 26, 54–56). Both cerebrovascular and mitochondrial dysfunction are observed to occur even before Aβ and tau accumulation and worsen throughout the progression of AD and CAA, causing a vicious loop of impaired clearance, toxic protein accumulation, neuroinflammation, exacerbated cerebrovascular, and mitochondrial/metabolic deficits. Because of the high impact of cerebrovascular and mitochondrial deficits on the development and progression of AD and CAA, it is evident that these are both ideal targets for the development of novel therapeutic strategies against AD and CAA. Current FDA-approved therapies for AD do not appear to improve cerebrovascular function. On the contrary, toxic effects of the new anti-Aβ antibodies, increasing Aβ deposits on the BBB, are likely to be responsible for the development of ARIA-E (edema) or ARIA-H (hemorrhages) in a substantial percentage of patients treated with these new drugs (5, 57), emphasizing the importance and need for treatments targeting vascular dysfunction in AD and CAA. Moreover, mitochondrial and metabolic deficits are also frequently discussed among the main pathways to target for the development of new drugs against AD and dementias. However, mitochondria-targeted therapies have not yet reached FDA approval for these diseases (14, 55, 56). A multitude of evidence demonstrates that lifestyle factors such as diet and exercise, as well as cardiovascular risk factors, can influence both BBB and mitochondrial function, resulting in the modulation of risk and progression of symptoms associated with the disease (1, 7, 58, 59). Cardiovascular risk factors, in particular, may increase cerebrovascular complications in patients with AD and CAA, through their toxic effects on CEC (7, 21). CAs catalyze the hydrolysis of carbon dioxide to produce bicarbonate and a proton. This is an essential reaction to maintain physiological cell health, and particularly within the mitochondria, due to the high production of carbon dioxide as a by-product of metabolism. The use of CA inhibitors has been first explored in cardiovascular disease (60, 61). Pan-CA inhibitors such as ATZ and MTZ were developed as legacy diuretics, being optimized to modulate the function on the kidney and water reabsorption. They are relatively polar (log P ATZ = −0.26, log P MTZ = 0.13) as they were optimized to work primarily at the level of the kidneys. They have then become FDA-approved for many diseases such as glaucoma, epilepsy, and high-altitude sickness (2, 34). More recently, they have also been demonstrated to be effective in models of obesity, cerebral edema, stroke, and diabetes-induced cerebrovascular pathology (27, 30, 62). Many of the studies, dissecting the involvement of CA in cardiovascular pathology or stroke, have used pan-CA inhibitors, most commonly ATZ and MTZ (2, 63). As previously mentioned, ATZ and MTZ have also been proved effective against Aβ-mediated toxicity in many NVU cells (23, 24, 26) and in a CAA mouse model, the TgSwDI (25). In addition, topiramate, another CA inhibitor, has been demonstrated to be protective in models of diabetes-induced cerebrovascular pathology (62, 64). Although many pan-CA inhibitors have protective effects, they are not ideal for therapeutic use, as their lack of selectivity requires significantly higher doses to achieve efficacy, which cause undesired side effects when used systemically in humans (65). Our previous studies have demonstrated that ATZ and MTZ prevent Aβ toxicity, specifically through mitochondria-mediated mechanisms (19, 23, 24, 26). Here, we tested for the first time a mitochondria-selective CA-V inhibitor, and we showed that it is effective at similar concentrations to ATZ (the most effective of the pan-CAIs we have tested in vitro and in vivo), against mitochondria-mediated apoptosis in hCMEC (24). Hence, we anticipate that selective CA-V inhibitors will be at least as effective as pan-CA inhibitors against AD/CAA while avoiding side effects that have been reported in humans for pan-CAIs. One of the possible side-effects to be considered when using high concentrations of pan-CA inhibitors, such as ATZ, is short-term memory impairment. A high dose of ATZ has been observed to inhibit long-term potentiation, acutely, in mice (66, 67). In addition, humans have reported memory reduction when taking ATZ at high concentrations (66). This effect may be due to the activity of nonselective inhibitors on a specific CA isoform, CA-VII, abundantly expressed in pyramidal neurons of the hippocampus (68). However, the dose of ATZ used in our animal models (20 mg/kg per day), which is the same dose of 4ITP in the present study, corresponds to a human dose of 2.6 mg/kg per day (182 mg/day for a 70-kg person) and is lower than the FDA-approved dose for CAIs (ATZ: up to 1 g/day), which is the dose known to induce short-term memory issues. Nevertheless, to confirm the safety and lack of this side effect for 4ITP, we have shown here that, while 4ITP treatment ameliorates memory in the 3xTG AD mouse model, the chronic use of 4ITP on WT animals at the proposed concentration does not impair spatial memory. In addition, when comparing two courses of treatment, starting at 6 or 12 months, we observed that the long treatment was much more effective than the later treatment. This result suggests CA-V inhibition is likely targeting early mechanisms of the disease such as cerebrovascular and mitochondrial dysfunction as well as neuroinflammation. Because of the diverse expression pattern and functions of CAs, it is evident that not all 15 CA isoforms are involved in the development of AD pathology. CA-V, expressed in the mitochondrial matrix, has been demonstrated to influence intermediates of the citric acid cycle through the supply of bicarbonate for enzymes such as pyruvate carboxylase and propionyl carboxylase, thereby affecting biosynthetic pathways, such as gluconeogenesis and ureagenesis. However, many of these studies have been done in the liver, where CA-VA is much more abundant. The involvement of CA-VB in metabolism is only evident in the absence of CA-VA (69). This is observed in cells where CA-VA is not expressed and in Car5a KO mice. Less is known about the function of CA-VB; however, it is much more abundant throughout the body and in the brain (2). Car5B global KO mice, differently from Car5A global KO mice, are viable and show normal growth and normal ammonia levels, suggesting that chronic inhibition of CA-VB or Car5B KO does not have toxic effects (69). We have shown here that knocking out CA-VB completely prevents mitochondria-mediated apoptosis induced by Aβ in hCMEC and mitigates the loss of barrier resistance. Thus, selectively targeting the mitochondrial isoform CA-VB, while avoiding undesired effects on other CA isoforms not involved in the progression of AD and CAA, may be of high interest for therapeutic development in AD-related disorders as well as, potentially, against cerebrovascular complications induced by current treatments with FDA-approved anti-Aβ antibodies (such as ARIA). While 4ITP, which inhibits both CA-VA and CA-VB, completely rescues barrier resistance in presence of Aβ, CA-VB KO only partially mitigates Aβ-induced BBB permeability. When treating CA-VB KO cells with 4ITP, however, we do see a complete rescue in resistance. These constitutes important knowledge, as 100% selectivity for CA-VB is not easily achievable, and these results suggest that CA-VA (or additional CA isoforms 4ITP may inhibit) could also be involved in the induction of BBB permeability in CEC. It is clear, however, that 4ITP and CA-VB KO do significantly improve barrier resistance and should be further explored as a potential therapy against cerebrovascular pathologies. Future studies comparing CA-V inhibitors and their selectivity profiles would help us to better understand the other players that may be involved. We are confident that 4ITP is selective for CA-V compared to other highly expressed CA isoforms (36); however, other, even more selective CA-VB inhibitors will be compared in future studies to further investigate the pharmacodynamics of CA-V inhibition. Besides CA-V’s role in metabolism, we have demonstrated here that CA-VB activity directly influences mitochondria-mediated apoptosis and mitochondrial membrane potential, as well as oxidative stress. We anticipate that while CA-VA may function mainly to regulate metabolism, CA-VB regulates mitochondrial pH and redox homeostasis. Mechanistically, the inhibition of CA activity in the mitochondrial matrix is expected to reduce the overproduction of hydrogen ions, especially in times of stress (such as amyloidosis), maintaining a lower proton level within the mitochondrial matrix compared to the intermembrane space. A lower H+ concentration in the matrix will aid in preserving a healthy mitochondrial membrane potential, in addition to an optimal proton motive force, especially in the presence of ROS. This is essential for proper mitochondrial function, cell signaling and, ultimately, cell survival. We have demonstrated that 4ITP prevents the loss of mitochondrial membrane potential, mitochondrial H2O2 production, and CytC release, confirming the mitochondria-specific mechanism of action of the drug, as also supported by the same results in CA-VB KO cells. A plethora of evidence indicates that preservation of mitochondrial function and mitigation of ROS prevents downstream neuroinflammatory mechanisms. ROS production and release have been observed to induce pro-inflammatory cytokine production, resulting in the activation of glial and vascular cell types. In hCMEC and mouse cortex, we observed a preservation in anti-inflammatory cytokine IL-10 with 4ITP treatment. The expression of IL-10 has been observed to negatively correlate with mitochondrial function and oxidative stress (70). This is an indicator that the prevention of ROS through 4ITP could be preventing the down-regulation of IL-10, allowing for improved cellular function and cell survival, as well as diminishing chronic inflammatory mechanisms observed to cause damage in AD. When CEC are activated, they increase their expression of adhesion proteins VCAM-1, E-selectin, P-selectin, and ICAM-1 as well as releasing chemokines and cytokines CXCL1/2 and CSF2/3 known to promote immune cell extravasation and EC activation (49). While recruitment of immune cells is important for the clearance of toxic substances from the brain, chronic inflammation and CEC activation are damaging to the BBB and thus to the NVU and the brain. The expression of VCAM-1 and ICAM-1 was reduced by 4ITP in the presence of AD pathology in hCMEC and 3xTG mice. In hCMEC, amyloid induced increased ICAM-1 expression at earlier time points and transiently compared to VCAM-1; thus, it is likely that ICAM-1 expression in the brain could be also attributed to tau or other mediators of pathology. The increase in CXCL2/MIP-2, a known recruiter of neutrophils, was also mitigated by 4ITP in vitro and in vivo. Neutrophils have been observed to be very damaging to brain tissue, particularly through the injury of capillaries (71, 72). BBB dysfunction and EC activation lead to many downstream negative consequences, exacerbating a viscous cycle of cerebrovascular dysfunction and neuroinflammation, including glial cell activation (2, 14, 25), and eventually neuronal loss. Intriguingly, the observed prevention of CEC activation and preservation of endothelial cell function as well as mitigation of astrogliosis correlate with a preserved area of hippocampal neurons and markedly improved cognition in 4ITP-treated 3xTG mice. The accumulation of Aβ around the brain vessels—found in AD, sporadic CAA, CAA-related inflammation (CAAri), and in ARIA complications due to high amounts of Aβ unable to be cleared by the cerebral vasculature—is known to exacerbate BBB dysfunction and induce permeability, enhancing the likelihood of ischemic and hemorrhagic strokes and white matter abnormalities (6, 73). Therefore, we anticipate that mitochondrial CA inhibitors could not only be a potential therapy for AD and CAA when used alone, but that the combination of CA inhibitors and anti-Aβ therapies could be a very beneficial cotreatment strategy. Further studies are needed to confirm this hypothesis. We observed a reduction of amyloid (particularly CAA-associated Aβ40) and phosphorylated tau accumulation in the hippocampus of 3xTG-treated female mice, likely due to preservation of cerebrovascular function and reduction of neuroinflammation, which facilitate improved clearance. We have demonstrated that both female and male mice are experiencing vascular abnormalities, endothelial cell activation, and gliosis, although males have very little/no amyloid accumulation. Notably, 4ITP rescues cognition, vascular health, and neuroinflammation in both females and males (Figs. 6 to 8, and fig. S12). These results suggest that CA-V inhibition is protecting against the toxic effects of amyloid but, more strikingly, against the underlying mechanisms, such as endothelial cell stress/apoptosis, endothelial and neuronal cell loss, and neuroinflammation, getting to the root causes of this multifactorial disease through multiple of the mechanisms that are known to cause it or mediate it. One limitation of this study is that it is currently impossible to have complete pharmacological selectivity for one CA-V isoform. 4ITP has been previously shown to have a Ki value of 8.6 and 8.3 nm for CA-VA and CA-VB respectively, indicating equal selectivity for the mitochondrial isoforms. 4ITP is about 8× less effective on CA-I and CA-II (Ki = 75 and 54 nm, respectively) and even less on plasma membrane isoforms CA-IX and CA-XII (136 and 212 nm) (35, 36). Although this study, by comparing the effects of 4ITP with CA-VB KO cells, provides strong evidence that 4ITP effects have CA-V specificity, the activity of 4ITP on other CA isoforms relevant to the brain/NVU activity remains to be determined. Of interest, CA-IV is highly expressed on the plasma membrane of CEC and should be considered when studying CA function in CEC. Furthermore, knowledge on the activity of 4ITP on CA-VII would be informative, as this isoform is involved in memory and highly expressed in neurons. The lack of an effect of 4ITP on WT animals’ behavior, however, suggests low affinity for this isoform. Studies providing more evidence about the selectivity of 4ITP would help us understand the mechanism of action and to design other selective CA-V inhibitors in future studies. Another limitation is that we did not measure BBB function or CBF in real time in vivo. However, we did uncover a decrease in blood vessel diameter in the brains of 3xTG mice compared to WT. This was rescued with 4ITP suggesting that it did in fact improve vessel function and integrity. We also showed a prevention of albumin leakage into the brain, suggesting reduced BBB permeability in vivo. Future studies under development in our laboratory intend to assess the effects of CA-V inhibitors and cerebral endothelial-specific CA-VB KO on CBF, BBB permeability, and neurovascular coupling in AD mouse models. These studies will also determine the effects of 4ITP on oxidative stress, microglia phenotype, and additional neurodegeneration markers in vivo, as we have previously shown for MTZ and ATZ (25). We did not focus on the effects of 4ITP on microglia in this manuscript, as this is an ongoing study in the laboratory and will be published with detail in the future. Understanding the direct effects of CA-V inhibitors on neurons is another avenue of research currently ongoing in the laboratory. We have previously demonstrated pan-CA inhibitors protect from amyloid-induced cell death in neurons; therefore, we anticipate that CA-V–selective inhibitors may also have a direct effect on neurons. The strong protective effects of 4ITP on CEC function and on cognition shown here suggest potential benefits in all those mechanisms in other neurovascular cells. We observed in this study that CA-V inhibition increased the % area of NeuN in the hippocampus while reducing NeuN and cleaved caspase-3 colocalization, particularly in the cortex. Studies using endothelial-specific CA-VB KO mice will be needed to understand if reduction of CA-V activity specifically in endothelial cells also causally prevents neuronal death. In addition, possible acute effects of endothelial dysfunction on neuronal cell survival remain to be determined, and future studies using coculture systems would allow to determine whether the inhibition/KO of CA-V in endothelial cells treated with amyloid can prevent acute neuronal toxicity. This study focused on the involvement of CA-V in amyloid-induced mitochondrial-mediated apoptosis; however, it would be interesting to consider other cell death pathways known to be associated with neurodegenerative disease such as pyroptosis and ferroptosis in future studies. In addition, only post-mortem measures of oxidative stress with 4-HNE were performed in this study. Further studies will be needed to assess mitochondrial function, metabolism, and oxidative stress outcomes in fresh brain tissue by seahorse assays of metabolic/mitochondrial function, single-cell analysis, or other methods. Overall, this study is pioneer in demonstrating the protective effects of selective CA-V inhibition in models of amyloidosis and AD. We have shown that the CA-V inhibitor 4ITP and CA-VB CRISPR KO protect CEC from Aβ-induced mitochondria-mediated apoptosis and BBB dysfunction and that 4ITP is effective to rescue brain apoptosis, endothelial cell survival, cerebrovascular and neuroimmune function, and cognitive impairment in 3xTG mice. Together, it is evident that the development and testing of CA-V inhibitors and understanding the role of mitochondrial CA-VB are critical areas of research for therapeutic targeting of cerebrovascular pathology and mitochondrial dysfunction in AD, CAA, ADRD, and the vascular contributions to cognitive impairment and dementias. MATERIALS AND METHODS Cell culture Immortalized human cerebral microvascular endothelial cells (hCMEC/d3) were provided by Babette Weksler (Cornell University). Cells were grown and maintained in endothelial basal medium (EBM-2) (Lonza, CC-3156) supplemented with EGM-2 growth factors (Lonza, CC-4176) and 5% fetal bovine serum (FBS). CRISPR-Cas9–mediated KO cell pool of CA-VB in hCMEC/D3 cells were generated by EditCo Bio Inc. (Redwood City, CA, USA). To generate these cells, ribonucleoproteins containing the Cas9 protein and synthetic chemically modified guide RNA produced by Synthego were electroporated into the cells using EditCo’s optimized protocol. Editing efficiency is assessed upon recovery, 48 hours postelectroporation. Genomic DNA is extracted from a portion of the cells, PCR amplified, and sequenced using Sanger sequencing. The resulting chromatograms are processed using EditCo’s Inference of CRISPR edits software. The WT cell line was generated similarly without guide RNA specific for the gene of interest. Upon receiving the cells and before experiments, CA-VB KO was confirmed via protein expression. Cells were grown and maintained in EBM-2 supplemented with EGM-2 growth factors and 5% FBS. Cells were grown in a humidified cell culture incubator, under a 5% CO2 atmosphere at 37°C. WT hCMEC were routinely checked to ensure that they maintained their proper function and TJP expression, as CRISPR-Cas9 CA-VB KO to ensure CA-VB KO stability. Synthesis and characterization of the CA-V selective inhibitor 4ITP The following materials were used as received: Phenylacetic acid and thionyl chloride were from Sigma-Aldrich (St. Louis, MO), 4-aminomethylbenzenesulfonamide hydrochloride was from Oakwood Chemical (Estill, SC), and N-methylmorpholine was from TCI America (Portland, OR). Organic solvents [high-performance liquid chromatography (HPLC) quality] were purchased from Thermo Fisher Scientific (Pittsburgh, PA), EMD (Gibbstown, NJ), and VWR International (West Chester, PA). The purity and the structure identity of the intermediary and final products were assessed by thin-layer chromatography (TLC), HPLC–mass spectrometry (MS), 1H–nuclear magnetic resonance (NMR), and 13C-NMR. TLC was carried out on SiO2-precoated aluminum plates (Alugram SIL G/UV254 20 cm by 20 cm with F254 indicator; layer thickness of 0.20 nm; pore size of 60 Å) from MACHEREY-NAGEL (Bethlehem, PA). Flash chromatography was performed using an Isco Combiflash RF (Teledyne Isco, Lincoln, NE) using Isco prepacked silica columns. The purity of compounds was also assessed via liquid chromatography–MS (LC-MS) using an Agilent 1200 HPLC DAD-MS system (Santa Clara, CA) equipped with a G1315A DAD and a 6130 Quadrupole MS via a ZORBAX SB-C18 column eluted with H2O (0.1% HCOOH)/MeCN (0.1% HCOOH) 95/5 to 0/100 linear gradient. NMR spectra were recorded at T = 300 K with a Bruker Avance III 400 Plus spectrometer equipped with a 5-mm indirect detection probe, operating at 400 MHz for 1H-NMR at 100 MHz for 13C-NMR. Chemical shifts are reported as δ values, using tetramethylsilane as internal standard for proton spectra and the solvent resonance for carbon. 4-(aminomethyl) benzenesulfonamide hydrochloride (1 eq, 4.5 mmol, 1.00 g) was suspended in dry acetonitrile (18 ml) in a 50-ml round bottom flask, and the flask was purged with argon. The mixture was cooled to 0°C using an ice bath, then 4-methylmorpholine (2.1 eq, 9.4 mmol, 0.95 g) was added under stirring. After 30 min, the ice bath was removed, and the reaction mixture was stirred overnight at room temperature. Separately, phenylacetic acid (1.05 eq, 4.73 mmol, 0.73 g) was treated with thionyl chloride (10 eq, 45 mmol, 5.35 g) and 1 drop of dimethylformamide in a 100-ml round bottom flask, and the mixture was heated to reflux, under stirring, for 1 hour. Excess thionyl chloride was removed in vacuo using a rotary evaporator, and the acid chloride was dissolved in dry acetonitrile (3 ml) and added dropwise to the previously made solution of 4-(aminomethyl)benzenesulfonamide at 0°C (ice bath cooling). The reaction was stirred at 0°C for 30 min abd then at room temperature overnight, when a precipitate was formed. The suspension was filtered, and the precipitate was washed subsequently with acetone and water, dried under vacuum, and absorbed onto SiO2 using MeOH/CHCl3. The product was purified by flash chromatography on silica gel (0 to 40% MeOH/CHCl3 gradients). The pure fractions were combined and evaporated to dryness to yield a while solid [55 to 71% (multiple batches made)]. 4ITP: 1H-NMR [400 MHz, dimethyl sulfoxide (DMSO)–d6, δ, parts per million (ppm)]: 8.63 (t, J = 5.8 Hz, 1H, NH), 7.74 (d, J = 8.4 Hz, 2H, H2,6-PhSO2NH2), 7.3 (d, J = 8.4 Hz, 2H, H3,5-PhSO2NH2), 7.20–7.35 (m, 5H, PhCH2CO), 7.29 (s, 2H, SO2NH2), 4.32 (d, J = 5.9 Hz, 2H, NHCH2), 3.49 (s, 2H, CH2CO). 13C-NMR (100.6 MHz, DMSO-d6, δ, ppm): 170.2, 143.5, 142.5, 136.2, 128.9 (2C), 128.1 (2C), 127.3 (2C), 126.3, 125.5, 42.2, 41.7. 4ITP: LC-MS (ESI) 305.1 (MH+), tR = 2.47 min, >96% purity. Determination of 4ITP solubility in water and PBS buffer The solubility of 4ITP in water and in phosphate-buffered saline (PBS) buffer was determined by the same HPLC method and instrument described above. A stock solution of 4ITP at a concentration of 10 mM were prepared in DMSO and serial dilutions (5, 2.5, 1.25, 0.625, 0.3125, and 0.156 mM) were done with the same solvent. A volume of 1 μl of these 4ITP solutions was injected in the instrument, and the resulting peak areas were plotted against the corresponding concentrations to obtain a calibration curve via the least squares linear regression analysis. 4ITP-saturated solutions in deionized water and in PBS buffer were made by suspending 10 mg of pure 4ITP in 1 ml of water or in 1 ml of PBS at pH 7.4 in two Eppendorf tubes, vortexing the suspensions for 1 min, and subsequently placing them in a mini rotator (Labnet International, Edison, NJ) overnight. The next day, the undissolved compound was centrifuged down using a bench centrifuge for 5 min, and 10 μl of each supernatant was collected in an LC-MS vial. A volume of 1 μl of these 4ITP solutions was injected in the HPLC instrument, and the concentration of 4ITP in each solution was determined from the resulting peak areas via the previously constructed calibration curve. The analysis was done in triplicate, and the values were reported as means ± SD [4ITP solubility in water: 0.057 ± 0.011 mg/ml (0.187 mM) and 4ITP solubility in PBS (pH = 7.4): 0.0399 ± 0.003 mg/ml (0.13 mM)] 4ITP lipophilicity (log P) determination The log P value of 4ITP was determined using the same HPLC method and instrument described above using an injection volume of 1 μl and a flow rate of 1 ml/min in all cases. DMSO solutions (1 mg/ml) of reference compounds of known log P (indomethacin, caffeine, ketoprofen, phenytoin, and testosterone) were injected into the HPLC system, and the retention time of each compound was determined in the same RP C18 column. A calibration curve was constructed by plotting the log P values of these reference compounds against their retention time (74). Subsequently, 1 mg of 4ITP was dissolved in 1 ml of DMSO and was injected in the same system to determine the retention time, which was interpolated within the calibration curve to determine the log P value of 4ITP (4ITP log P value: 1.69). 4ITP solubilization and delivery in vitro CA-V inhibitor, 4ITP, was dissolved in DMSO at a concentration of 10 mM and stored at −20°C. On the day of treatment, the 10 mM stock concentration was diluted to 2 mM in EBM2 with 50% FBS. Last, 4ITP was diluted to its final concentration (1, 10, or 30 μM) in treatment medium, EBM2, with 1% FBS. Aβ preparation Aβ40-Q22 was synthesized by peptide 2.0. To obtain monomeric Aβ40-Q22, the peptide was weighed and dissolved in 1,1,1,3,3,3-169 hexafluoro-2-propanol (Sigma-Aldrich, St. Louis, MI, USA) at a concentration of 1 mM and incubated overnight at room temperature to dissolve β sheet structures. The day after, the peptide was flash-frozen in dry ice-cold 100% ethanol, then lyophilized until the pellet was dry, and stored at −20°C. To maintain the peptide in its monomeric form, the pellet was resuspended at 10 mM in DMSO and then diluted to 1 mM with sterile dH2O immediately before treatment. Aβ40-Q22 was brought to a final concentration of either 25 or 10 μM in EBM2 with 1% FBS for treatment and added to the cells. DNA fragmentation/apoptosis Cells were seeded at a density of 20,000 cells per well in a 24-well plate 24 hours before treatment in complete cell culture media. Following treatments, cells were centrifuged at 200g for 10 min, and the appropriate lysis buffer was then added (Roche Applied Science). DNA fragmentation was measured using the Cell Death enzyme-linked immunosorbent assay (ELISA) kit (Roche Applied Science), as recommended by the manufacturer. Absorbance was measured with SpectraMax I3x Multimode micro-plate reader (Molecular Devices) at wavelength 405 nm. Absorbance values were analyzed by determining the relative change to the control through the calculation of fold change. Cell viability assay Cells were seeded at a density of 10,000 cells per well in a 96-well plate in complete cell culture media 24 hours before treatment. Cells were treated in EBM-2 supplemented with 1% FBS plus the desired concentrations of Aβ40-Q22 and 4ITP. Cell viability was measured after 24 hours of treatment. WST-1 reagent (Roche, Applied Bioscience) was added directly to the cell culture media (1:10 ratio) and then incubated for 3 hours at 37°C to allow for the colorimetric reaction. Absorbance was read with SpectraMax I3x at a wavelength of 450 nm. Cell event caspase-3/-7 A day before treatment, 10,000 cells per well were seeded in a 96-well plate in complete cell culture media. Cells were treated in EBM-2 supplemented with 1% FBS and the experimental concentrations of Aβ40-Q22 and 4ITP. After 24 hours of treatment, the Cell Event Caspase-3/7 reagent (Thermo Fisher Scientific) was diluted in Hank’s balanced Salt Solution (HBSS) at a concentration of 5 μM, added at a volume of 100 μl per well to the cells, and incubated at 37°C for 30 min. The reagent was then removed and replaced with HBSS during acquisition. Pictures were acquired on an EVOS M5000 microscope at ×10 magnification. Images were acquired with the brightfield and green fluorescent protein channels to visualize the total number of cells and the caspase-3/-7 activity, respectively. Three pictures per well were taken and then analyzed using Fiji/ImageJ software. The number of active caspase-3/-7–positive cells was normalized to the total number of cells in each image and expressed as percentage of caspase-3/-7–positive cells. Western blot In hCMEC, 1 day before treatment, cells were counted and seeded in a six-well plates at a density of 350,000 cells per well. Following treatment, proteins were isolated from cells. Briefly, medium was collected, and cells were washed with 1× PBS and lysed with radioimmunoprecipitation assay buffer (Invitrogen) + Halt protease/phosphatase inhibitors (Thermo Fisher Scientific). Cell debris were pelleted at 16,000g for 20 min at 4°C. Supernatant was collected, and protein concentration was quantified with BCA method and normalized to 25 μg in each sample. For WB analysis, cell lysates and homogenates from hippocampus and cortex of WT, 3xTG, and 3xTG treated with 4ITP were prepared using BOLT 4× loading buffer and 10× reducing agent or Laemmli buffer 6× (Invitrogen), depending on which gel was used. To further denature the samples, they were incubated for 5 min at 95°C. Proteins were separated using 4 to 12% BOLT bis-tris SDS polyacrylamide gels (Invitrogen) or 10% CRITERION SDS polyacrylamide gels (Bio-Rad) and transferred onto 0.45-μm nitrocellulose membrane using 1× Towbin buffer containing 20% methanol at 110 volts for 70 min. After transfer, membranes were blocked with 5% milk in TBS. Primary antibodies—Caspase-9 (1:500; ab202068), Bax (1:500; Novus Biologicals, NBP1-88682), Bcl-2 (1:500; ab196495), Bim (1:500; ab32158), Bak (1:250; ab104124), CA-VB (Novus Biologicals, NBP1-86090), Occludin (1:250; Invitrogen, 71-1500), Claudin-5 (1:500; Invitrogen, 352500), ICAM-1 (1:500; Invitrogen, MA5407), VCAM-1 (1:500; ab134047), and actin (1:3000; Millipore, MAB1501)—were incubated on the membrane O/N at 4°C or for 2 hours room temperature. Membranes were then washed with TBS-T (0.1% Tween) and incubated with the species-appropriate secondary antibody (1:20,000 in TBS; Licor) for 1 hour at room temperature. Images were acquired using LICOR Odyssey CLx Immunoblot imager and then analyzed with LICOR Image Studio Software. Uncropped membranes of each representative WB have been provided in the Supplementary Materials. CytC and Mito-Tracker staining Before seeding, eight-well chamber glass slides were precoated with attachment factor (Cell Systems). For immunocytochemistry experiments, 15,000cells per well were seeded 24 hours before treatment. Cells were treated with 25 μm Aβ40-Q22 with or without 4ITP. After 16 hours of treatment, Mito-Tracker CMH2xROS (Invitrogen) was diluted in 1% FBS EBM2 (1:2000) and added to the cells for 30 min at 37°C. This is a dye that only enters mitochondria with a healthy membrane potential. After live-cell staining was complete, cells were washed 1× with PBS and fixed with 4% paraformaldehyde for 15 min at 37°C and then washed again 3× with PBS for 5 min. Cells were then permeabilized with 0.2% Triton in PBS for 10 min at room temperature, followed by 1 hour of blocking with 3% bovine serum albumin (BSA) in PBS. Then, cells were incubated with CytoC–Alexa Fluor 488 antibody (BD Pharmigen, 560263) for 1 hour at room temperature. Last, to visualize the nucleus, we used DAPI mounting media (Southern-Biotech, 0100-20). Images were acquired using a Nikon Ti2-E fluorescence deconvolution microscope at 100× resolution. Mitochondrial H2O2 production HCMEC were seeded at 2 million cells per well in 10-cm dishes 24 hours before treatment and then treated with Aβ40-Q22 with or without 4ITP. After 16 hours of treatment, mitochondria were isolated in an isotonic buffer including 75 mM sucrose, 10 mM tris-HCl (pH 7.4), 2 mM EGTA, and 225 mM mannitol buffer (24). The mitochondrial fraction was quantified using BCA assay. H2O2 was measured with an Amplex red assay (Invitrogen). Levels of H2O2 were measured by absorbance detected at 560 nm, normalized to protein quantification of the mitochondria fraction, and then expressed as fold of change to the control. Lipid peroxidation 4-HNE ELISA The levels of 4-HNE in total cell lysates were measured with a commercially available ELISA kit (Abcam, 238538) as per the manufacturer’s instructions. Protein samples were collected after 16 hours of treatment and then flash frozen with dry ice. The levels of 4-HNE were calculated through a standard curve and detected with Spectra I3x Max at a wavelength 450 nm. Concentrations were further converted to % change of control. Mouse cortices were homogenized following brain tissue dissections. The same commercially available kit (Abcam, 238538) was used for samples in vivo normalized to equal milligram of protein. Measurement of barrier resistance by ECIS Zθ Transendothelial electrical resistance was measured using the ECIS ZΘ system (Applied Biophysics). All experiments were performed on eight-well ECIS (8WE10+, Applied Biophysics) 40 electrode gold-plated arrays, precoated as per the manufacturer’s instructions. Cells were seeded at a density of 200,000 cells per well, given 48 hours to form a monolayer, and visualized on the recorded plot as a plateau in Resistance over time. The frequency applied to the cells was 4000 Hz. At this point, cells were treated with 10 μM Aβ40-Q22, with or without 4ITP or 4ITP alone. The resistance was measured for 48 hours after treatment. Data were expressed as normalized to the untreated control, over a 48-hour period. MSD cytokine analysis Cytokine release was evaluated in CM from untreated hCMEC and Q22-treated cells with or without 4ITP. CM (900 μl) was collected and then centrifuged at 800g for 5 min at 4°C. CM was then flash frozen and stored at −80 for no longer than 1 month. Meso Scale Discovery (MSD) V-PLEX proinflammatory cytokine panel 1 was used to evaluate 10 cytokines (IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, tumor necrosis factor–α, and interferon-γ). The technique uses an electrochemical detection method, which produces a specific signal to evaluate all 10 cytokines in one sample per well. The amount of cytokine was calculated using a standard curve, normalized to the amount of protein in the cell sample, and then further converted to FOC of the untreated control. We also analyzed cytokines and chemokines in vivo. In mouse brain cortex homogenized in tissue homogenization buffer (THB), we normalized samples to equal protein amounts of 140 μg per well. The samples were loaded, and the protocol was followed per manufacturer’s instructions (MSD). qPCR array to measure cytokine mRNA levels HCMEC were seeded 1-day before treatment in six-well plates at a density of 350,000 cells per well. Cells were treated for 24 hours with amyloid plus or minus CA-V inhibitor 4ITP. CM was removed, and then the cells were washed with 1× PBS. Next, total RNA was extracted using miRNeasy mini kit (QIAGEN). Following total RNA purification, samples were quantified with NanoDrop and normalized to 1 μg per sample. This was followed by a 20-μl reverse transcription reaction using SuperScript IV VILO master mix to obtain cDNA. The cDNA samples were further diluted with 150 μl of nuclease-free water. Next, samples were prepared with Taq-man fast advanced master mix and loaded onto the customized Taq-man qPCR array. The targets analyzed were CCL2, CCL3, CXCL1, CXCL2, CSF-2/GM-CSF, CSF3, IL-10, IL-6, IL-12a, IL-12b, IL-4, VCAM-1, SELP/P-selectin, and SELE/E-selectin. Results were analyzed by normalizing the target genes to cyclophilin-D/PPIB and then further calculating the FOC of the nontreated control. Mouse model and treatment All procedures and experiments on animals were conducted and approved in accordance with Temple University Animal Care and Use Committee (IACUC) guidelines (protocol number 4890). The 3xTG AD mouse model was used to evaluate the protective effects of CA-V inhibition in vivo. The 3xTG mouse model carries human APP Swedish, Presenilin 1 (PSEN1 M146V) and tau (MAPT P301L) transgenes under the Thy1.2 neuronal promoter and thus presents both Aβ and tau accumulation within the brain, particularly in the hippocampus and cortex, with reports of Aβ accumulation within the vasculature (75). Mice are anticipated to present with mild amyloidosis around 6 months of age, while tau accumulation is reported later, around 12 months of age (37). Recent studies have demonstrated that vascular abnormalities (38), metabolic deficits, and chronic inflammation may occur in these mice before protein accumulation, particularly more in females than males, translating more accurately to the human progression of disease (38, 48). In this study, we compared male and female WT mice (C7BL/6;129X1/SvJ;129S1/Sv) (N = 20), WT + 4ITP-treated mice (N = 11) 3xTG mice (N = 21), and 3xTG + 4ITP-treated mice (N = 20). 4ITP treatment in 3xTG and WT mice began at 6 months of age until 16 months. CA-V inhibitor, 4ITP, was provided in the mice’ diet at 20 mg/kg per day (100 ppm) in a grain-based diet. This was done in collaboration with bio-Serv. Food was stored in a cool dry place. Body weight and activity levels were monitored and recorded throughout the study. Barnes maze When mice reached 15 months of age, spatial memory was evaluated through the Barnes Maze. The Barnes maze runs over the course of 7 days: day 0, habituation; days 1 to 5 training; and day 6, probe day. On habituation day, mice were put into the middle of the maze with a beaker, a white noise at 80 db for 60 s, after which mice were guided to the escape hole box and left there for 1 min with the noise off. After habituation day (0), every trial day (1 to 5), mice underwent two trials, 2 hours apart from each other, in which the mouse had 3 min to find the escape hole box. The number of mistakes, in addition to the time to find the escape hole (latency), was recorded on each trial. Between mice, the maze was cleaned, and each day the platform was rotated. On probe day (6), the mice are put into the maze, with the escape box removed. The number of mistakes and the time to find escape hole (when their nose is placed into the correct hole) were recorded with ANY-maze software. Fear conditioning To assess cue-induced memory, we used an optimized cue-induced fear conditioning protocol. On day 0 (habituation), mice were introduced to the conditioning chamber for 5 min without any stimulus. The following day (acquisition/training), mice were placed in the same chamber and exposed to a series of tone-foot shock pairings. The foot-shock was set to 0.5 mA/2 s. Following training, mice returned to their home cage. On day 2 (cue-induced testing phase), mice were placed into a different chamber (via changing the walls/floors, odor, and the lighting) and exposed to the same tone without foot shocks. The freezing time in both the acquisition phase (day1) and cue-induced testing phase (day2) were recorded to assess cue-induced memory recall. Freezing time was defined as the absence of all movements other than respiration and was recorded with ANY-maze software to avoid human bias. Novel object recognition test Recognition memory was evaluated using a novel object recognition paradigm. On day 0 (habituation) mice were allowed to explore the empty arena for 5 min. On day 1, trial 1 (training phase) mice were placed into the chamber with two identical objects (towers) and allowed to explore for 10 min. Two hours later, on day 1, trial 2 (testing phase), one of the familiar objects was replaced with a novel (dice) and placed into the chamber. Mice were allowed to explore for 3 min during the testing phase. The location of the novel object was alternated throughout the study to ensure there was no biased in cage side. All sessions were video-recorded, and the time spent focused on each object was captured through ANY-maze software. The % time exploring the novel object during the testing phase was calculated as (time spent exploring the novel object/total exploration time) × 100. Any mouse that did not explore the objects for more than 10 s was excluded from analysis. Rotarod Locomotor function, coordination, and balance were evaluated using an accelerating rotarod procedure across six trials spaced by 20-min intertrial interval. Five mice were run simultaneously on 9.5-cm diameter rods of a land rotarod apparatus. All mice were first given a habituation trial where they were required to stay on the platform for 1 min without falling. Animals were then tested in experimental trials, during which the rod was accelerating steadily from 4 to 40 rpm over the course of 5 min. The latency to fall from the platform was recorded. Open field test Locomotor activity and exploration were evaluated through open field test. This test evaluates the distance traveled by the mice and the preferred behavior of mice to spend more time exploring around the perimeter of the box rather than in the center. More time spent in the center suggests decreased anxiety. To start, mice were placed in the center of a rectangular box and then recorded for one 30-min session. After 30 min of exploration, mice returned to their home cage. The percentage of time spent in the center and the total distanced traveled was recorded. Tissue collection and mouse brain processing Following behavioral analysis, mice were anesthetized and perfused with ice-cold PBS. Each brain was split into hemispheres; one hemisphere was fixed with 4% paraformaldehyde and then further processed for immunohistochemistry analysis. The other hemisphere was dissected into pathological brain areas such as the cortex and hippocampus. After dissection, tissue was flash frozen and stored in −80°C. The tissue for immunohistochemistry was fixed for 2 days in 4% paraformaldehyde at 4°C. Following this, brains were placed in 15% sucrose for 1 day and then 30% sucrose for two. Following fixation and sucrose, brains were washed with PBS and then assembled in a plastic mold with Tissue-Tek O.C.T compound (Thermo Fisher Scientific). Brains were stored until sectioning at −80°C. For WB, hippocampus, and cortex tissue (WT, 3xTG, and 3xTG + 4ITP, N = 6 per group males and females) was homogenized in tissue homogenization buffer (20 mM tris-base, 0.25 M sucrose, 5 mM EDTA, and 1 mM EGTA + HALT protease/phosphatase inhibitors). The protein content of the homogenates was quantified with BCA and then normalized to 30 μg of protein. WB samples were prepared with 6x Laemmli Buffer (Invitrogen). Extraction and quantification of soluble/insoluble amyloid and tau Hippocampal and cortical brain samples were homogenized in THB buffer (as previously explained) and then used to fraction the soluble and insoluble proteins in 3xTG treated and nontreated mice as previously published (25). After homogenization, samples were quantified and normalized to equal levels of protein. To obtain soluble protein containing Aβ and tau monomers, oligomers, and protofibrils a diethylamine (DEA) extraction was performed. In addition, formic acid was used to obtain the insoluble fraction containing aggregated proteins such as Aβ and hyperphosphorylated tau. Both extracted fractions were then used in specific ELISA (Invitrogen) assays to measure the amount Aβ40, Aβ42, and pTau231. Immunohistochemistry Brains were sectioned at 8- and 20-μm thicknesses with a cryostat. Brain sections were collected on positively charged microscope slides and stored at −80°C until used for staining. For immunostaining evaluation, brain sections were fixed in 4% paraformaldehyde for 20 min at room temperature. Next, brain sections were blocked with 10% normal goat serum and 1% BSA solution for 2 hours at room temperature. Then, sections were incubated overnight at 4°C with primary antibodies diluted in 0.3% Triton X-100 (Sigma-Aldrich) blocking solution. The following primary antibodies were used: chicken anti-GFAP (1:1000; Aves, #GFAP), rat anti-CD31 (1:100; BD Pharmingen, 55370), rabbit anti-Albumin (1:500; Invitrogen, PA5-85166), conjugated Alexa Fluor 647–cleaved caspase-3 (1:100; Bioss, bs.20363R-bf647), and conjugated Alexa Fluor 488–NeuN (1:100; Millipore, ABN78A4). After overnight incubation, tissue was washed with 1× PBS and then the species-appropriate Alexa Fluor–conjugated secondary antibody diluted 1:1000 in PBS were placed on the tissue for 2 hours at room temperature. Further, nuclear staining with DAPI (1.5 μg/ml; Invitrogen) was placed on the tissue for 10 minutes at room temperature. Next, sections were washed with PBS, and lastly coverslips were placed with mounting media. Stained sections were imaged with a Nikon Ti2-E fluorescence deconvolution 860 microscope equipped with 340-/380-, 465-/495-, 540-/580-, and 590-/650-nm excitation filters, keeping identical settings within each session and using either 10×, 20×, or 60× zoom objectives. Images (20× and 60×) were acquired with a 0.5-μm Z-stack. To have a consistent examination for each animal, 5 to 10 different images were acquired in the same brain area of interest. Images were deconvolved, and then Z-stack images were merged into two-dimensional maximum signal projections of equal thickness using Fiji. All image analysis was conducted using the object colocalization FL module in HALO Image Analysis Platform version 4.2.6399 (Indica Labs Inc). TUNEL staining DNA fragmentation in vivo was measured through a commercially available TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling) staining kit (Biotium). The protocol was followed as per manufacturer’s instructions. Sections (8 μm) were used to costain DNA fragmentation (channel 647) and total nuclei with DAPI. The number of TUNEL+ nuclei was counted, and then the % of TUNEL+ DAPI was calculated for comparison. Vessel diameter measurement Vessels were stained using CD31 antibody described above using 20-μm-thick tissue sections. After image acquisition and image processing, three images per mouse per brain area (hippocampus and cortex) were selected. Twenty vessels per image were outlined via the annotation tool within HALO software. We then obtained six diameter measurements per vessel. These six values were averaged to obtain each blood vessels average diameter. Each vessel’s diameter was graphed allowing us to visualize the shift in the population of blood vessels within each group. Only measurements for microvessels (<10 μm in diameter) were analyzed.