Timosaponin B-II attenuates hemorrhagic transformation-driven acceleration of alzheimer disease-related pathology after ischemic stroke Authors: Yafen Dong, Xuesong Bai, Ye Chen, Jian Wang, Qiaoyan Cui, Liming Fan, Shushu Li, Yan Qiu Published: 2026 ABSTRACT AbstractHemorrhagic transformation (HT) is a major complication of reperfusion therapy after ischemic stroke, but its impact on Alzheimer disease (AD) progression remains unclear. In this study, we used an early-stage APP/PS1 mouse model and found that HT following transient middle cerebral artery occlusion(tMCAO) aggravated cognitive impairment and promoted AD-like pathological changes, which were markedly attenuated by Timosaponin B-II (TB-II) treatment. TB-II effectively alleviated HT following tMCAO(tMCAO/HT)-induced cognitive and recognition memory deficits in APP/PS1 mice and reduced hippocampal Aβ42 production. In vitro, TB-II pretreatment reversed oxygen-glucose deprivation/reoxygenation (OGD/R) and hemin-induced increases in Aβ42 production and apoptotic propensity, as well as decreases in mitochondrial membrane potential and cell viability in N2a/APPswe cells. Mechanistically, TB-II activated NRF2, which in turn indirectly modulating GSK-3β-mediated APP phosphorylation and negatively regulated BACE1 transcription through binding promoter binding, ultimately suppressing amyloidogenic processing. Together, our findings suggest that TB-II may serve as a potential therapeutic agent against HT-associated acceleration of AD-like pathology after ischemia-reperfusion.Keywords: Hemorrhagic transformation, Alzheimer disease, Acute ischemic stroke, Amyloid-β42, Neurovascular pathology, Timosaponin B-II Introduction Cerebrovascular dysfunction is increasingly recognized as a major contributor to the progression of Alzheimer disease (AD), highlighting that AD is not only a neurodegenerative disorder but also a disease closely associated with vascular pathology [1,2]. Among vascular insults, ischemic stroke is of particular importance, as stroke survivors exhibit a markedly increased risk of developing dementia, whereas patients with AD show a higher incidence of stroke than cognitively normal individuals [3,4]. Although these epidemiological associations have been well established, the mechanisms by which acute cerebrovascular events affect the short-term progression of pre-existing AD pathology remain poorly understood [5]. In contemporary clinical practice, most patients with ischemic stroke receive reperfusion therapies, such as thrombolysis or mechanical thrombectomy. However, these interventions are frequently complicated by hemorrhagic transformation (HT), which is characterized by blood-brain barrier (BBB) disruption, vascular leakage, oxidative stress, and neuroinflammation [6,7]. Such pathological changes generate a neurovascular environment that is distinct from ischemia alone and may be especially relevant in the context of AD, where blood-derived products and oxidative injury are known to promote amyloidogenic processing and neuronal vulnerability. In this context, an important unresolved question is whether HT acts as an acute vascular insult that accelerates the progression of pre-existing AD pathology after stroke. At the molecular level, glycogen synthase kinase-3β (GSK-3β) plays an important role in AD pathogenesis by regulating amyloid precursor protein (APP) phosphorylation and amyloid-β (Aβ) production [8,9]. Nuclear factor erythroid 2-related factor 2 (NRF2) is a master regulator of cellular redox homeostasis and is closely associated with GSK-3β-related signaling under oxidative stress conditions. Although GSK-3β has been more clearly characterized as an upstream negative regulator of NRF2, accumulating evidence suggests that NRF2 activation may also indirectly influence GSK-3β activity by alleviating oxidative stress, thereby affecting amyloidogenic APP processing and neuronal injury [[10], [11], [12], [13]]. These observations support the possibility that NRF2-targeted intervention may modulate GSK-3β-associated pathogenic signaling, particularly in the context of vascular injury and oxidative stress. Timosaponin B-II (TB-II) has been reported to attenuate brain injury in middle cerebral artery occlusion (MCAO) models and to activate NRF2 signaling [14,15], suggesting its potential to protect against vascular and oxidative insults relevant to AD. Based on these observations, we hypothesized that transient MCAO (tMCAO) accompanied by HT may serve as a trigger for the short-term progression of AD after stroke, and that TB-II may mitigate this process, at least in part, through NRF2-related mechanisms. To test this hypothesis, we employed an AD mouse model subjected to tMCAO with subsequent HT and examined the short-term effects of this injury on AD pathology and neuronal damage, together with the therapeutic potential of TB-II in this context. Materials and methods Ethical declarations All animal experiments in this study were approved by the Animal Ethics Committee of Shanghai Pudong New Area People's Hospital (Approval No. 2022-5-50). All procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and all efforts were made to minimize the number of animals used and their suffering. Animal model and experimental design Six-week-old male APP/PS1 transgenic mice and their wild-type (WT) littermates (n = 144 and 48, respectively) were obtained from the Shanghai Model Organisms Center, Inc. (Sanghai, China) and maintained under specific pathogen-free(SPF) conditions (22-24 °C, 50-60% humidity, 12 h light/dark cycle) with ad libitum access to food and water until seven months of age, when APP/PS1 mice exhibited early AD-like phenotypes confirmed by behavioral and biochemical assessments. Mice were randomly assigned to eight groups: Sham (WT), tMCAO/HT(WT), Sham (APP/PS1), tMCAO(APP/PS1), tMCAO/HT(APP/PS1), TB-II(L) + tMCAO/HT(APP/PS1), TB-II(H) + tMCAO/HT(APP/PS1) and TB-II(H) + ML385+tMCAO/HT(APP/PS1). Based on preliminary experiments with the tMCAO/HT model showing a cumulative postoperative mortality rate of approximately 25% within 7 days, the sample size was set at 24 animals per group to minimize animal use while ensuring adequate statistical power. This sample size also accounts for expected postoperative mortality during the 7-day observation period and provides sufficient numbers for all planned analyses. TB-II was administered by tail-vein injection at 10 mg/kg (low dose, L) or 30 mg/kg (high dose, H) using a vehicle composed of dimethyl sulfoxide (DMSO, Sigma-Aldrich, MO, USA), PEG300/400, Tween-80, and saline (final DMSO≤5%, sterile-filtered through a 0.22 μm membrane). ML385 (30 mg/kg) (Sigma-Aldrich) was administered intraperitoneally concurrently with TB-II. Both compounds were administered once daily for six consecutive days, beginning three days before and continuing for three days after surgery [16]. Twelve hours after drug administration, tMCAO was induced as described previously [17]. Mice were anesthetized with 1.5% isoflurane in a gas mixture of 70% N2 and 30% O2 and placed on a heating pad maintained at 37 ± 0.5 °C. Under a stereomicroscope, the left common, internal, and external carotid arteries were carefully exposed. The external and distal common carotid arteries were ligated, and a silicone-coated monofilament was gently inserted through the common carotid artery to occlude the middle cerebral artery. Cortical blood flow was continuously monitored using laser speckle imaging, and successful occlusion was confirmed by a≥70% reduction in regional cerebral perfusion. After 60 min of occlusion, the filament was withdrawn to allow reperfusion. Sham-operated mice underwent the same surgical exposure without filament insertion. HT was induced by stereotaxic injection of 8 μL autologous blood into the left striatum (AP-0.8 mm, ML+2.1 mm, DV-3.1 mm) at 0.5 μL/min; injection was paused after 3 μL for 5 min before the remaining 5 μL containing collagenase (0.2 U/μL) was delivered. The needle was kept in place for 5 min before withdrawal, and sham mice received needle insertion only. Postoperative care included maintaining normothermia, hydration, analgesia, and daily neurological assessment. Neurological deficits were evaluated postoperatively using a Longa-based neurological scoring system, and mice with a score of 1-3 were considered to have successful model establishment and were included in subsequent experiments. On postoperative day 3 and 7, six mice per group were randomly selected and deeply anesthetized with isoflurane (3%). Cerebrospinal fluid (CSF) was first collected via cisterna magna puncture (5-10 μL per mouse), then centrifuged (1000×g, 10 min, 4 °C) and stored at −80 °C for enzyme-linked immunosorbent assay (ELISA) detection of soluble Aβ42 and Aβ40. Following CSF collection, the mice were transcardially perfused with ice-cold PBS. The hippocampi were subsequently dissected, homogenized, and used for ELISA quantification of soluble and total Aβ42. The same hippocampal lysates were further analyzed by Western blotting to determine the expression of phosphorylated GSK-3β and phosphorylated APP. On postoperative day 7, six surviving mice per group were randomly selected for assessment of cognitive function. Following behavioral testing, these mice were deeply anesthetized and transcardially perfused with ice-cold PBS for subsequent immunofluorescence staining of hippocampal Aβ42. Morris water maze (MWM) The MWM test was performed to assess spatial learning and memory in mice. The apparatus consisted of a circular pool (120 cm in diameter) filled with opaque water maintained at 22 ± 1 °C. A circular escape platform (10 cm in diameter) was submerged 1 cm below the water surface and placed in the center of the target quadrant. Extra-maze visual cues were kept constant throughout the experiment. For the acquisition phase, mice were trained for 5 consecutive days, with 4 trials per day. In each trial, mice were released into the water from different starting points in a pseudo-random order, facing the wall of the pool. Each trial lasted for a maximum of 60 s. Mice that located the hidden platform within 60 s were allowed to remain on the platform for 15 s. If a mouse failed to find the platform within 60 s, it was gently guided to the platform and allowed to stay there for 15 s. The inter-trial interval was approximately 15-20 min. Escape latency, swimming speed, and swimming path were recorded using an automated video tracking system. A probe trial was conducted 24 h after the final training session. During the probe trial, the platform was removed, and each mouse was allowed to swim freely for 60 s. The time spent in the target quadrant, the number of crossings over the previous platform location, and the swimming path were recorded and analyzed. All behavioral assessments were performed by investigators blinded to the experimental groups. Novel object recognition (NOR) The NOR test was performed to assess short-term recognition memory in mice. Mice were tested in an opaque acrylic open-field box (40 × 40 × 40 cm). During habituation, mice were allowed to freely explore the empty arena for 5 min per day for 2 consecutive days. In the training session, two identical objects were placed symmetrically in the arena, and each mouse was allowed to explore freely for 5 min. After a 2 h retention interval, one familiar object was replaced with a novel object of similar size but different shape and texture, and mice were reintroduced into the arena for a 5 min test session. The position of the novel object was counterbalanced to avoid side preference. Exploration was defined as directing the nose toward the object at a distance of less than 2 cm and/or sniffing the object; climbing on or sitting against the object was not considered exploratory behavior. Exploration behavior was recorded by a camera positioned above the apparatus. The discrimination index (DI) was calculated as Tnovel/(Tnovel + Tfamiliar). The arena and objects were cleaned with 70% ethanol between trials. Mice with total exploration time<10 s during the test session were excluded from analysis. All behavioral assessments and data analyses were performed by investigators blinded to the experimental groups. Cell culture N2a/APPswe cells, previously established in our laboratory, are N2a cells (ATCC, VA, USA) stably expressing human APP with the Swedish mutation (KM670/671NL) and were used in this study as an in vitro model of AD-like amyloid pathology. This cell line exhibits increased β-secretase cleavage and elevated Aβ40 and Aβ42 production and is therefore widely used for studies of APP processing and Aβ-related pathology. Cells were maintained in high-glucose DMEM (Gibco, MA, USA) with 10% FBS at 37 °C in a humidified incubator with 5% CO2. The cells grew adherently and were passaged by trypsinization when confluence reached approximately 80%. TB-II (HPLC≥98%, Cayman Chemical, MI, USA) was applied at 1 or 5 μg/mL for 2 h before cell modeling. To mimic HT after ischemic stroke, an in vitro model was established using oxygen-glucose deprivation/reoxygenation (OGD/R) combined with hemin stimulation. Cells were incubated in glucose and serum free DMEM under hypoxia (5% CO2, 90% N2) for 4 h, followed by reoxygenation in complete DMEM containing 80 μM hemin (MP Biomedicals, CA, USA) for 48 h. The NRF2 inhibitor ML385 (5 μM) and activator CMF (10 μM) (Sigma-Aldrich, MO, USA) were administered following the same protocol as TB-II. TB-II was dissolved in DMSO to prepare stock solutions and diluted in culture medium before use. For NRF2 knockdown, siRNA targeting NRF2 (5′-GCUUUCAACCCGAAGCACG-3′) was transfected into N2a/APPswe cells using Lipofectamine 2000 (Invitrogen, MA, USA) 24 h before treatment. Hemin was prepared in glucose and serum free DMEM with 1% BSA, protected from light, stored at −20 °C, and diluted in complete medium before use. After treatment, both cells and culture supernatants were collected. Levels of Aβ42, Aβ40, CC90, sAPPα, and sAPPβ, along with Caspase-3/7 activity, cell viability, and proliferative/cell-cycle characteristics, were quantified. In parallel, cells were collected for analyses of NRF2 protein expression, NRF2 nuclear translocation, GSK-3β phosphorylation, and the mRNA expression of HMOX1 and NQO1. CCK-8 assay N2a/APPswe cells transfected with siRNA-NRF2 or siRNA-negative control (NC), as well as non-transfected controls, were trypsinized, seeded into 96-well plates at 5 × 104 cells/well, and cultured overnight. The cells were pretreated with TB-II and then subjected to OGD/R + hemin treatment. Cell viability was evaluated at 24, 48, and 72 h after reoxygenation using the CCK-8 assay. Following the addition of 10 μL CCK-8 reagent (Beyotime, Beijing, China) per well and a 2 h incubation, absorbance at 450 nm was recorded. OD values reflected relative cell viability, and all measurements were performed in triplicate. EdU staining After treatment, N2a/APPswe cells were incubated with 10 μM EdU (Click-iT, MA, USA) for 2 h, followed by fixation, permeabilization, Click-iT staining, and Hoechst 33342(5 μg/mL, Beyotime) counterstaining. Images were obtained under identical settings from at least five random fields per well in three biological replicates. EdU-positive cells were quantified using ImageJ and expressed as the percentage of total nuclei, with a no-EdU control used to define background signal. Caspase activity assay After 24 h of reoxygenation, N2a/APPswe cells were washed with cold PBS and lysed, and subsequently used for caspase activity measurement using the Caspase Multiplex Activity Assay Kit (Abcam, Cambridge, UK) according to the manufacturer's instructions. Equal amounts of clarified lysates were transferred to black 96-well plates and mixed with the caspase reaction buffer containing fluorogenic Caspase-3/7 substrates. Plates were incubated at room temperature in the dark, and fluorescence was measured at the recommended excitation/emission settings. Background signals from substrate-only blanks were subtracted. Caspase-3/7 activity was expressed as relative fluorescence units (RFU). All conditions were assayed in triplicate across at least three independent experiments. Mitochondrial membrane potential assay Mitochondrial membrane potential was evaluated using a JC-1 assay. After 24 h of reoxygenation, N2a/APPswe cells were stained with JC-1 working solution (2 μM, Abcam) according to the manufacturer's instructions, and fluorescence was measured with a microplate reader. The mitochondrial membrane potential was expressed as the ratio of red to green fluorescence. CCCP-treated cells were used as a positive control, whereas untreated cells served as a negative control. Reverse transcription quantitative polymerase chain reaction (RT-qPCR) Total RNA was extracted from cells using standard procedures, and cDNA was synthesized with M-MLV reverse transcriptase (Promega, WI, USA). RT-qPCR was performed with iTaq Universal SYBR Green Supermix (Bio-Rad, CA, USA) on a CFX96 Real-Time PCR System (Bio-Rad, CA, USA). The cycling conditions were: 95 °C for 10 s, 56 °C for 10 s, and 72 °C for 10 s, repeated for 45 cycles. Relative mRNA expression was calculated using the 2−ΔΔCt method, normalized to β-actin for NRF2(Nfe2l2),HMOX1, NQO1 and BACE1 mRNA. Primer sequences were as follows: NRF2, 5′-TGCCCACATTCCCAAACAAGA-3′ (F), 5′-GATATCCAGGGCAAGCGACTCA-3′ (R); HMOX1, 5′-AGCGCAACAAGCAGAACCCAGTC-3′ (F), 5′-CCAGTGAGGCCCATACCAGAAG-3′ (R); NQO1, 5′-CCATGAAGGAGGCTGCTGTAG-3′ (F), 5′-TCTGGAAATGATGGGGTTGAAGT-3′ (R); BACE1, 5′-TTGGGCCACTTAAAAGACGAGAG-3′ (F), 5′-GAGCCGCCTTTTCAGTAGATGGA-3′ (R); β-actin, 5′-GGCCGGGACCTGACAGACTACCTC-3′ (F), 5′-GTCACGCACGATTTCCCTCTCAGC-3′ (R); All reactions were performed following the manufacturer's instructions. Western blotting Total cellular proteins were extracted with M-PER, and tissue lysates with T-PER (Pierce, IL, USA). Protein concentrations were determined by BCA, and equal amounts of protein (15 μg) were separated on 12% SDS-PAGE and transferred to nitrocellulose membranes. After blocking in 5% non-fat milk/TBST for 1 h, membranes were incubated overnight at 4 °C with primary antibodies against BACE1 (1:400), p-GSK-3β (Ser9, 1:350), p-GSK-3β (Tyr216, 1:400), p-APP (Thr668, 1:500), and β-actin (1:1500) (Abcam). HRP-conjugated anti-rabbit secondary antibody (1:3000, Abcam) was applied for 1 h, and bands were visualized using ECL (Pierce) and captured on X-ray film. Densitometry was performed in ImageJ, with β-actin used for normalization. Enzyme-linked immunosorbent assay (ELISA) The concentrations of soluble and insoluble Aβ40 and Aβ42 in mouse hippocampus, CSF, and cell culture supernatants were measured using sandwich ELISA kits (Thermo Fisher Scientific, CA, USA) following the manufacturer's instructions. For tissue samples, hippocampi were homogenized in ice-cold PBS (pH = 7.4) with protease inhibitors and centrifuged at 100,000×g for 60 min at 4 °C. The supernatant was used for soluble Aβ40/42 measurement. The pellet was extracted with 5 M guanidine-HCl for 2 h at 4 °C and centrifuged, and the resulting supernatant was used to quantify total Aβ. CSF was collected by cisterna magna puncture, centrifuged at 1000×g for 10 min at 4 °C, and the supernatant was analyzed for soluble Aβ. For cultured cells, both medium and lysates were collected. Culture supernatants were used to measure Aβ40, Aβ42, sAPPα, and sAPPβ, and were pretreated with 5 M guanidine-HCl when total Aβ was required. Absorbance was read at 450 nm and concentrations calculated from standard curves. Cell lysates prepared in RIPA buffer were centrifuged at 14,000×g for 15 min, and the supernatant was used for C99 quantification. Results were expressed as pg/mL or pg/mg protein). Immunofluorescence staining Mice were deeply anesthetized and perfused with PBS followed by 4% PFA. Brains were post-fixed overnight at 4 °C, cryoprotected in 30% sucrose, and coronally sectioned (20 μm). After antigen retrieval with 70% formic acid (5 min), sections were permeabilized and blocked in 5% normal serum with 0.3% Triton X-100. Slices were incubated overnight at 4 °C with rabbit anti-Aβ1-42 (Thermo Fisher, 1:500), followed by Alexa Fluor-conjugated secondary antibody for 1 h at room temperature. Nuclei were counterstained with Hoechst and sections mounted in antifade medium. Images were acquired under identical microscope settings, and Aβ42 deposition was quantified in ImageJ as the percentage of Aβ-positive area in the hippocampus. Immunofluorescence was also performed to evaluate the effects of TB-II treatment on NRF2 expression and nuclear translocation, as well as GSK-3β phosphorylation at Tyr216, in N2a/APPswe cells. Primary antibodies were purchased from Santa Cruz Biotechnology (TX, USA) and were used at dilutions of 1:300 and 1:450, respectively. Chromatin immunoprecipitation (ChIP) First, N2a/APPswe cells were transfected with a C-terminal 3×FLAG-tagged NRF2 plasmid (pcDNA-FLAG-NRF2). After 24 h, the cells were re-seeded into 6-well plates at a density of 1 × 105 cells per well and allowed to recover overnight under standard culture conditions. The following day, the medium was completely replaced with fresh complete medium, and TB-II was added at a final concentration of 5 μg/mL. The cells were then further incubated for an additional 24 h under normal conditions. Cells were cross-linked with 1% formaldehyde for 10 min, quenched with 125 mM glycine, and lysed. Chromatin was sonicated to 200-500 bp and immunoprecipitated overnight at 4 °C with an anti-FLAG antibody (Abcam), followed by protein A/G bead capture. After reversing crosslinks at 65 °C for 6 h, DNA was purified by using QIAquick PCR Purification Kit (Qiagen, Venlo, Netherlands). Purified DNA was then used as template for qPCR amplification with BACE1 gene promoter-specific primers. The primer pair (forward, 5′-TCATCTCAATGAGTATGCT-3′; reverse, 5′-TTGATGGTTTCCTGTC-3′) generated a 124-bp product. ChIP-qPCR data were analyzed using the percent input method. Briefly, Ct values of immunoprecipitated DNA were normalized to the corresponding input DNA after correction for the input dilution factor, and enrichment was expressed as % input. A higher % input value indicated stronger binding of the target protein to the BACE1 promoter region. Luciferase assays N2a/APPswe cells were transfected with the pGL3-pro-BACE1 plasmid containing the mouse BACE1 promoter upstream of the luciferase reporter. These transfected cells were pretreated with TB-II (5 μg/mL) or DMF (10 μM) for 2 h before treated with OGD/R+ hemin (80 μM). After 24 h of reoxygenation, Luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega, WI, USA) according to the manufacturer's instructions and normalized to Renilla luciferase to control for transfection efficiency. Statistical analysis Data are presented as mean ± SD from at least three independent experiments, each performed in biological triplicate. Statistical analyses were performed using SPSS 20.0(IBM Corp., NY, USA) and GraphPad Prism 7.0(GraphPad Software, CA, USA). Data normality and homogeneity of variance were assessed prior to analysis. Two-group comparisons were conducted using two-tailed Student's t-tests, and multiple-group comparisons using one-way ANOVA followed by Tukey's post hoc test. A p-value<0.05 was considered statistically significant. All experiments were performed in biological triplicates. Results Hemorrhagic transformation after tMCAO acutely exacerbates cognitive impairment and AD-like pathology in APP/PS1 mice Behavioral performance was evaluated by the MWM on day 7 after tMCAO/HT. APP/PS1 mice in the Sham (APP/PS1) group already showed more disoriented swimming paths than WT mice of Sham (WT) group, and the combination of tMCAO/HT further increased path irregularity, with APP/PS1 mice displaying the most chaotic trajectories (Fig. S1A). Quantitatively, WT mice of tMCAO/HT(WT) group exhibited modest but non-significant increases in swimming distance, escape latency, and travel time compared with WT mice of Sham (WT) group (all p < 0.05). In contrast, APP/PS1 mice in the tMCAO/HT (APP/PS1) group showed significant increases in all three parameters relative to Sham (APP/PS1) group (p < 0.05 or p < 0.01) (Fig. 1A). These results indicate that HT after tMCAO aggravates cognitive impairment in early-stage APP/PS1 mice, suggesting an exacerbation of AD-like pathological progression under this condition. Cognitive Performance, Aβ Production, and GSK-3β Phosphorylation in APP/PS1 Mice After tMCAO/HT (A) Quantitative analysis of MWM performance, including total swimming distance, escape latency, and travel time. (B) ELISA determination of soluble and insoluble Aβ42 and Aβ40 levels, as well as the Aβ42/Aβ40 ratio, in hippocampal tissues. (C) Western blot analysis of GSK-3β phosphorylation at Ser9 and Tyr216, and APP phosphorylation at Thr668, in hippocampal tissues. Data are presented as mean ± SD (n = 6 per group). Statistical analysis was performed by one-way ANOVA followed by Tukey's multiple comparisons test. ∗p < 0.05, ∗∗p < 0.01. ELISA analysis showed that, in the early stages of AD, APP/PS1 mice in the tMCAO/HT (APP/PS1) group exhibited significant increases in total hippocampal Aβ42 and Aβ40 levels (including soluble and insoluble fractions) compared with Sham (APP/PS1) mice (p < 0.01 or p < 0.05 on Day 3; all p < 0.01 on Day 7). In contrast, the soluble Aβ42/Aβ40 ratio was significantly decreased in the tMCAO/HT(APP/PS1) group (p < 0.05 on Day 3; p < 0.01 on Day 7). Moreover, the insoluble Aβ42/Aβ40 ratio in the hippocampus was significantly increased in APP/PS1 mice subjected to tMCAO/HT compared with Sham (APP/PS1) mice (p < 0.05 on Day 3; p < 0.01 on Day 7). In WT mice, however, neither total Aβ42 and Aβ40 levels nor soluble or insoluble Aβ42/Aβ40 ratios differed significantly between the tMCAO/HT and Sham groups at either Day 3 or Day 7 (all p > 0.05) (Fig. 1B). Taken together, these findings indicate that HT after tMCAO increases Aβ40 and Aβ42 levels in APP/PS1 mice. Moreover, the decreased soluble Aβ42/Aβ40 ratio, together with the increased insoluble Aβ42/Aβ40 ratio and elevated total Aβ42 levels, suggests that Aβ42 is more likely to accumulate in insoluble aggregated or deposited forms in the hippocampus of APP/PS1 mice. This shift may contribute to increased neurotoxicity and cognitive impairment, as reflected in the MWM results. Western blot analysis revealed that, compared with APP/PS1 mice in the Sham (APP/PS1) group, those in the tMCAO/HT (APP/PS1) group exhibited a significant decrease in p-GSK-3β (Ser9) and a marked increase in p-GSK-3β (Tyr216) in the hippocampus (p < 0.01). In addition, phosphorylation of APP at Thr668, an event associated with the production of Aβ42 and Aβ40, was significantly elevated (Fig. 1C). ELISA analyses consistently showed that tMCAO alone did not significantly alter total hippocampal Aβ42 levels in APP/PS1 mice (p > 0.05), whereas the tMCAO/HT group displayed a marked increase in total Aβ42 (p < 0.01) (Fig. 2A). These findings indicate that HT after tMCAO, rather than tMCAO itself drives the elevated Aβ42 in the brain of AD mice. Evaluation of the effects of TB-II on Aβ deposition and cognitive performance in APP/PS1 mice with HT following tMCAO (A) ELISA analysis of total hippocampal Aβ42 levels in APP/PS1 mice subjected to different treatments. Aβ42 levels were normalized to total protein and expressed as pg/mg total protein. (B) MWM analysis of cognitive performance, including total swimming distance, travel time, and escape latency. (C) NOR analysis of recognition memory, presented as the discrimination index. (D) ELISA quantification of Aβ42 and Aβ40 levels in CSF on postoperative days 3 and 7. (E) Immunofluorescence analysis of hippocampal Aβ42 deposition across treatment groups. Left, representative immunofluorescence images. Right, quantification of fluorescence intensity. Relative fluorescence intensity was normalized to the Sham (APP/PS1) group and expressed as fold change. The blue signal represents the nuclei stained with Hoechst. Data are presented as mean ± SD (n = 6 per group). Statistical analysis was performed by one-way ANOVA followed by Tukey's multiple comparisons test. ∗p < 0.05, ∗∗p < 0.01. TB-II attenuates tMCAO/HT-induced cognitive impairment in early-stage APP/PS1 mice via NRF2 activation MWM testing showed that low-dose TB-II reduced total swimming distance, travel time, and escape latency in APP/PS1 mice after tMCAO/HT, with greater improvements observed in the high-dose TB-II group (p < 0.01). However, these benefits were largely abolished by ML385, the specific NRF2 inhibitor (p < 0.01) (Fig. 2B). Results of the NOR test showed that APP/PS1 mice of TB-II(L) + tMCAO/HT(APP/PS1) mice exhibited a higher DI than those of tMCAO/HT(APP/PS1) group, although the difference was not statistically significant (p > 0.05). The DI was further increased in TB-II(H) + tMCAO/HT(APP/PS1) group (p < 0.01). Compared with the TB-II (H) + tMCAO/HT(APP/PS1) group, the DI was significantly reduced in TB-II (H) + ML385+tMCAO/HT(APP/PS1) group (p < 0.01) (Fig. 2C). These results indicate that TB-II can improve the ability of tMCAO/HT(APP/PS1) mice to discriminate novel objects, and that this effect is associated with the activation of NRF2. TB-II reduces Aβ42 levels and deposition in early-stage APP/PS1 mice after tMCAO/HT ELISA revealed that CSF Aβ42 and Aβ40 levels were decreased on postoperative day 3 in APP/PS1 mice of TB-II(L) + tMCAO/HT(APP/PS1), with greater reductions in TB-II(H) + tMCAO/HT(APP/PS1) group (p < 0.01 or p < 0.05), and both were significantly lowered by day 7 (all p < 0.01). However, co-administration of the NRF2 inhibitor ML385 largely reversed these decreases, leading to a significant rebound in Aβ42 and Aβ40 levels (p < 0.01 or p < 0.05 on day 7) (Fig. 2D). Immunofluorescence further showed that tMCAO/HT markedly increased hippocampal Aβ42 deposition in APP/PS1 mice (p < 0.01), whereas high-dose TB-II significantly reduced this accumulation (p < 0.01) (Fig. 2E). TB-II reduces Aβ42 generation and Aβ42-Dependent cytotoxicity in N2a/APPswe cells subjected to OGD/R and hemin treatment APP-processing ELISAs further showed that OGD/R + hemin elevated intracellular C99 and supernatant sAPPβ while reducing supernatant sAPPα. TB-II normalized all three indices, and these effects were blocked by NRF2 inhibition. ELISA further showed that OGD/R+ hemin markedly increased the Aβ42/Aβ40 ratio in the cell culture supernatant of N2a/APPswe cells (p < 0.01), whereas TB-II significantly reduced this change (p < 0.01). ML385 largely reversed the effect of TB-II (Fig. 3A). The RT-qPCR results showed that 48 h after siRNA transfection, the NRF2 mRNA level in N2a/APPswe cells was significantly reduced in the siRNA-NRF2 group compared with the control group (p < 0.01), whereas no significant change was observed in the siRNA-NC group (p > 0.05) (Fig. S1B). CCK-8 assays demonstrated that TB-II restored the proliferation loss induced by OGD/R+ hemin, whereas siRNA-NRF2 negated this improvement (Fig. S1C). Caspase-3/7 activity assays confirmed that TB-II reduced OGD/R + hemin-induced apoptosis, and this effect was reversed by NRF2 silencing (Fig. 3B). EdU staining showed that TB-II increased the proportion of S-phase cells suppressed by OGD/R + hemin, while NRF2 silencing again counteracted this effect (Fig. 3C). JC-1 staining revealed parallel changes in mitochondrial membrane potential, consistent with the alterations in cell-cycle progression across groups (Fig. 3D). Effects of TB-II Pretreatment on Aβ Production, APP Processing, Cell Proliferation, and Mitochondrial Function in N2a/APPswe Cells Under OGD/R and Hemin Treatment (A) ELISA analysis of C99 levels in cell lysates, as well as the sAPPβ/sAPPα ratio and Aβ42/Aβ40 ratio in the supernatant of N2a/APPswe cells across treatment groups. (B) Caspase-3/7 activity assay by using a luminescence-based method. Relative light units were normalized to controls. (C) EdU incorporation assay. Left, EdU-positive nuclei were observed in five randomly selected microscopic fields per group and quantified using ImageJ. Right, the number of EdU-positive nuclei was counted and compared across groups. (D) JC-1 staining. Red/green fluorescence intensity ratios were calculated from three random fields per experiment using ImageJ and expressed in arbitrary units. Data are presented as mean ± SD (n = 3). Technical replicates(n = 3) were included for the CCK-8 and ELISA assays. Statistical analysis was performed by one-way ANOVA followed by Tukey's multiple comparisons test. ∗p < 0.05, ∗∗p < 0.01. TB-II suppress BACE1 and modulate GSK-3β Ser9 phosphorylation through NRF2 activation Luciferase assays revealed that OGD/R + hemin markedly reduced intracellular luciferase activity in N2a/APPswe cells transfected with pGL3-pro-BACE1(p < 0.01, vs. pGL3-pro-BACE1 group). In contrast, both pGL3-pro-BACE1+OGD/R + hemin + DMF and pGL3-pro-BACE1+OGD/R + hemin + TB-II treatments significantly increased luciferase activity (p < 0.01, vs. pGL3-pro-BACE1 group + OGD/R + hemin group) (Fig. 4A). ChIP analysis further demonstrated that TB-II enhanced NRF2 binding to the BACE1 promoter (Fig. 4B). Immunofluorescence staining showed that TB-II treatment resulted in a significant increase in NRF2 expression along with a marked enhancement of GSK-3β Tyr216 phosphorylation compared with the vehicle group (p < 0.01). In addition, immunofluorescence analysis indicated that, compared with the vehicle group, NRF2 protein in the TB-II group showed a marked translocation from the cytoplasm to the nucleus (p < 0.01)(Fig. 4C).RT-qPCR analysis showed that, compared with the vehicle group, TB-II treatment significantly increased the mRNA levels of HMOX1 and NQO1 in N2a/APPswe cells (p < 0.01) (Fig. 4D). RT-qPCR analysis also confirmed that OGD/R + hemin significantly upregulated BACE1 mRNA and protein levels in N2a/APPswe cells (p < 0.01, vs. Control group), whereas both DMF and TB-II treatments markedly reduced BACE1 expression in OGD/R+ hemin treated cells (p < 0.01, vs. OGD/R + hemin group) (Fig. 4E and F). Collectively, these results indicate that TB-II counteracts OGD/R + hemin induced BACE1 upregulation by increasing NRF2 expression and enhancing its transcriptional repression activity, with effects comparable to those of the NRF2 activator DMF. Analysis of NRF2-mediated regulation of BACE1 transcription and GSK-3β phosphorylation in N2a/APPswe cells exposed to OGD/R + hemin following TB-II or DMF pretreatment (A) Luciferase reporter assay. Relative luciferase activity was normalized to Renilla control values and expressed in arbitrary units. (B) ChIP analysis. Top, immunoblot verification of the target protein. Bottom, qPCR analysis of changes in the amount of BACE1 promoter DNA bound by NRF2, with input used as the reference control (%). (C)Immunofluorescence analysis of NRF2 expression and GSK-3β phosphorylation at Tyr216 in N2a/APPswe cells. The red fluorescence signal represents NRF2, the green fluorescence signal indicates GSK-3β phosphorylated at Tyr216, and the blue signal marks nuclei stained with Hoechst. Fluorescence intensity was quantified from five randomly selected microscopic fields per group using ImageJ and expressed in arbitrary units. The nuclear-to-total cellular NRF2 ratio was analyzed to evaluate the change in NRF2 nuclear localization. (D) RT-qPCR analysis of the mRNA expression levels of HMOX1 and NQO1 in N2a/APPswe cells treated with TB-II or vehicle. β-Actin served as the internal control. (E) RT-qPCR and (F)Western blot analyses of BACE1 mRNA and protein expression in N2a/APPswe cells subjected to OGD/R + hemin following TB-II or DMF pretreatment. β-Actin served as the internal control, and relative expression levels were normalized to the Control group. Data are presented as mean ± SD (n = 3 independent experiments per group). ∗p < 0.05, ∗∗p < 0.01. Discussion Recent studies increasingly support a close link between cerebrovascular pathology and AD, with stroke recognized as an important accelerator of AD progression [18]. Large cohort studies show that post-stroke cognitive decline often begins with an acute stepwise deterioration followed by a persistently accelerated trajectory, and that early post-stroke cognitive impairment strongly predicts post-stroke dementia. Mechanistically, stroke can induce prolonged BBB disruption, permitting peripheral inflammatory mediators to enter the brain and sustain neuroinflammation, while simultaneously impairing CSF-interstitial fluid exchange and glymphatic clearance [19]. These vascular and metabolic disturbances may reduce Aβ and promote amyloidogenic APP processing, thereby facilitating AD-related neurodegenerative changes [20]. In line with these mechanisms, imaging and biomarker studies have shown that white-matter lesions, microbleeds, and elevated plasma neurofilament light chain levels are associated with cognitive decline and accelerated AD-related pathology after stroke [21]. HT typically arises during thrombolytic or endovascular reperfusion and is characterized by BBB disruption, vascular leakage, and robust inflammatory and oxidative responses. These features closely resemble neurovascular unit dysfunction in AD, suggesting that HT after ischemic stroke may create a permissive microenvironment that accelerates AD-related pathological processes [22]. Clinically, the impact of HT may be particularly pronounced in patients with pre-existing cerebrovascular vulnerability. Older age, greater infarct burden, coagulopathy, cerebral small-vessel disease (CSVD), and especially cerebral amyloid angiopathy (CAA) are all associated with an increased risk of HT. Because amyloid-laden and structurally fragile vessels are more prone to hemorrhagic injury, patients with CAA or extensive cerebral microbleeds (CMBs) may be especially susceptible to hemorrhagic complications after thrombolysis, supporting a “double-hit” model in which pre-existing AD-related vascular fragility is further aggravated by reperfusion stress [23].Mechanistically, prolonged endothelial injury, biphasic blood–brain barrier (BBB) disruption, iron deposition, and oxidative stress may sustain neuroinflammation and impair protein clearance, thereby further accelerating AD-related pathology [[23], [24], [25], [26]]. Our animal study suggests that tMCAO/HT is an important driver of the short-term acceleration of AD-related pathology. APP/PS1 mice showed greater tMCAO/HT-associated cognitive impairment than WT mice, indicating increased vulnerability of AD-prone neural circuits to this combined insult. In APP/PS1 mice, tMCAO alone caused only modest changes in hippocampal Aβ42 levels, whereas the addition of HT markedly increased Aβ42, suggesting that hemorrhagic transformation after ischemic stroke creates a microenvironment favorable for rapid amyloid accumulation. At the molecular level, tMCAO/HT increased both soluble and insoluble Aβ42 and Aβ40 in the hippocampus, with a particularly marked increase in the Aβ42/Aβ40 ratio, while reducing the soluble Aβ42/Aβ40 ratio, pointing to disrupted amyloid homeostasis with preferential Aβ42 accumulation. Together, these findings support a model in which HT after tMCAO accelerates AD-related pathology, likely by enhancing Aβ accumulation and altered amyloid dynamics. Nevertheless, because this study was limited to the acute phase (up to 7 days post-stroke), it remains unclear whether these Aβ alterations reflect a transient injury response or sustained acceleration of AD-like pathology. In addition, HT was modeled by tMCAO followed by stereotaxic injection of collagenase and autologous blood at reperfusion, with the aim of reproducing key features of hemorrhagic injury after ischemia-reperfusion, including vascular damage, BBB breakdown, and blood extravasation into ischemic tissue. Thus, this paradigm is better regarded as an HT-relevant injury model rather than a full recapitulation of spontaneous clinical HT after reperfusion therapy. Although it is useful for mechanistic investigation, its translational interpretation should be made with caution, and further validation in more clinically relevant reperfusion-associated HT models is warranted. GSK-3β is a key mediator linking ischemia-reperfusion injury to AD-related amyloid pathology. Previous studies have shown that ischemia-reperfusion robustly activates GSK-3β through oxidative stress and inflammatory signaling, thereby promoting APP Thr668 phosphorylation, amyloidogenic processing, and Aβ production [27,28]. In the present study, the addition of HT to tMCAO markedly intensified oxidative stress and inflammation, reduced inhibitory Ser9 phosphorylation, and enhanced GSK-3β activity. Consistently, HT after tMCAO in AD mice decreased hippocampal Ser9 phosphorylation while increasing Tyr216 phosphorylation of GSK-3β, accompanied by elevated APP Thr668 phosphorylation. In parallel, iron overload and hemoglobin breakdown within hemorrhagic regions may further amplify reactive oxygen species generation and reinforce a positive feedback loop between oxidative stress and GSK-3β-driven APP dysregulation [29,30]. Together, these findings support an oxidative stress driven GSK-3β overactivation-APP phosphorylation axis as an important mechanism underlying HT-mediated short-term acceleration of Aβ pathology [31]. Given the multifactorial nature of AD and the limited efficacy of current Aβ-targeting therapies, upstream regulatory pathways are of particular interest. NRF2 is a master regulator of antioxidant and anti-inflammatory responses and is increasingly recognized as a critical link between ischemic injury and AD pathology. NRF2 activity is markedly reduced in AD brains and inversely correlates with Aβ burden. Experimental studies show that NRF2 activation suppresses Aβ production and neuroinflammation, whereas NRF2 deficiency exacerbates amyloid pathology and cognitive impairment [32]. In ischemic stroke and reperfusion models, NRF2 upregulation limits oxidative stress, preserves BBB integrity, and reduces hemorrhagic injury [33]. These observations suggest that HT may accelerate AD pathology, at least in part, by suppressing NRF2 signaling, thereby aggravating oxidative stress and facilitating GSK-3β-associated APP amyloidogenic processing and Aβ42 accumulation. Collectively, this supports a mechanistic framework in which HT-induced oxidative stress, impaired NRF2 signaling, and GSK-3β overactivation cooperate to link hemorrhagic reperfusion injury with accelerated AD-related neurodegeneration. Notably, the relationship between NRF2 and GSK-3β is likely more complex than a simple linear inhibitory cascade. Although GSK-3β has been more clearly characterized as an upstream negative regulator of NRF2, current evidence suggests that the reciprocal influence of NRF2 on GSK-3β is more likely to occur indirectly, primarily through restoration of intracellular redox balance rather than through direct molecular repression [[10], [11], [12], [13]]. This distinction is important for interpreting our findings. In the present study, our focus was not to establish a direct NRF2-to-GSK-3β regulatory mechanism, but rather to determine whether NRF2-targeted intervention is sufficient to modulate GSK-3β activity and its downstream APP amyloidogenic processing under HT-related oxidative conditions. From this perspective, the biological and therapeutic relevance of NRF2 does not depend on direct inhibition of GSK-3β per se. Instead, NRF2 activation may function as an upstream modulatory event that reshapes the oxidative microenvironment, thereby limiting persistent GSK-3β overactivation and attenuating downstream amyloidogenic signaling. This potentially overlooked feedback interplay may be particularly relevant in HT, where abrupt and severe oxidative stress could otherwise drive a self-reinforcing pathogenic cycle that accelerates early AD related changes after ischemic stroke. Beyond its modulatory effect on the oxidative stress-driven GSK-3β/APP axis, our study further suggests that NRF2 may regulate Aβ42 production through an additional and potentially more direct pathway involving transcriptional repression of BACE1. As the rate-limiting β-secretase in amyloidogenic APP processing, BACE1 is a critical determinant of Aβ generation. We found that NRF2 binds to the BACE1 promoter and normally suppresses its transcription, whereas OGD/R plus heme exposure disrupted this repression in N2a/APPswe cells. Both the NRF2 agonist DMF and TB-II reversed this effect, indicating that HT-related oxidative stress may enhance BACE1 expression, at least in part, through inhibition of NRF2. These findings suggest that, in addition to promoting GSK-3β-associated APP dysregulation, HT may also accelerate Aβ42 overproduction by relieving NRF2-dependent repression of BACE1. Collectively, this identifies NRF2 as a central regulator of HT-driven AD-like pathology and supports TB-II as a promising therapeutic candidate capable of targeting this pathogenic cascade. Although disease-modifying therapies for AD have advanced in recent years, no treatment currently addresses the acceleration of AD pathology associated with stroke or hemorrhagic injury, which remains an important unmet clinical need [34]. In this context, the present study provides the first systematic evidence that HT after ischemic stroke can rapidly accelerate AD-related pathology in the short term. By integrating in vivo and in vitro approaches, we established a coherent mechanistic framework in which NRF2 acts as a central protective regulator that limits Aβ accumulation by suppressing oxidative stress, restraining GSK-3β activation, and repressing BACE1 transcription. Under HT-mimicking conditions or after HT following tMCAO, NRF2 signaling was markedly suppressed, leading to aberrant activation of the GSK-3β-BACE1 axis, a shift of APP processing toward the β-secretase-dependent pathway, and excessive Aβ42 production. Together, these findings identify NRF2 dysfunction as a key mechanistic link between HT-related injury and accelerated AD-like amyloid pathology. Building on our previous finding that TB-II directly binds NRF2 and prevents its ubiquitin-mediated degradation [15], the present study further demonstrates that TB-II can effectively counteract this pathogenic process. In both in vivo and in vitro models, TB-II reduced Aβ42 accumulation, alleviated injury-related functional impairment, and restored APP processing toward a less amyloidogenic state. These protective effects were largely dependent on NRF2, as NRF2 silencing substantially abolished the benefits of TB-II. Collectively, our findings suggest that TB-II is a promising therapeutic candidate for limiting HT-driven short-term acceleration of AD-related pathology through NRF2-associated modulation of amyloidogenic signaling. Several limitations should be acknowledged. First, although our findings suggest that NRF2-targeted intervention modulates GSK-3β-associated amyloidogenic signaling under HT-related oxidative stress, a PI3K/Akt-dependent mechanism cannot be conclusively established because PI3K/Akt activation was not directly assessed. Second, this study mainly focused on Aβ-related pathology and did not evaluate other relevant processes, including tau-related changes, glymphatic dysfunction, and microglial activation. Third, the in vivo and in vitro models used here do not fully recapitulate the complexity of clinically reperfusion-associated HT, particularly lacking key features such as blood-brain barrier disruption, inflammatory cell infiltration, and thrombin-mediated signaling. Moreover, the in vitro OGD/R plus hemin model primarily reflects oxidative stress and hemoglobin breakdown product toxicity rather than the full spectrum of HT pathology. In addition, the one-week observation period precluded assessment of longer-term pathological evolution. Finally, although TB-II appeared to suppress Aβ42 production through NRF2-dependent regulation of both GSK-3β and BACE1, the relative contribution and precise relationship of these downstream pathways remain to be clarified. Nevertheless, these limitations do not fundamentally alter the main conclusion that HT can accelerate AD-related pathology in the short term and that TB-II may mitigate this process through NRF2-associated anti-amyloidogenic mechanisms. Despite these limitations, the present study provides initial evidence that HT after ischemic stroke can rapidly accelerate AD-related pathology in the short term, with NRF2 emerging as a key protective regulator. By restoring NRF2 activity, TB-II may counteract this process and represent a potential therapeutic strategy for stroke- or hemorrhage-associated AD progression. In summary, this study provides the first systematic evidence that hemorrhagic transformation after ischemic stroke can rapidly accelerate AD-related pathology in the short term, and identifies NRF2 as a key protective regulator in this process. By restoring NRF2 activity, TB-II may interrupt HT-driven amyloidogenic progression, highlighting a potential therapeutic strategy for early intervention against stroke-associated exacerbation of AD pathology. Consent for publication All contributors give consent for unrestricted publication of this work. CRediT authorship contribution statement Yafen Dong: Conceptualization, Funding acquisition, Investigation, Methodology. Xuesong Bai: Data curation, Funding acquisition, Investigation, Project administration, Writing – original draft. Ye Chen: Investigation, Visualization. Jian Wang: Data curation, Investigation, Software. Qiaoyan Cui: Formal analysis, Investigation, Supervision. Liming Fan: Formal analysis, Investigation, Software, Validation. Shushu Li: Supervision, Validation. Yan Qiu: Conceptualization, Funding acquisition, Resources, Writing – review & editing. Declaration of competing interest The authors declare that they have no competing financial interests or personal relationships that could have influenced the work reported in this manuscript. Acknowledgements This study was supported by the Medical Discipline Construction Program of Shanghai Pudong New Area Health Commission (the Key Weak Disciplines Program) (Grant No. PWZbr2022-13). Acdamic Leaders Training Program of Pudong Health Bureau of Shanghai (Grant No. PWRD2025-13). National Natural Science Foundation of China (82474107). The Outstanding Leaders Training Program of Shanghai Pudong New Area Health Commission (Grant No. PWRl 2020-08). The Young Medical Talents Training Program of Shanghai Pudong New Area Health Commission (PWRq2024-44). The Scientific Research Program of Shanghai Pudong New Area Health Commission (the General Program) (Grant No.PW2023A-06). Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbrep.2026.102654. Supplementary data The following are the Supplementary data to this article: Multimedia component 1 MWM Swimming Trajectories, Validation of NRF2 Silencing, and Analysis of APP-Processing Intermediates (A) Representative swimming trajectories in the MWM on postoperative day 7. (B) Evaluation of NRF2 silencing efficiency in vitro. N2a/APPswe cells were transfected with siRNA-NRF2 or NC, and total RNA was extracted 48 h later. The mRNA expression level of NRF2 in each group was determined by RT-qPCR. (C) CCK-8 assay for cell viability. Absorbance at 450 nm was measured and normalized to control values from three independent experiments. (D) ELISA analysis of APP-processing intermediates, including sAPPβ and sAPPα, in the culture supernatants of N2a/APPswe cells across treatment groups. Data are presented as mean ± SD (n = 3). Statistical analysis was performed by one-way ANOVA followed. ∗p < 0.05, ∗∗p < 0.01. Data availability Data will be made available on request. References Associated Data Supplementary Materials Multimedia component 1 Data Availability Statement Data will be made available on request.