Model mice (Stock Number: 034830-JAX, RRID:MMRRC_034830-JAX) with APP, PS1 and Tau mutations were purchased from the Jackson Laboratory (Bar Harbor, ME, USA), and were housed and bred in the SPF class animal house of the Key Laboratory of Basic Pharmacology, Ministry of Education, Zunyi Medical University. During the breeding process, the offspring mice were genotyped according to the method provided by Jackson Laboratory, and the APP, PS1 and Tau genes containing mutations were identified. Non-transgenic B6129SF2/J wild-type (WT) mice (Stock Number: 101045, RRID:IMSR_JAX:101045) were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). Wild-type mice and model mice were bred in the same environment and at the same time period. All mice were fed and watered freely in a 12-h day/night cycle with relatively constant room temperature (18–25°C) and humidity (40–70%). This experiment complied with the guidelines related to experimental animal welfare and ethics.

The Morris water maze apparatus (TopScan Version 3.00, CleverSys Inc) consisted of a white circular water basin with a diameter of 120 cm and a height of 50 cm, a hidden platform and an automatic recording and analysis system. The water basin was divided into four quadrants: 1, 2, 3 and 4, and the hidden platform was always fixed in the third quadrant. During the experiment, the mice were placed into the water facing the basin wall, and the first 1 to 4 days were used for positioning navigation test, and the time to find the hidden platform within 60 s was recorded. If the mice did not find the hidden platform within the specified time, the latency period of escape was 60 s and the mice were guided to the platform for 15 s. On day 5, the platform was removed. The trajectory of the mice within 60 s and their residence time in the target quadrant were recorded.

ZH-300 fatigue rotating rod instrument was purchased from Anhui Zhenghua Biologic Apparatus Facilities Co., Ltd. The rotating bar experiment was used to test the motor function of mice. The mice were first acclimatized. The experiment was started at 10 rpm for each mouse, and the rpm was increased by 5 rpm every 60 s. The upper limit of the test was 300 s. If the movement time was greater than 300 s, it was recorded as 300 s. The mice were trained for 3 days, once a day, and the average of the 3 times was taken for statistical purposes.

The open field (TopScan Version 3.00, CleverSys Inc) was composed of four experimental boxes of the same size, divided into the middle area and the peripheral area. Mice were placed from a fixed position and the camera simultaneously tracked the movement trajectory of four mice in each experimental box for 5 min. To avoid any effect on the subsequent experiments, 75% alcohol was used to wipe the residues of feces and urine after each mouse was tested.

Nesting experiments are commonly used to assess daily activities in cognitively impaired mice, and increasing age and damage to the hippocampus and other parts of the body can lead to a gradual decrease in nesting behavior. Therefore, we referred to the literature (Xu et al., 2003) to observe the nesting behavior of mice. Clean cages and bedding were changed 1 h before the experiment, wood chips covered the bottom of the cages with a thickness of approximately 0.5 cm, and a piece of 5 cm × 5 cm compressed cotton was placed in each cage, which was observed and recorded after 12 h. The nesting scoring criteria were as follows: 1: no visible changes in nesting material; 2: partial shredding of nesting material; 3: most of the nesting material was shredded but disorganized without a definite direction; 4: most of the nesting material was shredded but the nest built was flat; and 5: the nest was completely built.

After behavioral testing of the learning memory ability and autonomous exploration ability of mice, three mice were randomly selected from each group, anesthetized with sodium pentobarbital (Sigma-Aldrich; Merk KGaA; Darmstadt, Germany) intraperitoneally, and fixed on a homemade surgical plate. The heart was exposed by opening the chest, and a perfusion needle was inserted and fixed from the apical position. PBS buffer (0.1 M) was slowly injected until the liver and lungs turned white in color, followed by 4% paraformaldehyde solution. The brain was fixed by perfusion, and the head was severed. The dissected brains were quickly and carefully fixed in 4% paraformaldehyde solution for 48 h (4°C). After complete fixation, brain tissues were cut into appropriate sizes and then neatly placed in the embedding boxes, washed to remove the fixative and dehydrated transparently, and routinely paraffin-embedded. The rest of the mice were executed by the same method, and the brains were quickly removed on the ice box and stored in the −80°C refrigerator for backup.

Aβ_25–35 was synthesized by Sangon Biotechnology (P26152). Aβ_25–35 powder was added to Phosphate Buffer Solution (P1020, Solarbio) to make the final concentration of 1 mM. The solution was placed in a 37°C incubator for 7 days and then stored at −20°C.

HT22 cells were cultured in DMEM medium (C11995500BT, GIBCO) containing 10% FBS (C04001500, VIVACELL) and 1% penicillin-streptomycin (P1400, Solarbio) at 37°C and 5% CO₂, and passaged every 2–3 d. The lentiviruses of hnRNP A1 and HK1 were provided by HanBio. Sodium pyruvate (P8380, Solarbio) and medium (PM150271, Procell) were used when sodium pyruvate was used as the sole carbon source for the cells. The following inhibitors were used in the cell experiments: 2-Deoxy-D-glucose (HY-13966, MCE), UK-5099 (HY-15475, MCE), VPC-80051 (Sangon), p38 MAPK-IN-1 (HY-12839, MCE). Dosage of drugs: 2-DG (1 mM); VPC-80051 (50 μM); UK5099 (30 μM); p38 MAPK-IN-1 (40 μM); Aβ_25–35 (20 μM).

Brain tissues of model and wild-type mice within 24 to 48 h after birth were collected and placed in precooled DMEM/F-12 medium. The tissue was then transferred to a 15 mL glass centrifuge tube and the hippocampus was mashed with a disposable plastic tip dropper. F-12 containing 0.125% trypsin (C25200056, GIBCO) was then added, and the tissue homogenate was placed in an incubator at 37°C with a volume fraction of 5% CO₂ saturated humidity for 15 min. An appropriate amount of DMEM/F-12 medium was added to terminate the digestion, then filtered through a mesh sieve into a 15 mL centrifuge tube and centrifuged for 8 min (1000 rpm/min). The supernatant from the centrifuge tube was discarded, and the resultant cell pellet was resuspended with neuron cell seeding medium and mix evenly. The cells were counted with an automatic counter. According to the needs of the experiment, the cells were seeded into 6-well culture plates coated with poly-D-lysine at a density of 5 × 10⁵/ml, and cultured in an incubator at 37°C with a volume fraction of 5% CO₂ saturated humidity. After culturing for 4 h, the seeding solution was replaced with cell maintenance solution, and cytarabine solution (A607079, SANGON BIOTECH) was added 48 h after neuron inoculation to inhibit the proliferation of glial cells. After that, the cell maintenance solution was changed every 2–3 d. After 7 days of culture, Aβ_25–35 was added to make a cellular AD model. Planting solution used in this study: 10% fetal bovine serum (C04001500, VIVACELL), 90% DMEM/F-12 (C11330500BT, GIBCO); maintenance solution: 97% Neurobasal-A (10888022, GIBCO), 2% B27 (17504044, GIBCO), 1% Glutamine (A2916801, GIBCO).

Paraffin was cut into 5.0 μm thick sections and baked at 60°C for 30 min. Slices were dewaxed with xylene and gradient ethanol. HE staining: hematoxylin solution 10 min → distilled water wash 3 s → hydrochloric acid ethanol fractionation 2 s → tap water rinse 10 min → eosin staining 2 min → tap water rinse 10 min. Finally, the sections were routinely dehydrated, cleared and sealed. Nissl staining: 1% toluidine blue staining: After dewaxing, toluidine blue solution was added dropwise to the sections for 30 min (60°C) and rinsed in running water for 5 min. Finally, the sections were routinely dehydrated, transparent and sealed. The sections were observed using a light microscope (BX43; Olympus).

Paraffin embedded tissues were cut at a thickness of 5 μm. Oven baking the pieces at 60°C for 35 min. They were deparaffinized by xylene and dehydrated in graded series of pure alcohol and graded alcohol until distilled water. 1% H₂O₂ pipette was applied to the tissue sections and placed in a damp box sheltered from light for 15 min. After being washed with PBS, the citric acid buffer was placed in microwave oven at high temperature for antigen repair. Then, tissue were blocked with goat serum (ZSGB-BIO) for 50 min (37°C). Tissue sections were combined with Aβ primary antibody (1:400, ab201060, Abcam) and incubated overnight at 4°C. Sections were then incubated with fluorescent secondary antibody (1:400; SA00013-2, Proteintech) for 50 min at 37°C in a light-proof environment. The sections were then stained with DAPI (C0065, Solarbio) for 6 min at 37°C. Slides were mounted with anti-fluorescence quencher (S2100, Solarbio). Finally, they were observed under a fluorescent microscope and photographed (OLYMPUS, BX53).

HT22 was spread on 24-well plates at 20,000/well. After completion of modeling, 24-well plates were washed with PBS and cells were fixed with 4% fixative (P1110, Solarbio) for 20 min and 0.5% Triton X-100 (T8200, Solarbio) for 2 h at room temperature. Cells were sealed with goat serum (ZLI-9022, ZSBG-BIO) for 50 min. Cells were incubated overnight at 4°C with primary antibody hnRNP A1 (1:100; #8443S, Cell Signaling Technology). The sections were then stained with DAPI (C0065, Solarbio) for 6 min. Slides were mounted with anti-fluorescence quencher (S2100, Solarbio). Finally, they were observed and photographed under a 60-fold oil-based fluorescence microscope (OLYMPUS, BX53).

Brief steps were as follows: extraction of total protein from cells (R0010, Solarbio) → protein quantification using BCA kit (GK5012, GENERAY) → protein denaturation at 95°C for 10 min → preparation of SDS-PAGE gels at 10% → loading volume (20 μg) → electrophoresis → electrotransfer (7 min, 25 V, 2.5 A) → blocking of non-specific antigens for 2 h (5% skim milk) → reaction with primary antibody (overnight at 4°C) → reaction with secondary antibody (1 h at room temperature) → imaging (Bio-Rad). The band intensities were quantified with ImageJ software. The following primary antibodies were used: p38 MAPK (1:1500; #8690, Cell Signaling Technology), p-p38 MAPK (1:1500; #4511, Cell Signaling Technology), hnRNP A1 (1:1000; #8443S, Cell Signaling Technology), HK1 (1:2000; 19662-1-AP, Proteintech), Aβ (1:1000; ab201060, Abcam), APP (1:1000; 25524-1-AP, Proteintech), β-Tubulin (1:5000; 10068-1-AP, Proteintech), IgG (1:5000; SA00001-2, Proteintech).

Mitochondrial membrane potential was assayed according to the manufacturer’s (M8650, Solarbio) instructions. First, the medium of the 6-well plate was removed and washed with PBS. Then, 1 ml cell culture medium and 1 mL JC-1 (1 ×) staining working solution were added to each well. The cells were incubated at 37°C for 30 min. After washing with JC-1 buffer, 1 mL medium was added to each well and observed under an inverted fluorescent microscope (OLYMPUS, IX73P2F).

Brief description of the procedure: RNAiso Plus Kit (9108, TaKaRa) to extract RNA from HT22 cells → PrimeScript RT kit (RR037A, TaKaRa) and thermal cycler (C1000 Touch, BIO-RAD) for reverse transcription of RNA → amplification of cDNA on a real-time PCR detector (CFX Connect, BIO-RAD) to amplify cDNA. The thermal cycling procedure was as follows: 95°C (3 min) → 95°C (10 s) → 60°C (45 s) → 40 cycles repeated. Reactions consisted of 7.5 μL TB Green Premix Ex Taq II (RR820A, TaKaRa), 4 μL nucleic acid-free water, 0.5 μL primer, and 3 μL cDNA. Data were analyzed using the comparative cycle threshold (Ct) method (ΔΔCt). The synthesized primers (Sangon Biotech) are listed in Table 1.

The ATP content was detected using a Luminescent ATP Detection Assay Kit (ab113849, abcam). HT22 cells were digested with trypsin and spread into 96-well plates in the same number of cells in a 50 μL system. Detergent solution (solution A) was added and incubated for 5 min, then substrate solution (solution B) was added and incubated for 5 min. The 96-well plates were placed in a light-proof environment for 9 min and finally the luminescence was measured with a multifunctional microplate reader (Varioskan LUX, ThermoFisher). The excitation wavelength is 488 and the emission wavelength is 525.

Pyruvate content was measured according to the manufacturer’s instructions (BC2200, Solarbio). HT22 cells were collected into a centrifuge tube, and the supernatant was discarded after centrifugation (200 g, 5 min). The extract was added in appropriate proportions and crushed by ultrasound (Scientz, Scientz-IID). Hippocampal tissue was added to the extract according to the ratio of about 0.1 g tissue to 1mL extract and homogenized in an ice bath. The solution was allowed to stand for 30 min, centrifuged at 8000 g for 10 min at room temperature, and the supernatant was taken to be measured. Reagent II was added after standard solution or sample was mixed with reagent I in a 96-well plate, and absorbance A was determined at a wavelength of 520 nm using a microplate reader (MuitisKan, ThermoFisher). Finally, the pyruvate content was calculated after measuring the protein concentration of the samples using the BCA method.

The enzymatic activities of HK, PFK and PK were assayed according to the instructions of the commercial kits (BC0745, Solarbio; BC0535, Solarbio; and BC0545, Solarbio). Cells or tissues were collected into centrifuge tubes and extracts were added in appropriate proportions. Cells were crushed with ultrasound (Scientz, Scientz-IID) and tissues were homogenized in ice bath. The extracts were then centrifuged at 8000 g for 10 min at 4°C. The supernatant was taken for testing. The reagents were added sequentially to the 96-well plate according to the instructions. After mixing well, the absorbance difference values were measured at different times on a microplate reader (MuitisKan, ThermoFisher) according to the instructions. The wavelength of the microplate reader was set to 340 nm.

The thiazolyl blue powder (IM0280, Solarbio) was dissolved in PBS (P1020, Solarbio) and stored in the dark at 4°C. HT22 cells were plated at 2000 cells/well in a 96-well plate in a volume of 100 μL. Following drug intervention for 24 h, HT22 cells were given 20 μL of thiazolyl blue solution. The 96-well plate was placed in a 37°C incubator and incubated in the dark for 4 h. After aspirating the medium, 100 μL of DMSO was added to dissolve the purple precipitate at the bottom. Finally, absorbance was determined at a wavelength of 490 nm using a microplate reader (MuitisKan, ThermoFisher).

The experiment was performed by IEMed Guangzhou Biomedical Technology Co, Ltd. Crosslinking-immunprecipitatio assays were performed by using the CLIP Kit (IEMed, K319) along with the hnRNP A1 antibody (#8443S, Cell Signaling Technology). HT22 cells were irradiated with 254 nm UV lamp at a dose of 0.15 J/cm² for cross-linking for 1 min. PBS was added into the culture plate. Cells were scraped off with a cell scraper and transferred to a centrifuge tube, centrifuged at 1000rpm for 5 min at 4°C, and supernatant was removed. 2 mL of cell lysis buffer and 20 μL of protease inhibitor were added to cell precipitation. IP and IgG groups were incubated on a vertical mixer with shaking at room temperature for 1 h. The beads were collected on a magnetic rack. 500 uL of CLIP wash buffer was added and washed with shaking at room temperature for 3 min, and the beads were collected on a magnetic rack. Input group, IP group and IgG group were added with 50 μL CLIP washing buffer and 1 μL RNase T1 respectively. Each group was incubated at 37°C for 1 h. Beads of IP group and IgG group were collected on the magnetic rack, and then mixed upside down with 200 μL CLIP washing buffer. Beads were collected on the magnetic rack and repeated twice. The Input group, IP group and IgG group were added with 50 μL PK buffer and 10 μL protease K for digestion at 55°C for 1 h. 120 μl nucleic acid BF C and 10 μL magnetic beads were added to each tube, incubated at room temperature for 20 min, and the magnetic beads were collected on the magnetic rack. Each group was added with 60 μL nucleic acid BF C, which was collected on the magnetic rack immediately and repeated twice. After drying at room temperature for about 10 min, 20 μL RNase-free water was added and the supernatant RNA was collected on the magnetic rack after 3 min of elution at room temperature. Finally, qPCR was performed to detect the enrichment. The synthesized primers are listed in Table 2.

RNA immunoprecipitation assays were performed by using the RIP Kit (IEMed, K303) along with the hnRNP A1 antibody (1:50; #8443S, Cell Signaling Technology). Two 10 μL of Protein A/G beads (labeled as IP and IgG, respectively) were added to 100 μL of AB binding buffer and mixed by flicking, then the beads were collected on a magnetic stand and the supernatant was removed; IP group: 100 μL of AB binding buffer and 2 μg of hnRNP A1 antibody were added; IgG group: 100 μL of AB binding buffer and 2 μL of IgG antibody were added and mixed by flicking, then the beads were collected on a magnetic stand and the supernatant was removed. For the IgG group, 100 μL of AB binding buffer and 2 μL of IgG antibody were added and mixed with flicking. The mixture was incubated at room temperature for 30 min, then the beads were collected on a magnetic rack and the supernatant was removed; for the IP and IgG groups, 100 μL of AB binding buffer were added and mixed upside down, then the beads were collected on a magnetic rack and 80 μL of supernatant were removed; collect 10⁷ HT22 cells were washed twice with 1 × PBS, and the last time the residual PBS was aspirated; 700 μL of 2 × RIP buffer, 20 μL of Protease inhibitors, 100 μL of RNase inhibitor were added to the cell precipitate, mixed with sufficient blowing, and then lysed at room temperature for 20 min; 700 μL of Binding enhancer was added. Binding enhancer, mix upside down, centrifuge at 12000g for 10 min at 4°C, transfer supernatant to new RNA-free centrifuge tube; 5 μL of Clean beads were added to supernatant, incubate on vertical mixer at room temperature with shaking for 20 min; 800 μL of supernatant were transfered to IP sample tube on magnetic stand; 400 μL of supernatant were transfered to IgG sample tube. 800 μL of supernatant were transfered to the IP sample tube and 400 μL of supernatant to the IgG sample tube; 100 μL of supernatant were transfered (labeled as Input group and stored at −80°C for use); The IP and IgG groups were combined by incubation on a vertical mixer with shaking at room temperature for 60 min; The beads were collected on the magnetic stand and removed the supernatant; 500 μL of RIP washing buffer were added to each of the IP and IgG groups and washed with shaking at room temperature for 5 min; then the beads were collected on the magnetic stand and removed the supernatant. 50 μL of Polysome elution buffer were added to the IP and IgG groups, mixed with a light flick, and then eluted at 75°C for 10 min; 100 μL of Nucleic acid BF C were added to the IP and IgG groups and 200 μL of Nucleic acid BF C to the Input group. 10 μL Magnetic beads were added, shaked and combined for 10 min at room temperature; The beads were collected on the magnetic rack to remove the supernatant; 100 μL Nucleic acid BF C were added to each group, flicked and mixed, then removed the supernatant directly on the magnetic rack; after drying at room temperature (about 10 min), 20 μL RNase-free water were added to each group elute at room temperature for 3 min, and collected the supernatant on the magnetic rack RNA. HK1 mRNA levels in the immunoprecipitates were measured by qPCR. Finally, the binding situation was judged by the enrichment efficiency.

No less than 5 million cells were collected according to glutathione determination instructions (BC1175, Solarbio). The cells were washed with PBS and then resuspended with 1 mL reagent A. The cells were ultrasonically broken (Scientz, Scientz-iID) and centrifuged. Then, supernatant measurements were collected. The wavelength was adjusted to 412 nm with an enzyme marker (MuitisKan, ThermoFisher). Finally, the GSH content was calculated according to the number of cells.

A concentration of 1 ml DCFH-DA (10 μM) was added to a six-well plate with the medium removed according to the instructions of the Reactive Oxygen Species Test Kit (CA1410, Solarbio). The cells were cultured in an incubator for 30 min. HT22 cells were digested with trypsin and spread into 96-well plates in the same number of cells in a 50 μL system. Finally, the luminescence was measured with a multifunctional microplate reader (Varioskan LUX, ThermoFisher). The excitation wavelength is 488 and the emission wavelength is 525.

The experiment was completed under the experimental platform provided by the Sichuan Key Laboratory of Nuclear Medicine and Molecular Imaging. Mice were injected intraperitoneally with sodium barbiturate, and after complete anesthesia, approximately 14.8–16.5 mBq of ¹⁸F-FDG (prepared by the Sichuan Key Laboratory of Nuclear Medicine and Molecular Imaging) was injected via the tail vein. The mice were fixed in prone position on the scanner bed. The long axis of the head was parallel to the long axis of the scanner and within the field of view of the scanner. 20 min of ¹⁸F-FDG uptake in the mice was followed by 10 min of PET image acquisition and 10 min of CT image acquisition. The data were then processed using SIEMENS Inveon Research Workplace V4.2 software. ROI selection: The 3D ROI technique was used to manually select the 3D ROI of the whole brain, hippocampal and cortical regions in the cross-sectional, coronal and sagittal planes with reference to the mouse brain ex vivo localization spectrograms, and the average ROI per gram of tissue within the ROI was calculated of uptake rate.

Statistical analysis was performed using GraphPad software Prism 8.0. All data were expressed as mean ± standard error (mean ± SEM). Student’s t-test and one-way analysis of variance (ANOVA) were used to determine the levels of statistical differences between two and multiple groups, respectively. Statistically significant differences were set at *p < 0.05, **p < 0.01, *** p < 0.001.

Numerous studies have pointed that AD is inextricably linked to impaired glycolysis (Lee et al., 2016). The 3 × Tg-AD mice have been used successfully to model Alzheimer’s disease for years (Chen et al., 2022). As illustrated in the schematic diagram (Supplementary Figure 1A), the selected time points represented corresponding age stages for conducting animal experiments. Our previous studies have shown Aβ deposition in the brains of 9-month-old AD mice, so we selected 9-month-old mice for follow-up experiments (Yan et al., 2023). ¹⁸F-FDG MicroPET showed reduced glucose utilization in the hippocampal region of 3 × Tg-AD mice compared to WT group mice in 9-month-old (Figures 1A, B). This phenomenon is mainly related to both glucose uptake and reduced HK activity (Figure 1C). We found that reduced HK activity was associated with reduced HK1 expression in the hippocampal tissue of 9-month-old 3 × Tg-AD mice (Figure 1D). It is suggested that reduced HK1 expression is involved in AD pathogenesis. HK is the first rate-limiting enzyme of the glycolytic pathway and controls the entire glycolytic flux. Of these, brain tissue is predominantly dominated by HK1, so we proceeded to assess whether hippocampal tissue glycolytic function is impaired in the AD model. We assessed the content of pyruvate, the end product of glycolysis, in the hippocampal tissue of 9-month-old mice. The pyruvate biochemical kit showed a significant decrease in pyruvate levels in hippocampal tissue of 3 × Tg-AD mice compared to the WT group (Figure 1E). This was associated with reduced activity of HK (Figure 1C) and PK (Figure 1F) enzymes among the three key enzymes of glycolysis in hippocampal tissue of 9-month-old 3 × Tg-AD mice. No change in PFK enzyme activity was found (Figure 1G). The HT22 cell line is an immortalized mouse hippocampal neuron cell line with properties similar to those of mature hippocampal neurons in vivo, such as neurite growth, expression of functional cholinergic markers and receptors, which allows HT22 cells to be utilized to characterize the molecular events underlying neurodegeneration and neuronal injuries. We used HT22 to administer exogenous Aβ_25–35 to simulate the AD model to explore the effect of Aβ on neurons. Similarly, we observed a significant reduction in HK1 protein expression (Figure 1H), HK enzyme activity (Figure 1I), PK enzyme activity (Figure 1J) and its end product pyruvate (Figure 1K) in Aβ_25–35-induced HT22 cells, which is in line with the findings of animal tests. To further verify that the decrease in pyruvate during glucose metabolism was a result of decreased glycolytic production rather than increased oxidative phosphorylation into the mitochondria, we designed to use a glucose analog (2-DG) to inhibit glycolysis and a mitochondrial pyruvate carrier inhibitor (UK5099) to inhibit mitochondrial uptake of pyruvate. When 2-DG was intervened, the 2-DG + Aβ_25–35 group exhibited increased pyruvate levels compared to the 2-DG group (Figure 1K). Aβ_25–35 may have inhibited oxidative phosphorylation of pyruvate into mitochondria, resulting in increased pyruvate abundance. At the time of the UK5099 intervention, pyruvate was reduced by approximately 15% in the Aβ_25–35 group compared with the control group (Figure 1L), while the Aβ_25–35 + UK5099 group showed a greater reduction of 35% compared with the UK5099 group (Figure 1L). Our study supported the function of Aβ_25–35 injury on neuronal glycolysis, whose dysfunction is involved in AD.

Using MWM to test learning and memory abilities is dependable (Tucker et al., 2018). It showed that 9-month-old 3 × Tg-AD mice exhibited prolonged escape latency on the fourth day (Supplementary Figure 1B) and impaired spatial perception (Supplementary Figure 1C), which is in agreement with earlier study (Jiang et al., 2015). Subsequently, we confirmed this with different behavioral tests. Turning bar experiment showed that 9-month-old 3 × Tg-AD mice with poor coordination could fall off the batting in a short period of time (Supplementary Figure 1D). The open-field experiment suggested that autonomous and exploratory behaviors were diminished in 9-month-old 3 × Tg-AD mice (Supplementary Figure 1E). Nesting experiments showed reduced daily living skills in 9-month-old 3 × Tg-AD mice (Supplementary Figure 1F). These lines of evidence suggest that 9-month-old 3 × Tg-AD mice already exhibit cognitive impairments that affect motor behavior. The hippocampus is a site of memory storage and spatial orientation (Zhang et al., 2022). Next, we observed structural changes and neuronal survival in DG and CA3 regions of hippocampal tissue by tissue sectioning. HE staining showed that the structure of neurons was obviously intact and densely arranged in WT mice, whereas neurons appeared with obvious pyknosis and loose arrangement in 3 × Tg-AD mice (Supplementary Figure 2A). Nissl staining suggested that neuron morphological structure destruction and the number of surviving neurons were reduced in 3 × Tg-AD mice (Supplementary Figures 2B–D). Anti-hnRNP A1 antibodies are known to cause mislocalization and increased nuclear depletion of hnRNP A1 in neurons and induce neuronal death (Libner et al., 2022). One study also reported that hnRNP A1 regulates glycolytic function by selective cutting of PK RNA (David et al., 2010). We hypothesized that hnRNP A1 may be an important molecule that links AD neuronal death and glycolytic dysfunction. Therefore, we assessed the hnRNP A1 expression in brain tissue from 6-, 9-, 16-, and 21-month-old AD and WT mice to investigate whether there is an association between AD and hnRNP A1. Western blotting revealed that hnRNP A1 in mice brain tissue gradually decreased from 9 months of age. Meanwhile, hnRNP A1 expression in the brain tissue of 3 × Tg-AD mice was consistently lower than that of WT mice at the same month of age (Figure 2A). This result supported that 9-month-old 3 × Tg-AD mice exhibited a variety of behavioral abnormalities and reduced hnRNP A1 expression compared to control mice. Furthermore, we hypothesized that Aβ_25–35 induced neuronal death by regulating the down-regulation of hnRNP A1 As expected, immunofluorescence showed that Aβ_25–35 resulted in a downregulation of the mean fluorescence intensity of hnRNP A1 (Figure 2B). Western blotting revealed that Aβ_25–35 induced a downregulation of hnRNP A1 expression in primary mice hippocampal neurons (Figure 2C). We used VPC-80051 (Carabet et al., 2019) to inhibit the RNA binding domain (RBD) of hnRNP A1 and detected changes in the mitochondrial membrane potential. JC1 emits red or green fluorescence to represent high or low mitochondrial membrane potential, respectively. At the same time, according to the effect of VPC-80051 on the viability of HT22 cells, we decided to use 50 μM for the experiment (Supplementary Figure 3A). We found that mitochondrial membrane potential in VPC-80051 group showed a significant decrease in HT22 cells (Figure 2D), which is an early sign of apoptosis (Jin et al., 2019). These results suggested that neuronal death is closely associated with the reduced hnRNP A1 expression in AD models. The reduced hnRNP A1 expression is involved in AD pathogenesis, and hnRNP A1 may be a predisposing factor for age-related neurodegeneration.

It has been reported that hnRNP A1 regulates glycolytic function by selective cutting of PK RNA (David et al., 2010). We also confirmed the cooperative relationship between hnRNP A1 and PKM through RIP experiments (Supplementary Figure 3B). We speculated that the decreased HK1 might be related to the lower regulation of hnRNP A1. To test this conjecture, we used HT22 cells as cell model in vitro to elucidate the mechanism of hnRNP A1 in AD. Then, we cultured HT22 cells with VPC-80051 to inhibit the RNA binding domain (RBD) of hnRNP A1 for 24 h and found a significant down-regulation of HK1 protein (Figure 3A) and HK1 mRNA expression (Figure 3B). Next, we constructed a stable overexpression cell line of hnRNP A1 (Figure 3C) in cultured HT22 cells. We found that the mRNA (Figure 3D) and protein expression (Figure 3E) levels of HK1 were significantly increased after hnRNP A1 overexpression by western blotting and RT-qPCR. To further verify the existence of their partnership, we used hnRNP A1 antibody to pull down all RNAs bound to hnRNP A1. By RT-qPCR, we found a significant enrichment of HK1 compared to the IgG group (Figure 3F). Finally, to find the base sequence of HK1 mRNA binding to hnRNP A1, we designed 19 primers targeting HK1 mRNA sequences (Figure 3G). The CLIP-qPCR revealed a clear enrichment at primer number P7070, which was analyzed to act on the 2605-2821 region of the HK1 mRNA (Figure 3H). The binding site may be related to the best sequence 5′-UAG-3′ (Jain et al., 2017) in the sequence as determined by analysis of the affinity distribution, or to the sequence 5′-UAG-3′ (Bartys et al., 2022) that can be used as an efficient binding platform for hnRNP A1 by binding screening assays, Kd assays and UV melting experiments (Figure 3I). In conclusion, the above results supported that hnRNP A1 binds directly to HK1 mRNA to regulate the protein expression of HK1 and provided a possible base sequence for binding to HK1 mRNA.

To investigate the effect of hnRNP A1 on the glycolytic function of hippocampal neurons in mice. First, we found that the hnRNP A1 overexpression could ameliorate Aβ_25–35-induced cell axonal damage (Figure 4A) and cell death (Figure 4B) in HT22 cells. Mechanistically, hnRNP A1 overexpression caused elevation of pyruvate (Figure 4E) and lactate (Supplementary Figure 3C) from glycolysis by improving protein expression and enzymatic activity of HK1 (Figure 4C) in Aβ_25–35-induced HT22 cells. Here, a point to note was that the pyruvate content of hnRNP A1 overexpression did not change when not given the UK5099 intervention compared to the normal group, which may be related to increased mitochondrial uptake of pyruvate by hnRNP A1 overexpression (Figure 4D). We then demonstrated that hnRNP A1 overexpression ameliorated energy impairment in Aβ_25–35-induced HT22 cells by assaying the ATP content (Figure 4F). Reverse validation by VPC-80051 showed that inhibition of hnRNP A1 by RBD can cause a decrease in neuronal ATP (Figure 4G) and death (Supplementary Figure 3A). Secondly, to further confirm that hnRNP A1 overexpression is mitigating HT22 damage by improving HK1 expression. We also constructed a stable overexpression cell line of HK1 in cultured HT22 cells. As expected, HK1 overexpression also improved the Aβ_25–35-induced reduction of pyruvate (Figure 4H) in HT22. Finally, to show that overexpression of hnRNP A1 ultimately ameliorated neuronal death by promoting pyruvate production. We used sodium pyruvate as the sole carbon source for neurons in culture, and the results showed that increasing the amount of sodium pyruvate under normal conditions significantly increased neuronal survival (Figure 4I). Similarly, increased levels of sodium pyruvate improved Aβ_25–35-induced impairment of energy metabolism (Figure 4J) and neuronal death (Figure 4K) in HT22. In conclusion, Aβ in neurons caused glycolytic damage by downregulating hnRNP A1 expression. Impaired energy metabolism and neuronal death could be reversed by enhancing hnRNP A1. At the same time, we found that the protection of hnRNP A1 on neurons was multifaceted, which may be related to the improvement of glycolysis function (Supplementary Figures 3D–G). We described this in more detail and expanded discussion on it in the later discussion section.

As is well known, damage to neurons exposed to Aβ is considered to be a major pathological feature of AD (Yan et al., 2023). It has been reported that hnRNP A1 is involved in alternative splicing of amyloid precursor protein (APP) RNA in neuronal cells differentiated from malignant pluripotent embryonal carcinoma, and increasing hnRNP A1 reduces the formation of Aβ (Donev et al., 2007). At the same time, immunofluorescence revealed that Aβ was widely distributed in DG and CA3 regions of 9-month-old 3 × Tg-AD mice, whereas no significant deposition was detected in WT mice at the same month of age (Supplementary Figures 2E–H), which is consistent with previously reported clinical observations (Colom-Cadena et al., 2020). To verify whether the mechanism as described above exists between Aβ deposition and hnRNP A1 in mice hippocampal neurons, follow-up experiments were performed using HT22 cells. As expected, the expression of APP, Aβ oligomers were significantly elevated after inhibition of hnRNP A1 using VPC-80051 (Figure 5A). Next, we further designed different primers for CLIP-qPCR experiments based on APP mRNA (Figure 5B). The results showed the presence of four main binding regions with significant enrichment, namely base sequences 104-204, 873-902, 1169-1343, and 1475-1575 (Figures 5C, D). Exon 7 have the similarity as previously reported, but the binding sites of exon 2, exon 6 and exon 11 were not reported. After searching all transcripts known to APP, no transcripts were found that did not include exons 2, 6 and 11. These results suggest that hnRNP A1 binds directly to APP mRNA to regulate Aβ production. Furthermore, recent studies have reported that Aβ-induced phosphorylation of p38 MAPK increases ROS, leading to neuronal cell death (Lee and Kim, 2017). In fibroblasts hnRNP A1 expression levels are regulated in a p38 MAPK-dependent manner, probably through its phosphorylation (Shimada et al., 2009). Therefore, we hypothesized that Aβ_25–35 induced downregulation of hnRNP A1 probably through activation of p-p38 MAPK. As expected, Western blotting results showed that Aβ_25–35 significantly increased the expression of p-p38 MAPK (Figure 6A), but not the expression of p38 MAPK (Figure 6A). To test this conjecture, the p38 MAPK activity inhibitor p38 MAPK-IN-1was used, and the results showed that p38 MAPK-IN-1 significantly inhibited Aβ_25–35-induced downregulation of hnRNP A1 (Figure 6B). Our data suggested that Aβ_25–35-induced downregulation of hnRNP A1 may be achieved through activation of p38 MAPK phosphorylation. Taken together, the evidence suggested a bidirectional regulatory mechanism between hnRNP A1 and Aβ, which may further induce hnRNP A1 downregulation and ultimately exacerbate neuronal glycolytic dysfunction.