Abstract Introduction Liver dysfunction contributes to Alzheimer’s disease (AD) pathogenesis, and evidence suggests that the liver is involved in amyloid β (Aβ) clearance, and regulates Aβ deposition in the brain. However, the specific regulatory mechanism remains elusive. Objectives Angiopoietin-like protein 8 (ANGPTL8), a high expression of liver-specific secreted proinflammatory factor, crosses the blood‒brain barrier from the bloodstream to abnormally activate microglia and promote AD progression. Methods The ANGPTL8^−/− mice and 5 × FAD mice were crossed mutated and subjected to the Morris water maze test and novel object recognition test to assess cognitive ability in different cohorts. Thioflavin-S, NeuN, and Nissl staining were used to assess Aβ deposition and neuron loss. The number of phagocytic microglia was evaluated with Fitc latex beads. Adeno-associated virus 8 (AAV8) hydrodynamically injected restored the liver ANGPTL8 levels of ANGPTL8^−/− 5 × FAD mice for further experiments. Single-cell RNA sequencing, bulk RNA sequencing and transmission electron microscopy were used to explore the role of ANGPTL8 in regulating AD progression, and drug screening was carried out to identify an effective inhibitor of ANGPTL8. Results ANGPTL8 knockout improved cognitive function and reduced Aβ deposition by reducing microgliosis and microglial activation in 5xFAD mice. Mechanistically, ANGPTL8 crossed the blood‒brain barrier and interacted with the microglial membrane receptor PirB/LILRB2. This interaction subsequently activated the downstream NLRP3 inflammasome, leading to microglial pyroptosis and exacerbating the Aβ-induced release of inflammatory factors, thereby accelerating AD progression. Furthermore, the administration of metformin, an ANGPTL8 inhibitor, improved learning and memory deficits in 5 × FAD mice by negating microglial pyroptosis and neuroinflammation. Conclusions ANGPTL8 aggravates microglial pyroptosis via the PirB/NLRP3 pathway to accelerate the pathogenesis of AD. Targeting high expression of ANGPTL8 in the liver may hold potential for developing therapies for AD. Supplementary Information The online version contains supplementary material available at 10.1186/s12974-025-03487-3. Keywords: ANGPTL8, Neuroinflammation, Liver‒brain axis, Microglial pyroptosis, Alzheimer's disease Introduction Alzheimer's disease (AD) is the most prevalent form of dementia and is characterized by progressive synaptic dysfunction and the accumulation of extracellular amyloid-beta (Aβ) plaques and intracellular neurofibrillary tangles (NFTs) [[60]1]. Increasing evidence highlights the critical role of neuroinflammation in AD pathogenesis, with Aβ acting both as a causative factor and a consequence of neuroinflammation [[61]2]. An increasing number of investigations have focused on understanding how neuroinflammation contributes to aberrant Aβ deposition and disease progression in AD patients. Microglia, the primary immune cells in the central nervous system, play pivotal roles in AD [[62]3, [63]4]. These cells act as drivers of innate immunity, and prolonged activation leads to functional impairment, exacerbating the progression of AD [[64]3, [65]4]. Chronic neuroinflammation results in microglial pyroptosis, a novel inflammatory form of regulated cell death (RCD) characterized by cell membrane rupture and proinflammatory effects [[66]5]. Excessive pyroptosis contributes to severe inflammatory responses, resulting in neuronal damage and loss [[67]5]. Furthermore, certain peripheral inflammatory factors cross the blood‒brain barrier (BBB), activating microglia and promoting neuroinflammation within the brain [[68]6]. Emerging evidence indicates that the liver is a key peripheral organ involved in this process, with inflammatory mediators from the liver reaching the brain via the circulatory system, a phenomenon referred to as the liver‒brain axis [[69]7, [70]8]. Notably, disease conditions such as obesity, type 2 diabetes, and nonalcoholic fatty liver disease (NAFLD) are reportedly associated with an increased risk of AD [[71]9, [72]10][.] However, the specific mechanisms driving the peripheral liver‒brain axis in AD development remain elusive. Angiopoietin-like protein 8 (ANGPTL8), also known as lipasin, RIFL, or TD26, is a secreted protein predominantly expressed in liver cells and adipose tissue. It plays a critical role in the regulation of glucose and lipid metabolism [[73]11–[74]13]. Recent studies have indicated that ANGPTL8 levels are increased during oxidative stress and inflammatory diseases [[75]14][.] It has been reported that inflammatory cytokines such as TNF-α and IL-1β upregulate ANGPTL8 expression, suggesting that ANGPTL8 is closely involved in the development of inflammation [[76]15]. Other studies also reported increased serum ANGPTL8 levels in patients with aging-related chronic inflammatory conditions such as NAFLD, diabetes, and metabolic syndrome [[77]16]. Our previous work revealed that the secretory protein ANGPTL8, which is present in the blood microenvironment, bidirectionally regulates insulin resistance in type 2 diabetes [[78]17, [79]18]. We have also demonstrated that ANGPTL8, a proinflammatory factor in the liver microenvironment activates hepatic stellate cells and immune cells, thereby promoting the development of aging-related diseases such as nonalcoholic fatty liver disease-related liver fibrosis [[80]19] and liver cancer [[81]20]. ANGPTL8 in the circulation regulates mesenchymal stem cell differentiation and pathological cardiac remodeling [[82]21, [83]22]. These findings indicate that the secretory protein ANGPTL8 within the microenvironment may play a pivotal role in the regulation of aging-related diseases. AD, the most representative age-related neurodegenerative disorder, is associated with a significantly increased risk of onset with age [[84]23]. In our preliminary study, we unexpectedly found that ANGPTL8 knockout (KO) delayed age-related cognitive decline and hair greying in high-fat-induced liver fibrosis model mice, suggesting that ANGPTL8 may influence both neuroinflammation and cognitive function. However, its precise role in AD remains to be fully elucidated. In this study, we provide compelling evidence that high liver-specific expression of ANGPTL8 acts as a proinflammatory mediator, accelerating AD pathogenesis. We showed that conventional ANGPTL8 knockout significantly reduces Aβ deposition and alleviates cognitive impairments in an AD mouse model. Mechanistically, we revealed that ANGPTL8 interacts with the microglial membrane receptor PirB/LILRB2. This interaction subsequently activates the downstream NLRP3 inflammasome, leading to microglial pyroptosis and exacerbating the Aβ-induced release of inflammatory factors, thereby accelerating AD progression. Additionally, we demonstrate that metformin (Met), a drug known to inhibit ANGPTL8 expression, ameliorates cognitive deficits in a 5 × FAD mouse model of AD. Our findings suggest that ANGPTL8 is an important key mediator of the liver‒brain axis and a potential therapeutic target for AD. Materials and methods Animal experiments All animal experiments were approved by the Institutional Animal Care and Use Committee at HuBei University of Medicine, Shiyan, China. The 5 × FAD (C57BL/6J) mouse (the selection of the 5xFAD model is based on its applicability in aging-related research—this model not only recapitulates early-stage AD mechanisms but also aligns closely with the focus of this study on the role of ANGPTL8 in the occurrence of AD), which expresses five familial AD gene mutations, that were provided by Li Ding (HuBei University of Medicine, Hubei, China). ANGPTL8^−/− (C57BL/6 J) mice were crossed with 5 × FAD mice to yield four phenotypes: WT, 5 × FAD, ANGPTL8^−/−, and ANGPTL8^−/− 5 × FAD. The mice were housed in our institute's animal care facility under ambient conditions (26 ± 1°C) with a 12-h light/12-h dark cycle. They had unrestricted access to standard rodent chow and clean water. The experimental protocol, treatments and time points are shown in the online supplement and supplemental Fig. [85]1 (Fig. S1) and are briefly described below. Fig. 1. [86]Fig. 1 [87]Open in a new tab ANGPTL8 KO improved the cognitive function of 5xFAD mice. A Representative images of the tracks of 9-month-old WT and ANGPTL8 KO mice in the Morris water maze (MWM) test. B-C Statistical results of the escape latency (B) and number of platform crossings (C) of 9-month-old mice (n = 8) on d 7. D Schematics of the experimental procedures for the four groups of mice (WT, ANGPTL8 KO, 5 × FAD, and ANGPTL8 KO:5xFAD; every group included half male and half female mice). E Representative plot of the MWM test results of 3-month-old mice in the four groups (n = 8–10). F The time needed to reach the hidden platform (escape latency) was plotted across training days. Statistical analysis revealed a difference in 3-month-old ANGPTL8 KO:5 × FAD mice compared with WT, ANGPTL8 KO and 5 × FAD mice on d 7 (n = 8–10). G‒I. Escape latency, number of platform crossings and swimming speed (mm/s) were analysed during the MWM test (d 7) in 3-month-old mice. J‒K The escape latency and number of platform crossings were analysed during the MWM test (d 7) in 6-month-old mice (n = 6‒8). L‒M Recognition indices of the four groups of 3- and 6-month-old mice were detected via a novel object recognition experiment (n = 6–8). N‒P. IHC staining (N) and statistical analysis (O‒P) of Aβ plaques per pixel area in the hippocampus and cortex of four groups of mice (WT, ANGPTL8 KO, 5xFAD, and ANGPTL8 KO:5xFAD). (n = 7—11 slices from three mice per group). *P < 0.05, ** P < 0.01, *** P < 0.001, ns = not significant Grouping Six- to eight-week-old mice were randomly divided into the following groups: WT, WT + Met, 5 × FAD, 5 × FAD + Met, ANGPTL8^−/−, ANGPTL8^−/− + Met, and ANGPTL8^−/− 5 × FAD: ANGPTL8^−/− 5 × FAD + Met. Each group consisted of an equal distribution of male and female mice. The Met treatment group received 300 mg/kg/d of Met (n = 8–10/group). After four weeks of oral administration, the mice were subjected to behavioural tests and their brains were harvested for immunohistochemical examinations. Hydrodynamic tail vein injection Some cohorts of mice were subjected to hydrodynamic tail vein injection following all the methods described previously with some modifications [[88]19, [89]24] Plasmid Some cohorts received a total of 20–30 μg (1 μg/g body weight) of pEGFP-N1 or pEGFP-N1-human ANGPTL8 plasmid DNA in sterile saline via tail vein injection (n = 3–5/group). The volume of sterile saline solution was 10% of the body weight, and the injection time was between 8 and 10 s. The mouse liver and brain were collected after plasmid expression for 48 h via frozen sectioning to examine the expression of GFP and GFP-ANGPTL8. The stained sections were examined via a confocal microscope (Olympus Corporation) equipped with NIS Elements Advanced Research Software. AAV8 In some cohorts, pAAV8–mANGPTL8(mouse ANGPTL8)-GFP and control viruses (2 × 10^13 genome copy/kg, tail vein injection, pAAV8-mANGPTL8-Flag and control viruses purchased from Genomeditech Co., Shanghai, China) in sterile saline were injected into the tail vein (8–10 mice per group). The volume of sterile saline solution was 10% of the body weight, and the injection time was between 8 and 10 s. The mice received two injections of the AAV8 virus at 8 and 16 weeks of age, followed by cognitive behavioral tests and pathological examinations at week 24. Behaviour assessment The Morris water maze and novel object recognition (NOR) tests were carried out as reported previously with some modifications [[90]10]. Morris water maze test Spatial learning memory tests were conducted via the Morris water maze. In this test, the mice (6–10 mice per group) were placed in a circular pool (100 cm diameter, 50 cm height) filled with water at 22°C. The pool was divided into four quadrants of equal size, each containing a submerged platform (10 cm diameter) positioned 1 cm below the water surface. Each mouse underwent three trials per day for 7 consecutive days. In each trial, a mouse was placed in a randomly selected quadrant and allowed to swim freely for a maximum of 60 s. If the mouse reached the platform within 60 s, the escape latency, defined as the time taken to reach the platform, was recorded. If the mouse failed to locate the platform within 60 s, it was placed on the platform for 20 s, and the escape latency was recorded as 60 s. On the 7th d, the MWM probe trial, involving the removal of the platform from the pool, was conducted. The mice were placed in the water to swim for 60 s to search for the platform. An analysis-management system was used to measure swimming speed, escape latency, time spent in the target quadrant, and the number of target crossings. Novel object recognition Five days after MWM testing, we conducted a novel object recognition (NOR) test to evaluate animal behaviour when novel and familiar objects were introduced (6–10 mice per group). The NOR test consists of habituation, training, and testing sessions. Initially, the mice were placed in an open field for 5 min on the first day of the habituation procedure. During the training procedure, two identical objects were positioned in opposite corners of the arena. The mouse was then placed in the center of the arena, equidistant from the two identical objects. The mice were permitted to explore freely for 5 min. After 1 h, the mouse was placed in the open arena with one of the original objects and a new novel object for 5 min. We recorded the percentage of time spent examining the new object. Both the novel and original objects were identical in material and size, differing only in shape. The recognition index (RI) was calculated as follows: RI = (T[n])/(T[n] + T[o]), where T[n] represents the time spent with the new object and T[o] represents the time spent with the old object. A higher RI indicates superior retention and memory at 24 h. Brain tissue collection Following the completion of the behavioural assays, the mice were anaesthetized with terminal anesthesia and transcardially perfused with PBS, followed by ice-cold 4% paraformaldehyde (PFA). The brains were carefully extracted, postfixed overnight in 4% PFA at 4°C and subsequently embedded in paraffin for thioflavin-S staining and immunohistochemistry examination. For biochemical assays, the remaining mice were transcardially perfused with PBS to remove blood from the body. The brain or hippocampus was dissected and promptly stored at -80°C for the subsequent assays described below. RNA sequencing Total RNA was isolated from the hippocampal tissue of 6-month-old 5 × FAD and ANGPTL8-/- 5 × FAD mice. RNA-seq was then performed by Genomics Co. via the Illumina instrument, followed by computational analysis. The criteria for differential genes were set as a P value < 0.01 and a fold change > 1.5 or < 0.5. Finally, the differentially expressed genes (DEGs) were characterized via gene set enrichment analysis (GSEA) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. Single-cell RNA sequencing Cortical brain tissue was harvested as previously described [[91]25], with some modifications. Briefly, scRNA-seq libraries were prepared via the Chromium controller and Chromium Single Cell 3' Reagent Kit v3 (10 × Genomics; Pleasanton, CA, USA) according to the manufacturer's instructions. The libraries were subsequently sequenced by BGI Genomics Co. on an MGISEQ-2000 platform. The sequencing depth of each sample was 700 M raw reads, and sample demultiplexing, barcode processing and single-cell counting were performed via Cellranger-6.1.2 (10X Genomics). Brain pathological analysis All analyses were based on the Bregma coordinate system (Paxinos & Franklin’s Mouse Brain Atlas, 4th ed.): Hippocampus: bregma -1.34 mm to -2.92 mm (coronal sections), focusing on the dorsal CA1 region and dentate gyrus. Cortex: bregma -1.58 mm to -2.30 mm (prefrontal and somatosensory cortices), with laminar specificity (e.g., layers II-IV) confirmed by histological markers (e.g., NeuN). Three sections (spaced 120 μm apart to avoid overlap) per mouse. Rationale: Hippocampal and cortical structures are anatomically intact within this range, with stable pathological marker distributions (e.g., Aβ plaques). Interval sampling (every 4th section) ensures spatial independence while covering the entire region of interest. Thioflavin-S staining The cortex and hippocampal sections (n ≥ 3 mice per group), which were taken at various depths along the rostrocaudal axis, were washed three times in PBS, immersed in 1% thioflavin-S solution (Sigma‒Aldrich) at 37°C for 30 min in the dark and subsequently differentiated in 70% ethanol. Aβ plaques were assessed with a fluorescence microscope (Nikon A1R) and quantitatively analysed with ImageJ. Immunohistochemical and immunofluorescence staining After antigen retrieval, the sections were blocked with 1% goat serum and 0.1% Triton X-100 in PBS and then probed. primary and then secondary antibodies (see online supplemental Table 1). NISSL staining Mouse brains were harvested from 3-month-old WT, ANGPTL8 KO, 5xFAD, or ANGPTL8 KO 5xFAD mice(n ≥ 3 mice per group) under terminal anaesthesia for Nissl staining as reported previously [[92]26]. We collected different groups of mouse brains and fixed them with 4% paraformaldehyde in 1 M PBS at pH 7.4 for 24 h. Serial Sects. (20 µm) were dewaxed with graded alcohols (75%, 95%, 100% and 10 min each) and then brought into distilled water. The sections were then immersed in a 50% (weight/volume) aqueous solution of potassium metabisulfite for 10 min and rinsed with distilled water. Afterward, the sections were stained with a 1–2 aqueous solution of cresyl violet (Sangon Biotech, E607316) for 20 min at room temperature. The slices were progressively dehydrated in 70% ethanol once, 96% ethanol twice, and 100% ethanol for 5 min. Finally, the slices were cleared in xylene for 5 min and mounted in conventional neutral resin. The slices were subsequently observed with a microscope, and the cell density was subsequently determined in 3 fields per slice, with 3 slices per mouse. ELISA We assessed ANGPTL8 levels in serum, hippocampal tissue, and homogenates from WT and 5 × FAD mice (n = 3–5 mice per group) via a Mouse ANGPTL8/Betatrophin ELISA Kit (LSBio, LS-F56163) [[93]13], with some modifications. After the mice were anaesthetized with diethyl ether, they were transcardially perfused with PBS buffer containing a protease inhibitor. We carefully removed the brains and bisected them along the midline. The bilateral hippocampal and cortical tissues were separated on the basis of their anatomical location. The bilateral hippocampus and cortex were homogenized in NP-40 buffer (P0013F, Beyotime Institute of Biotechnology). After thorough grinding, the tissue homogenate was centrifuged at 5,000 × g for 10 min at 4°C, and the supernatant was collected. The protein concentration was determined via the BCA method. We followed the manufacturer's instructions to conduct the ELISA, and the interassay and intra-assay coefficients of variation for ANGPTL8 were < 15% and < 10%, respectively. The kit demonstrated high specificity for ANGPTL8 (ANGPTL8 kit, LSBio LS-F56163), with no significant cross-reactivity or internal interference. In situ hybridization mRNA in situ hybridization was performed via a PinpoRNA RNA in situ hybridization kit (GD Pinpoease Biotech Co. Ltd., Cat#: PIT0001) as previously reported [[94]27], with some modifications. Short probes, which are specifically complementary to the ANGPTL8 target RNA sequence covering regions 420–680, were designed via patented algorithms (CHINA patent number ZL202110581853.9). The brain and liver tissues were treated with pre-A solution to inhibit endogenous peroxidase activity for 10 min at room temperature. The samples were subsequently boiled with pre-B Solution B at 95–100°C for 15 min to recover the RNA binding site. Protease II treatment was used to expose the target RNA molecules, followed by probe hybridization for 2 h at 40°C. The signal was then sequentially amplified through reactions 1, 2, and 3 at 40°C for 15 min, followed by incubation at room temperature for reactions 4 and 5 (China patent number ZL202110575231.5) for 15 min in the dark. Finally, the slides were incubated at 37°C in Fast Red solution in a humidified chamber for 5 min and washed with PBS three times. Hematoxylin staining was used, and the slides were sealed to capture images under a Leica DM6B microscope with Leica Application Suite X software, where a positive signal was identified as precipitated red dots. Cell line and primary microglial culture The immortalized mouse BV2 microglial cell lines and Hippocampal Transformed cell line 22(HT22) (obtained from the China Centre for Type Culture Collection) were cultured in high-glucose DMEM supplemented with 10% FBS and 200 µM L-glutamine (Invitrogen Corporation) at 37°C in a humidified incubator supplied with 5% CO[2] balanced with air. In addition, mouse primary microglia were isolated and characterized as previously described [[95]28]. Briefly, primary glial cells were derived from neonatal mice aged 1–3 days. Their brains were extracted after decapitation and placed in sterile D-Hanks buffer on ice. After removal meninges and cerebellum, brain tissues were then cut into 1 mm × 1 mm pieces, mechanically homogenized, and filtered through a 100 μm nylon cell strainer with approximately 10–15 ml of sterile D-Hanks supplemented with 0.2% glucose into a 50 ml centrifuge tube. The homogenate was subsequently centrifuged at 1,000 rpm for 6 min at 4°C. The pellet was resuspended in DMEM supplemented with 10% FBS in a 15 ml poly-L-lysine-coated flask and incubated at 37°C with 5% CO[2] balanced with air for 24 h. Next, the culture medium was changed to high-glucose DMEM with 10% FBS and 25 ng/ml GM-CSF. After one week, the mixed glial cells were agitated at 80 rpm on a gyratory shaker for 20 min. For microglial isolation, floating cells were collected and reseeded in DMEM/F12 containing 10% FBS. Microglia were identified with immunocytochemistry assessment for ionizing calcium-binding adaptor molecule 1 (Iba1). Amyloid β (1–42) peptide and protein aggregation The amyloid β (1–42) peptide (BACHEM, 401,447) was initially dissolved in HFIP (1,1,1,3,3,3-hexafluoro-2-propanol) to achieve a concentration of 1,000 ng/ml. To generate A-Aβ, the method was performed as previously described [[96]26]. RNA interference For the RNA interference assay, the following siRNA target sequences were used: NLRP3 siRNA, 5'-GTACTTAAATCGTGAAACA-3', and NC. The siRNA sequence was 5'-TTCTCCGAACGTGTCACGTdTdT-3'. The PIRB siRNA sequence was 5'-CAAGGACATGGAGCAATGA-3', and the NC siRNA sequence was 5'-TTCTCCGAACGTGTCACGTdTdT-3'. All siRNAs were synthesized by RiboBio (Guangzhou, China) and transfected at a final concentration of 10 nM. Transmission electron microscopy BV2 cells were treated with Aβ1-42 prepared (BACHEM, 401,447, 400 ng/ml) or rANGPTL8 (recombinant human ANGPTL8, 600 ng/ml) for transmission electron microscopy (TEM) assessment (Hitachi H-7500 transmission electron microscope, Hitachi, H-7500, Japan) as previously described [[97]29]. Treated BV2 cells were fixed in 4% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.4, for 24 h at 4°C. The cells were subsequently rinsed in 0.1 M PBS for 20 min and then fixed in 1% osmium tetroxide in 0.1 M PBS for 2 h at room temperature. These samples were embedded in fresh Epon within predried molds after dehydrating through a graded series of ethanol. Subsequently, 200-nm-thick sections were stained with 2% uranyl acetate and alkaline lead citrate for 20 min. Transmission electron microscopy (TEM) was conducted using a Hitachi H-7500 transmission electron microscope (Hitachi, H-7500, Japan). Flow cytometry BV2 cells and mouse primary microglia were plated in 6-well dishes at a density of 1 × 10^5 per well and incubated with Aβ protein (400 ng/ml) and/or rANGPTL8 (600 ng/ml) for 6 h. The cells were washed, digested with trypsin, rinsed, and resuspended in 200 μL of cold PBS at a density of 1 × 10^5 cells/ml. Then, APC-conjugated monoclonal mouse CD86 antibodies and PE-conjugated monoclonal mouse CD206 antibodies were used to detect the corresponding cell membrane proteins. The cells were incubated with CD86 or CD206 antibodies in the dark for 30 min at room temperature and rinsed twice with PBS. Flow cytometry (Beckman Coulter Cytoflex ) was employed to analyse the light scatter characteristics of each sample (10^5 cells). The Q1-UL quadrant displayed CD206-positive cells, and the Q1-LR quadrant displayed CD86-positive cells. The value of CD86/CD206 was then calculated and statistically analysed after normalizing CD206-positive and CD86-positive cells in the treatment group to those in the control group. Phagocytosis assay In vitro phagocytosis assays were performed following established protocols [[98]30]. Primary microglia were plated in 12-well plates for 12 h. The cells were subsequently exposed to 1 g/ml aggregated Aβ1-42 (BACHEM, 4,014,447, Switzerland) with or without rANGPTL8 (6 µM). After 24 h, the cells were rinsed twice with PBS and digested with trypsin to obtain a cell suspension. The phagocytic index was assessed via flow cytometry. Another method was employed to assess the phagocytic activity of microglia. BV2 cells were treated with Aβ1-42 and/or rANGPTL8 for 4 h, followed by the introduction of fluorescent latex beads (4 µM) preopsonized in 50% FBS and PBS. BV2 cells were loaded with preopsonized beads at a concentration of 100 beads per primary microglia and incubated at 37°C for 2 h. The cells were subsequently washed twice with PBS to eliminate residual beads and fixed with 4% PFA at room temperature for 5 min. Iba1 was added to label the microglia, which were then stained with DAPI to count the number of cells. Images were acquired via a Zeiss Observer Z1 (LSM510 META) inverted microscope, which captured DAPI, Iba1, and bead fluorescence images. Image analysis was conducted via ImageJ software (version 2.0.0). Thanks to the Research Institute of Biomedicine at Hubei University of Medicine for providing large-scale equipment. Additional assessments qRT‒PCR We extracted total RNA from brain tissue or BV2 microglia via TRIzol (Invitrogen) reagent following the manufacturer's instructions. We subsequently synthesized cDNA via a reverse transcription kit (TIANGEN, Beijing, China). The mRNA levels were determined via quantitative real-time PCR via SuperReal PreMix QuantiNova™ SYBR® Green PCR Master Mix (QIAGEN, Germany). Each sample was analysed in triplicate, and relative expression levels were calculated via the 2^−ΔΔCt method with endogenous RNA for normalization. All primer sequences are listed in Supplementary Table 1. Western blotting Total protein was extracted from microglia and BV2 cells. The protein concentration was determined via a BCA kit. Proteins (40 µg) were separated via 10% sodium dodecyl sulfate‒polyacrylamide gel electrophoresis (SDS‒PAGE) and transferred to polyvinylidene fluoride (PVDF) membranes in transfer buffer. The membranes were blocked with 5% skim milk for 1 h at room temperature and then incubated with primary antibodies against cytosolic gasdermin-D (GSDMD)-FL (1:1000), GSDMD-N (1:1000), NLRP3 (1:1000), NFκB (1:1000), P-NFκB (1:1000), IL-6 (1:1000), cleaved caspase-1 (1:1000), β-tubulin (1:1000, Absin, Shanghai, China), β-actin (1:2000, Beyotime Biotechnology), and ANGPTL8 (1:1000, Sigma, St. Louis, Missouri, USA) at 4°C overnight. The next day, the membranes were incubated with a horseradish peroxidase (HRP)-conjugated secondary antibody (1:1000, Beyotime Biotechnology, Shanghai, China) diluted in TBS-T for 1 h at room temperature. Following three washes with TBS-T for 10 min each, the protein bands were detected with enhanced chemiluminescence (ECL). Images were captured via a gel imaging system (Bio-Rad GelDoc XR + , California, USA), and the band density was analysed with Fiji Image J software. We conducted three independent experiments. All the antibodies used are listed in Supplementary Table 2. Statistical analysis All data are presented as mean ± SEM. Individual data points identified as greater than two standard deviations away from the mean were designated as outliers and removed. Data were assessed for normality using a Shapiro-Wilks test. Normally distributed data were analysed by independent samples t-test, one-way ANOVA, two-way ANOVA, two-way repeated measures ANOVA or mixed effects model, as described for statistically significant data in relevant figure legends. Data not normally distributed were analysed by MannWhitney test or Kruskal–Wallis test, as described for statistically significant data in relevant figure legends. All statistical analyses were performed using GraphPad Prism 9, with p values < 0.05 considered significant. Results ANGPTL8 knockout improves cognitive function and reduces Aβ deposition in 5xFAD mice ANGPTL8 knockout (ANGPTL8 KO) led to a reduction in hair graying in the mice (Fig. S2A). Furthermore, after 9 months of normal diet feeding, the spatial learning and memory abilities of ANGPTL8 KO mice improved compared with those of WT mice, as demonstrated by a 56% reduction in escape latency and a 3.18-fold increase in time spent in the target quadrant (Fig. [99]1A-C). To investigate the impact of ANGPTL8 on AD development, we crossed the ANGPTL8-KO mouse line with the 5xFAD mouse line (Fig. [100]1D). The genotypes of the WT, 5xFAD, and ANGPTL8 KO mice were confirmed through PCR and sequencing analysis (Fig. S2B). Next, we evaluated the impact of ANGPTL8 KO on cognitive impairment via the Morris water maze test. Compared with 5xFAD mice, ANGPTL8 KO:5xFAD mice presented a significant increase in target crossings at both 3 (increased by 8.1-fold, p < 0.001) and 6 (increased by 10.6-fold, p < 0.05) months of age, along with a nearly 50% reduction in escape latency (Fig. [101]1E-K, Fig. S2C-D). Similar improvements in learning and memory were observed in the novel object recognition test (Fig. [102]1L-M). Importantly, there were no significant differences in swimming speed among the groups (Fig. S2E). The above groups of mice showed no significant sex differences in the results of the behavioral tests (Fig. S3A-L). These results clearly demonstrate that ANGPTL8 knockout ameliorates cognitive deficits in learning and memory in 5xFAD mice. Aβ accumulation is a hallmark pathological feature of AD, and 5xFAD mice exhibit extensive Aβ deposition. We directly evaluated the Aβ burden in the cortex and hippocampus via Aβ immunohistochemistry (Fig. [103]1N-P) and thioflavin-S (Thio-S) staining (Fig. S4A-F). No Aβ plaques were detected in the brains of WT or ANGPTL8 KO mice at 3 or 6 months of age (Fig. [104]1N). In contrast, 5xFAD and ANGPTL8 KO:5xFAD mice presented numerous Aβ-positive plaques in both the cortex and hippocampus. However, the number of Aβ plaques in ANGPTL8 KO:5xFAD mice was significantly lower than that in 5xFAD mice, with approximately half (61.7%, p < 0.01) the number observed in the latter (Fig. [105]1N‒P, Fig. S4A‒F). Notably, the difference in the number of Aβ plaques between the 5xFAD and ANGPTL8 KO:5xFAD groups became more pronounced with increasing age (Fig. S4A-F). Aβ plaques contribute to synaptic damage and neuronal loss in AD. Nissl staining revealed that ANGPTL8 knockout reduced neuronal loss in the cortex, dentate gyrus (DG), and CA3 region in 6-month-old 5xFAD mice (Fig. S5A-E). Consistent with these findings, immunofluorescence staining revealed significantly fewer neurons in the cortex, DG, and CA1 regions in 5xFAD mice than in ANGPTL8 KO:5xFAD mice (Fig. S5F-I). Collectively, these findings provide strong evidence that ANGPTL8 knockout significantly reduces Aβ deposition and neuronal loss in 5xFAD mice. ANGPTL8 promotes AD progression by inducing microgliosis and microglial activation To investigate the potential mechanisms underlying the role of ANGPTL8 in AD progression, we performed Gene Ontology (GO) biological process analysis via bulk RNA-seq data from the hippocampi of 5xFAD:ANGPTL8 KO and 5xFAD mice. The DEGs were predominantly involved in innate immune responses, Aβ clearance, and AD-related pathways (Fig. S6A-B). To further explore the effects of ANGPTL8 on brain cell populations, we conducted single-nucleus RNA sequencing (snRNA-seq) on 6-month-old 5xFAD:ANGPTL8 KO and 5xFAD mice (Fig. [106]2A-D). A total of 17 major cell clusters were identified in the cortical tissue of both groups (Fig. [107]2A-D, Fig. S6C-D). Notably, the proportion of microglia (cluster 3) in ANGPTL8 KO mice (7.52% of the total) was significantly lower than that in 5xFAD mice (12.97% of the total) (Fig. [108]2B-C). Microglia, as resident innate immune cells in the brain, undergo inflammatory polarization in response to pathological stimuli, leading to increased production of proinflammatory cytokines and increased immune activation [[109]31, [110]32]. On the basis of both the snRNA-seq and bulk RNA-seq results, we hypothesize that ANGPTL8 may accelerate AD progression by modulating the number and function of microglia. Sub-clustering of the microglial population revealed five distinct microglial subpopulations. ANGPTL8 KO significantly reduced the proportions of MG1 and MG2 clusters, which correspond to disease-associated microglia (DAM) states (inflammatory I and II) [[111]33] (Fig. [112]2D-F, Fig. S6E-F). Immunohistochemical analysis of Iba-1-positive microglia in the hippocampus of 5xFAD mice revealed a significant increase in activated microglia, characterized by shortened processes and amoeboid morphology. In contrast, ANGPTL8 KO:5xFAD mice presented fewer activated microglia, with no significant differences between the WT and ANGPTL8 KO groups (Fig. [113]2G-H, Fig. S6E-F). Double staining for Iba-1 and Aβ revealed that Iba-1-positive microglia were more abundant around Aβ plaques in the 5xFAD group than in the control group, whereas ANGPTL8 KO:5xFAD mice presented a marked reduction in activated microglia surrounding Aβ deposits. This effect was particularly evident in 6-month-old 5xFAD and ANGPTL8 KO:5xFAD mice (Fig. [114]2G-H, Fig. S6G-H). Collectively, these results suggest that ANGPTL8 knockout reduces microgliosis and microglial activation, thereby decreasing Aβ deposition in 5xFAD mice. Fig. 2. [115]Fig. 2 [116]Open in a new tab ANGPTL8 promotes AD progression by inducing microgliosis and microglial activation. A Changes in cortical cells in the brains of 6-month-old 5xFAD: ANGPTL8-KO and Con-5xFAD mice were analysed via single-nucleus RNA sequencing (snRNA-seq). Joint uniform manifold approximation and projection (UMAP), colored by total major cell type (17 main clusters) in mice. B-C. Global breakdown, region composition, enrichment and number of nuclei for 17 main cluster subtypes in the Con-5xFAD mouse group (B) and 5xFAD:ANGPTL8-KO mouse group (C). D-F. Subpopulations of microglia. UMAP embeddings of single nuclei profiled via snRNA-seq, colored according to clusters, and annotated for each subpopulation, indicating the proportion of each subpopulation in the Con-5xFAD mouse group (E) and the 5xFAD:ANGPTL8-KO mouse group (F). G. Immunostaining for Iba-1 and DAPI in the CA1 area of the hippocampus. Microglia were labelled with an Iba-1 antibody. H. Colocalization of Aβ and Iba-1 in the hippocampus of the mice in the indicated groups. Aβ plaques and microglia were labelled with anti-Aβ (red) and anti-Iba-1 antibodies (green), respectively Liver-specific expression of ANGPTL8 promotes AD progression by inhibiting microglial phagocytosis To investigate the origin of the effects of ANGPTL8 on brain microglia, we first performed alignment analyses of single-cell RNA-sequencing libraries from mouse brains ([117]GSE140511 dataset) and AD patient brains ([118]GSE157827 dataset). ANGPTL8 was expressed at very low levels in the brains of either normal or AD patients or in the brains of 5xFAD mice (Fig. S7A-D). Additionally, ANGPTL8 mRNA was very low in human clinical glioma samples, brain trauma tissue and primary neuronal cells or their cell lines (HT22 and SHSY5Y), but was undetectable in primary microglial or BV2 cell lines (Fig. S7 E–F and Fig. S8A-B). In situ hybridization (ISH) further confirmed that ANGPTL8 mRNA was nearly unexpressed in the brains of 6-month-old WT or 5xFAD mice, but was highly expressed in the liver (Fig. S8C). We subsequently detected the expression of ANGPTL8 of different nuclei (including the hippocampus, cortex, thalamus, pituitary gland and olfactory bulb) in the brains of WT and 5XFAD mice at 3, 6, and 9 months. The results revealed that ANGPTL8 expression was extremely low in the hippocampus and cortex, with no significant differences observed between WT and 5XFAD mice (Fig. S9A-W). These findings suggest that ANGPTL8 may be secreted by other tissues to promote AD progression. Consistent with previous studies indicating that ANGPTL8 is predominantly expressed in the liver [[119]19], we observed significantly higher ANGPTL8 mRNA in the livers of 3-, 6-, and 9-month-old 5xFAD mice than in those of WT mice (Fig. [120]3A-D), and the Aβ protein could upregulate ANGPTL8 expression in the hepatocyte cell line LO2 (Fig. [121]3E). Notably, we also found that the protein level of ANGPTL8 was greater in the serum of 3-, 6-, and 9-month-old 5xFAD mice than in that of WT mice (Fig. [122]3F-I). Furthermore, ANGPTL8 protein levels were markedly elevated in the hippocampus (1.67-fold, P < 0.05) and cortex (1.96-fold, P < 0.05) of 6-month-old 5xFAD mice compared with those of WT mice, as measured by ELISA (Fig. [123]3J-K). Therefore, we speculate that highly expressed ANGPTL8 in the liver may cross the blood–brain barrier and promote Aβ deposition in the brain. Our follow-up experiments revealed that hydrodynamic transfection of the ANGPTL8-GFP fusion plasmid into the mouse liver resulted in the presence of the ANGPTL8-GFP in the brain but not in control GFP-transfected mice, suggesting that ANGPTL8 in the circulation can cross the BBB (Fig. S10A‒C). These results support the hypothesis that the elevated expression of ANGPTL8 in the brain may originate from the peripheral liver. Fig. 3. [124]Fig. 3 [125]Open in a new tab Liver-derived ANGPTL8 activated microglia. A-D. ANGPTL8 mRNA was detected by qPCR in liver of 1, 3, 6 and 9 months old WT and 5xFAD mice. E. The effect of Aβ stimulation on ANGPTL8 expression was analyzed by qPCR in LO2 cells. F-I. Serum ANGPTL8 levels were detected by ELISA in 1-, 3-, 6-, and 9-month-old WT and 5xFAD mice (n ≥ 3). J-K. ANGPTL8 protein levels in the hippocampal and cortical homogenates of 6-month-old 5xFAD and WT mice were determined via ELISA. *P < 0.05, **P < 0.01, ***P < 0.001, ns = not significant. L. Schematic of the establishment of a model for long-term complementation of the liver expression of ANGPTL8 in ANGPTL8 KO:5xFAD mice via hydrodynamic tail vein injection of AAV8. M. Representative images of the tracks of 6-month-old ANGPTL8 KO:5xFAD-AAV8-con and ANGPTL8 KO:5xFAD-AAV8-ANGPTL8 mice in the Morris water maze (MWM) test. N‒O. Statistical analysis of the number of platform crossings (N) and escape latency (O) of 6-month-old ANGPTL8 KO:5xFAD-AAV8-con and ANGPTL8 KO:5xFAD-AAV8-ANGPTL8 mice (n = 8) on d 7. P‒R. IHC staining (P) and statistical analysis (Q‒R) of Aβ plaques per pixel area in the hippocampus and cortex of two groups of mice (ANGPTL8 KO:5xFAD-AAV8-con and ANGPTL8 KO:5xFAD-AAV8-ANGPTL8; n = 7–11 slices from three mice per group). S‒U. Thioflavin-S (S) and statistical analysis (T‒U) of Aβ plaques per pixel area in the hippocampus and cortex of two groups of mice (ANGPTL8 KO:5xFAD-AAV8-con and ANGPTL8 KO:5xFAD-AAV8-ANGPTL8; n = 7–11 slices from three mice per group). *P < 0.05, **P < 0.01, ***P < 0.001, ns = not significant To further assess the contribution of liver–specific expression of ANGPTL8 to AD pathology, we established a model for long-term liver-specific expression of ANGPTL8 in 2-month-old ANGPTL8 KO:5xFAD mice via hydrodynamic tail vein injection of AAV8-ANGPTL8 (or AAV8 control virus). The mice were injected twice, once every 2 months. At 6 months of age, we performed behavioral and pathological analyses. Compared with the control groups (ANGPTL8 KO:5xFAD-AAV8-con and ANGPTL8 KO:5xFAD-AAV8-ANGPTL8) mice, that expressed liver-derived ANGPTL8 exhibited significant impairments in spatial learning and memory, as evidenced by a 55% reduction in escape latency and a 1.73-fold increase in time spent in the target quadrant (Fig. [126]3L-O). Additionally, compared with control group mice, ANGPTL8 KO:5xFAD-AAV8-ANGPTL8 mice presented a greater number of Aβ plaques in the cortex (1.74-fold, P < 0.05) and hippocampus (1.26-fold, P < 0.05) (Fig. [127]3P-U). Nissl staining further revealed that neurons in the cortex (0.831-fold, P < 0.05), dentate gyrus (DG, 0.823-fold, P < 0.05), CA1 (0.78-fold, P < 0.01) and CA3 (0.66-fold, P < 0.05) regions of ANGPTL8 KO:5xFAD-AAV8-ANGPTL8 mice were thinner than those in the control groups (Fig. S10D-H). Together, these findings demonstrate that liver-derived ANGPTL8 accelerates AD progression by inhibiting microglial phagocytosis and increasing Aβ accumulation. ANGPTL8 induces microglial activation by promoting pyroptosis To further investigate the impact of ANGPTL8 on microglial function, we treated cultured BV2 cells with recombinant ANGPTL8 protein (rANGPTL8, 600 ng/ml). Compared with control BV2 cells, rANGPTL8-treated BV2 cells exhibited significant morphological changes, transitioning to an amoeboid and dystrophic appearance (Fig. [128]4A). We next assessed microglial phagocytic activity and found that the phagocytic index was lower in Aβ-treated microglia (both BV2 cells and primary microglia) than in vehicle-treated control microglia. Notably, rANGPTL8 further exacerbated the inhibitory effect of Aβ on microglial phagocytosis, increasing it by approximately twofold (Fig. [129]4B-E). A marked reduction of approximately 50% in phagocytic capacity was also observed in primary microglia and BV2 cells following co-treatment with rANGPTL8 and Aβ, as indicated by a decrease in the number of cells containing phagocytosed FITC-labelled beads (Fig. [130]4F-I). Collectively, these findings suggest that ANGPTL8 potentiates Aβ-induced microglial dysfunction. Fig. 4. [131]Fig. 4 [132]Open in a new tab ANGPTL8 reduces microglial phagocytosis to inhibit the clearance of Aβ. A Morphological changes (resting ramified microglia and activated amoeboid microglia) in BV2 cells treated with rANGPTL8 (600 ng/ml) for 6 h under a bright field. Scale bar: 50 μm. B-E. Flow cytometry analysis of the phagocytic ability of fluorescence microspheres in BV2 cells or primary microglia after treatment with Aβ1-42 (400 ng/ml) and/or ANGPTL8 (600 ng/ml) for 6 h. Representative scatter plots showing the frequencies of the cells containing fluorescent microspheres (B, C) and summary bar graphs (D, E). F‒I. Microscopic analysis of the phagocytosis of fluorescent microspheres by BV2 cells or primary microglia after treatment with Aβ1‒42 (400 ng/ml) and/or ANGPTL8 (600 ng/ml) for 6 h. Representative images of fluorescent microspheres (green) engulfed by BV2 cells (F) or primary microglia (G) (marked with anti-Iba1 (red) and DAPI for the nucleus (blue)) and a summary bar graph (H‒I). Scale bar: 50 μm. *P < 0.05, ** P < 0.01, *** P < 0.001, ****P < 0.0001 To elucidate the underlying mechanism, we performed KEGG pathway enrichment analysis using bulk RNA-seq data from 5xFAD:ANGPTL8 KO and 5xFAD mice. A key finding was a significant alteration in the NF‒kB signalling pathway between these two groups (Fig. [133]5A‒B). NF-kB is a well-known mediator of inflammation that is crucial for microglial activation and is associated with changes in microglial morphology, cytokine release, neurotoxic mediator production, migration, and phagocytosis [[134]34]. snRNA-seq data also confirmed that ANGPTL8 regulates the NF-kB pathway in microglia (Fig. S11A). We found that proinflammatory and promigration factors were highly expressed in the microglia of 5xFAD mice, whereas anti-inflammatory markers were enriched in 5xFAD:ANGPTL8 KO microglia (Fig. [135]5C-E). Given these findings, we focused on the expression of proinflammatory factors. We observed that the combination of rANGPTL8 and Aβ1-42 significantly increased the levels of phosphorylated NF-kB and NLRP3 by 1.52-fold in BV2 cells, similar to LPS + PMA (12-O-tetradecanoylphorbol 13-acetate) treatment (Fig. [136]5F). Furthermore, rANGPTL8 upregulated the mRNA levels of NLRP3, IL-1β, and markers of reactive microglia (pro-inflammatory phenotype), such as CD86 and iNOS, in a time-dependent manner (Fig. S11B-E). Notably, co-treatment with rANGPTL8 and Aβ1-42 synergistically increased iNOS mRNA levels and decreased Arg-1 mRNA levels approximately 1.5-fold (Fig. S11F-G). Additionally, the ratio of CD86/CD206, indicative of the pro-inflammatory/anti-inflammatory phenotypes of reactive/homeostatic microglia, was elevated in response to rANGPTL8 and/or Aβ1-42 (Fig. [137]5G-H). These findings suggest that rANGPTL8, in combination with Aβ, promotes microglial activation, particularly by amplifying the transition from a resting phenotype to an activated phenotype. Fig. 5. [138]Fig. 5 [139]Open in a new tab ANGPTL8 induced inflammatory factor expression and microglial polarization. A. Volcano plot visualizing the differential expression of the genes of the heart transcriptome (n = 3 in each group). B. Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis of the identified differentially expressed genes on the basis of the RNA-seq dataset from the hippocampi of 6-week-old 5 × FAD (WT) and ANGPTL8-KO:5 × FAD mice (n = 3 in each group). C. Representative immunoblots probed with antibodies against NLRP3, p-NF-κBp65, NF-κBp65, and β-tubulin and the statistical results of Western blotting. D-F. UMAP gene expression patterns of NF-κB pathway markers in 17 labelled clusters (D). Gene expression patterns of NF-κB pathway markers in UMAPs of microglial subpopulations (E). Distinct gene expression patterns of NF-κB pathway markers in microglial subpopulations in 5 × FAD (WT) and ANGPTL8-KO:5 × FAD mice. The dot color shows the mean expression in expressing cells (column scale); the dot size shows the percentage of cells expressing the gene (F). G. Representative images of CD86 and CD206 expression detected by flow cytometry in the indicated groups. H. Quantification of CD86/CD206 in BV2 cells in the indicated groups. LPS (500 ng/ml) + PMA (500 ng/ml) was used as a positive control. *P < 0.05, ** P < 0.01, *** P < 0.001, ****P < 0.0001 Interestingly, our results revealed a distinctive phenomenon: treatment with rANGPTL8 and/or Aβ1-42 for 36 h induced pyroptotic features in BV2 cells (Fig. [140]6A-C). The combination of rANGPTL8 and Aβ1-42 led to a significant increase in pyroptotic cells (Fig. [141]6A-C). Pyroptosis, a form of programmed cell death characterized by cell lysis, facilitates inflammasome activation (including that of NLRP3) and the release of inflammatory factors such as IL-1β, contributing to neurotoxicity in AD [[142]5, [143]35]. The pyroptosis marker GSDMD-N was significantly elevated approximately 2.1-fold in the hippocampi of 3-month-old 5xFAD mice, but this increase was attenuated nearly 1.4-fold in 5xFAD:ANGPTL8 KO mice (Fig. [144]6D-E). Furthermore, rANGPTL8 treatment upregulated the levels of GSDMD-N, NLRP3, cleaved caspase-1 (c-caspase-1), and phosphorylated NF-kB (p-NF-kB) in a time-dependent manner (Fig. S12A-B). Compared with 5xFAD mice, 5xFAD:ANGPTL8 KO mice presented nearly onefold lower levels of NLRP3 and ASC in the hippocampus and cortex (F [145]ig. [146]6F-I). The colocalization of NLRP3 and ASC was observed in Iba1-positive microglia (Fig. [147]6J-M). SnRNA-seq confirmed that NLRP3 was predominantly expressed in the microglial cluster, which is consistent with prior reports indicating that NLRP3 inflammasome-induced GSDMD pyroptosis occurs primarily in AD microglia. In vitro, rANGPTL8 treatment also increased GSDMD-N, ASC, and cleaved caspase-1 in BV2 cells exposed to Aβ1-42 (Fig. [148]6N-O). Interestingly, GSDMD-N cleavage was even more pronounced when ANGPTL8 expression was restored in the livers of 5xFAD:ANGPTL8 KO mice (Fig. S12C-D). These results strongly suggest that hepatic ANGPTL8 promotes microglial activation and the release of proinflammatory cytokines by inducing pyroptosis. Fig. 6. [149]Fig. 6 [150]Open in a new tab ANGPTL8-induced microglial pyroptosis. A. Representative morphological images of pyroptosis in BV2 cells treated with Aβ1-42 (400 ng/ml) and/or rANGPTL8 (600 ng/ml) for 36 h. B. Statistical analysis of pyroptotic cell numbers/mm.^2 in each group. C. Representative morphological images of the indicated groups observed via transmission electron microscopy. LPS (500 ng/ml) + PMA (500 ng/ml) was used as a positive control. D. Quantification of the levels of GSDMD-N normalized to those of β‐tubulin. Experiments were performed in the indicated groups (n = 3 mice/group) E. Representative western blot bands of GSDMD-FL, GSDMD-N, and β-tubulin in 3-month-old mice in each group. F-I. IHC staining (F&H) and statistical analysis (G&I) of NLRP3 and ASC in the hippocampus and cortex of four groups of mice (WT, ANGPTL8 KO, 5xFAD, and ANGPTL8 KO:5xFAD). N = 7 to 11 slices from three mice per group, *P < 0.05, ** P < 0.01, *** P < 0.001, ns = not significant. J‒K. Colocalization of ASC and Iba-1 (J) or NRLP3 and Iba-1 (K) in the hippocampus of the mice in the indicated groups. L. Representative immunoblots probed with antibodies against GSDMD-FL, GSDMD-N, cleaved caspase-1, ASC and β-tubulin in BV2 cells treated with Aβ1-42 (400 ng/ml) and/or rANGPTL8 (600 ng/ml) for 36 h. M‒O. The levels of GSDMD-N, cleaved caspase-1, and ASC were quantified and normalized to those of β-tubulin. GSDMD-FL represents full-length GSDMD, and c-caspase-1 represents cleaved caspase-1. *P < 0.05, ** P < 0.01, *** P < 0.001, ****P < 0.0001 ANGPTL8 promotes microglial pyroptosis via the PirB/NLRP3 pathway To investigate the role of the NLRP3 inflammasome in ANGPTL8-induced microglial pyroptosis, we treated microglia with NLRP3-specific siRNA. Knockdown of NLRP3 abolished the nearly 0.5-fold effect of rANGPTL8 on GSDMD-N levels (Fig. [151]7A-C). Since ANGPTL8 is secreted and not expressed in microglia, we next explored potential receptors on the microglial cell membrane. PirB, a transmembrane receptor that is upregulated in CNS injury and linked to axonal regeneration [[152]36], has also been implicated as an Aβ receptor [[153]37]. Our previous work demonstrated that ANGPTL8 increased PirB expression and bound to PirB [[154]21]. Thus, we examined the role of PirB in ANGPTL8-induced activation of the NLRP3/GSDMD pathway. Both Aβ1-42 and rANGPTL8 increased the mRNA levels of PirB (2.37-fold, P < 0.001) and PirA2 (2.89-fold, P < 0.001) in BV2 cells (Fig. S13A-B). When PirB siRNA was applied, the rANGPTL8-induced nearly onefold increases in the NLRP3 and GSDMD-N protein levels were significantly reduced in BV2 cells (Fig. [155]7D-F), and the immunofluorescence results of primary microglia further demonstrated that the knockdown of PirB counteracted the rANGPTL8-induced upregulation of the pyroptosis-related key genes NLRP3 and ASC (Fig. [156]7G-I). Moreover, PirB siRNA reversed the inhibitory effect of ANGPTL8 on microglial phagocytosis in BV2 or primary microglia, as evidenced by an increase in the number of cells containing phagocytosed FITC beads (Fig. [157]7J-N). These findings demonstrate that ANGPTL8 inhibits microglial phagocytosis by promoting pyroptosis via the PirB/NLRP3 pathway. Fig. 7. Fig. 7 [158]Open in a new tab ANGPTL8 promoted microglial pyroptosis via the PirB/NLRP3 pathway. A. To analyse the role of NLRP3 in BV2 cell pyroptosis after treatment with ANGPLT8 and/or Aβ1-42, a siRNA strategy was used to manipulate the expression of NLRP3 in BV2 cells. Immunoblots of NLRP3, GSDMD-FL, GSDMD-N, and β-tubulin (loading control) in BV2 cells and NLRP3-knockdown cells treated with Aβ1-42 (400 ng/ml) and/or rANGPTL8 (600 ng/ml) for 6 h. LPS (500 ng/ml) + PMA (500 ng/ml) was used as a positive control. B-C. The levels of NLRP3 and GSDMD-N were quantified and normalized to those of β-tubulin. D. To analyse the role of PirB in BV2 cell pyroptosis after treatment with ANGPLT8 and/or Aβ1-42 by western blotting, a siRNA strategy was used to manipulate the expression of PirB. E–F. Quantification of NLRP3 and GSDMD-N in BV2 and PirB-knockdown BV2 cells treated with Aβ1-42 with or without rANGPTL8. G-H. Representative images of NLRP3 and ASC expression detected by immunofluorescence in the indicated groups. I‒J. Flow cytometry analysis of the phagocytic ability of fluorescent microspheres in BV2 and PirB-knockdown BV2 cells after treatment with ANGPTL8 (600 ng/ml) for 6 h. Representative scatter plots showing the frequencies of the cells containing fluorescent microspheres (I) and summary bar graphs (J). K‒N. Confocal microscopic examination of the phagocytosis of fluorescent microspheres by BV2 cells or primary microglia and PirB-knockdown BV2 cells or primary microglia after treatment with ANGPTL8 for 6 h. Representative images (K-L) of fluorescent microspheres (green) engulfed by BV2 cells or primary microglia (marked with anti-Iba1 (red) and DAPI for nuclei (blue)) and a summary bar graph (M–N) are shown Metformin improves learning and memory in 5xFAD mice by inhibiting ANGPTL8 expression Our study revealed a novel role for ANGPTL8 in AD development, suggesting that targeting ANGPTL8 expression may be a promising therapeutic strategy. To identify compounds that inhibit ANGPTL8, we screened drugs for AD treatment and found that metformin significantly reduced ANGPTL8 expression in LO2 cells (Fig. S14A-E). Metformin also decreased ANGPTL8 mRNA levels approximately fourfold in LO2 cells and nearly onefold in the liver tissue of 5xFAD mice (Fig. S14F-G). As a well-established FDA-approved drug, metformin can cross the BBB and influence the central nervous system. We evaluated the therapeutic potential of metformin in AD via four groups of mice: WT, ANGPTL8 KO, 5xFAD, and ANGPTL8 KO:5xFAD. Metformin (300 mg/kg) was administered for 4 weeks (Fig. [159]8A). Metformin treatment significantly improved spatial learning and memory in the 5xFAD group, as evidenced by better performance in the Morris water maze (approximately 4.5-fold greater number of platform crossings and recognition indices and approximately 0.7-fold greater escape latency) than in the control group. No significant differences were observed between the metformin-treated WT, ANGPTL8 KO, and ANGPTL8 KO:5xFAD groups (Fig. [160]8B-F). Additionally, the serum ANGPTL8 level was nearly 0.3-fold lower in the 5xFAD + Met group than in the 5xFAD group (Fig. [161]8G). Furthermore, Tio-S staining revealed that metformin reduced Aβ plaque deposition in the hippocampus and cortex of 5xFAD mice. The effect was more pronounced in ANGPTL8 KO:5xFAD mice (Fig. [162]8H-J). These findings underscore the potential of metformin to improve cognitive function in AD patients, likely through the inhibition of ANGPTL8 expression in vivo. Fig. 8. [163]Fig. 8 [164]Open in a new tab Metformin improved cognitive dysfunction in 5xFAD mice by inhibiting ANGPTL8. A. Schematics of the experimental procedures for the eight groups of mice. B. A representative locus plot of the MWM test results of 4-month-old WT, ANGPTL8KO, 5 × FAD, ANGPTL8KO: 5 × FAD mice that received without or with Met (300 mg/kg, i.g.) for 4 weeks is shown. C‒E. The swimming speed, number of platform crossings, and escape latency of the mice in the indicated groups were analysed during the Morris water maze test. F. The recognition index was detected by a novel object recognition experiment in 4-month-old mice in the indicated groups. G. Serum ANGPTL8 levels were detected by ELISA in 4-month-old 5 × FAD mice without or with Met (300 mg/kg, i.g.) treatment for 4 weeks. H. Amyloid plaques were detected by thioflavin-S staining (scale bars, 100 μm) in the cortex and hippocampus of the brains of 4-month-old mice from the indicated groups. I‒J. Quantitative analysis of amyloid plaques in the cortex (I) and hippocampus (J) in each group (the average value of 8‒10 slices from each mouse, n = 3 mice/group), *P < 0.05, ** P < 0.01, ns = not significant. K. Schematic representation of the proposed mechanism by which ANGPTL8, as a liver-secreted inflammatory trigger, aggravates microglial pyroptosis via the PirB/NLRP3 pathway to accelerate the pathogenesis of AD. The elevated Aβ levels in the AD model of 5 × FAD mice upregulated liver ANGPTL8 expression, and secreted ANGPTL8 acted as an inflammatory trigger and directly interacted with the microglial membrane receptor PirB to activate downstream PirB/NLRP3 signalling. Furthermore, the NLRP3 inflammasome initiates cleaved GSDMD to release a GSDMD-N fragment that forms pores on microglia, leading to the extracellular release of inflammatory factors such as IL-1β. Metformin inhibited the expression of ANGPTL8 and decreased Aβ deposition in 5 × FAD mice Discussion Our study underscores the pivotal role of liver-derived ANGPTL8, a secreted protein, in promoting the progression of Alzheimer’s disease (AD). Specifically, we demonstrate, for the first time, that ANGPTL8 not only exacerbates the deposition of amyloid-beta (Aβ) but also acts as a proinflammatory mediator, potentiating neuroinflammation through microglial activation and pyroptosis via the PirB/NLRP3 pathway to accelerate the pathogenesis of AD. Targeting liver-derived ANGPTL8 may hold great potential for developing therapies for AD. Our data revealed that ANGPTL8 KO improved cognitive function and significantly reduced Aβ deposition in 5xFAD mice. These mice, which are known for their rapid Aβ accumulation and plaque formation, typically exhibit early signs of Aβ deposition and gliosis by 2 months, followed by neuronal loss at approximately 4 months [[165]25]. This model not only recapitulates the early-stage pathological mechanisms of AD but also serves as a valuable tool for investigating the role of ANGPTL8 in AD pathogenesis [[166]24]. Consistent with these observations, our data demonstrated that ANGPTL8 KO slows the progression of Aβ deposition in 5xFAD mice at both 3 and 6 months. Interestingly, single-cell sequencing libraries and our results revealed very low expression of ANGPTL8 in normal or AD patient brains. Studies have shown that under normal dietary conditions, ANGPTL8 is expressed at low levels in neurons of the brain [[167]38, [168]39]. However, a high-fat diet significantly upregulates ANGPTL8 expression in the hippocampus and accelerates diabetes-related cognitive impairment [[169]39]. Therefore, we hypothesized that ANGPTL8 expression in the brain promotes the progression of AD. However, in contrast to WT mice, ANGPTL8 expression in various brain nuclei of 5XFAD mice does not significantly increase with age, suggesting that brain-derived ANGPTL8 may not promote AD progression. Previous studies have shown that the liver and adipose tissue are the primary sites of ANGPTL8 expression, but only the ANGPTL8 protein expressed by the liver has secretory properties [[170]40]. Our results further support the notion that highly expressed ANGPTL8 in the liver could permeate the cerebral cortex and hippocampus, reinforcing the hypothesis that peripheral ANGPTL8 contributes to AD progression. To understand how ANGPTL8 influences AD pathology, we performed bulk RNA-seq and snRNA-seq in the hippocampus and cortex of ANGPTL8 KO and 5xFAD mice. Our results indicated that ANGPTL8 KO reduced Aβ deposition by suppressing microgliosis and microglial activation. Notably, ANGPTL8 was found to regulate several genes related to innate immunity, with the NLRP3 inflammasome gene being particularly significant. Pyroptosis, a form of programmed cell death characterized by inflammatory cytokine release, is known to be driven by the NLRP3 inflammasome, especially in the context of AD [[171]41]. Recent research has shown that pyroptosis, which is mediated by the caspase-1/GSDMD pathway, contributes to microglial activation and neuroinflammation, exacerbating AD progression [[172]42]. In our study, we found that NLRP3 is expressed predominantly in microglia, which is consistent with previous reports suggesting that NLRP3 inflammasome-induced pyroptosis is most pronounced in microglia in AD [[173]5]. Our results indicate that ANGPTL8-induced microglial pyroptosis is also dependent on NLRP3 expression and that its promotion of AD progression likely occurs through the stimulation of microglial activation, the induction of pyroptosis through the NF-κB/NLRP3 pathway, and the impairment of microglial phagocytic capacity, which further over-activates microglia to accelerate Aβ plaque deposition. The overactivation of microglia is closely related to immune system disorders [[174]3, [175]4], and a new theory has been proposed that AD is not merely a brain disease but essentially a chronic autoimmune disease (AD2 theory) [[176]43]. Under the stimulation of various pathogens and injury factors, Aβ is synthesized and released as an early immune peptide [[177]44], triggering an innate immune cascade that promotes the continuous activation of microglia and the secretion of proinflammatory cytokines, leading to nonspecific autoimmune inflammation in the brain and resulting in neuronal death [[178]45]. This innovative theory also explains why transplanting peripheral blood monocytes from normal mice into the brains of AD mice causes monocytes to quickly become overactivated, similar to microglia, and lose their ability to phagocytose Aβ [[179]46]. However, whole blood exchange therapy can effectively reduce the abnormal activation of microglia and decrease the formation of Aβ plaques in the mouse brain, significantly delaying the progression of AD [[180]47]. Our findings also highlight a positive feedback loop between Aβ and ANGPTL8. We demonstrated that Aβ stimulates the expression of ANGPTL8, which is then secreted into the bloodstream, crosses the BBB, and interacts with microglial receptors to exacerbate neuroinflammation. Moreover, our study underscores the pivotal role of ANGPTL8 in promoting Aβ-induced microglial dysfunction, which contributes to cognitive deficits in AD. Aβ can exit the brain and enter the peripheral circulation through various routes, including the BBB and lymphatic pathways [[181]48]. The dynamic equilibrium between the brain and peripheral Aβ, even in the presence of an intact BBB, can be triggered via elevated circulating Aβ levels, further prompting the liver to secrete ANGPTL8. Given that ANGPTL8 is a secretory protein not expressed by primary microglia or BV2 cells, as shown in our study, we investigated potential receptors for ANGPTL8 on microglial cell membranes. Our studies revealed PirB expression in both primary microglia and BV2 cells. PirB, along with LILRB2, has been recognized as an Aβ receptor, and blocking this receptor has therapeutic potential for AD treatment [[182]36]. Our previous work revealed that ANGPTL8 binds to LILRB2/PirB [[183]19]. Thus, our current study provides further evidence that ANGPTL8 promotes microglial pyroptosis via the PirB/NLRP3 pathway, accelerating AD progression. However, the exact interaction between ANGPTL8 and the LILRB2 receptor is unknown and warrants further study. Our study also revealed that metformin is a potential therapeutic strategy for AD, as it inhibits ANGPTL8 expression. Metformin has been shown to have anti-AD effects, improving learning and memory in AD models and patients [[184]49–[185]51]. In our study, we found that metformin, an ANGPTL8 inhibitor, delayed the progression of AD. Furthermore, its effects on AD are not effective after the knockout of ANGPTL8, suggesting that the anti-AD effect of metformin is partially mediated by ANGPTL8 inhibition. Our study and other previous studies revealed that the serum levels of ANGPTL8 are elevated in patients with diabetes [[186]18, [187]52] and nonalcoholic fatty liver disease-related liver fibrosis [[188]19, [189]53], and glycolipid metabolism disorders caused by these diseases are risk factors for AD [[190]54]. Therefore, our work may also suggest that if ANGPTL8 is highly accumulated in the serum of patients at an early stage, metformin may have some benefits for patients who suffer from these diseases. While our research provides valuable insights into the role of ANGPTL8 in AD pathogenesis, future studies should explore its effects on other cell types, including neurons, astrocytes, and oligodendrocytes. Although we found that ANGPTL8 expression in the brains of 5xFAD mice, particularly in neurons, is very low and that this study focuses primarily on how high hepatic expression of ANGPTL8 promotes AD pathogenesis, it does not exclude the possibility that other factors may trigger neuron- and liver-derived ANGPTL8 to jointly drive AD progression. Therefore, as ANGPTL8 is predominantly expressed in peripheral tissues, understanding how it crosses the blood–brain barrier and regulates interactions between microglia and neuronal cells in the immune microenvironment of the brain will be crucial for developing targeted therapies. Our study demonstrated that brain-derived Aβ promotes hepatic ANGPTL8 expression, but the functional regulation of ANGPTL8 by Aβ originating from other tissues requires further investigation. Additionally, we identified metformin as a suppressor of ANGPTL8 expression; metformin has numerous targets in vivo and is not an ANGPTL8-specific inhibitor. It may also delay aging-related cognitive impairment through various mechanisms, such as modulating the AMPK pathway and gut–brain axis to influence metabolism and neuroinflammation [[191]55, [192]56]. Therefore, future efforts should focus on screening for ANGPTL8-specific inhibitors. In conclusion, our study provides evidence that ANGPTL8 accelerates AD progression by modulating the immune response, specifically through microglial activation and pyroptosis. Aβ may stimulate hepatic ANGPTL8 expression, which then crosses the BBB and activates microglia via the PirB/NLRP3 pathway, impairing microglial function and exacerbating Aβ deposition per se (Fig. [193]8K). Notably, our findings suggest that inhibiting ANGPTL8 with metformin or other developed strategies may be promising therapies for AD. Our findings also illustrate how the axis of the peripheral organ-secreted factor ANGPTL8/brain microenvironment/microglia mediate neuroimmune disturbances to promote the development of AD. Therefore, our study provides novel insight into disease pathogenesis, diagnosis, and intervention for AD from a systemic perspective and additional evidence for the targeting of peripheral Aβ for drug development. Supplementary Information [194]Supplementary Material 1.^ (15.2KB, docx) [195]Supplementary Material 2.^ (14.3KB, docx) [196]Supplementary Material 3.^ (16.3MB, docx) [197]Supplementary Material 4.^ (328.7MB, docx) Authors’ contributions Qiufang Zhang and Xingrong Guo designed the work and wrote and revised the main manuscript text. Jiarui Wei, Lin Hu, Shufan Xu, Fan Yang and Xin Shen prepared all the figures; Yifan Li performed some behavioral tests, such as the MWM test; Fusheng Liao, Xiaoqiao Zhang, and Xinggang Fang helped prepare some slides; Li Ding helped feed the mice and drug administration; Zhuo Chen and Shanchun Su performed some of the PCRs and WBs; Yong Huang and Junhua Cheng performed single-cell sequencing analysis; and Qian Chen and Daqing Ma performed further data and writing validation. All the authors reviewed the manuscript and approved it for publication. Funding This research was funded by the National Natural Science Foundation of China (No. 82371596, 82471628), the Project of Creative Research Groups of Hubei Province (2023AFA023), the Advantages Discipline Group (Medicine) Project in Higher Education of Hubei Province (2025XKQZ551), the Natural Science Foundation of Hubei Province (2022CFB446), the Hubei Province Science and Technology Department of Natural Science Foundation (2024AFB787, 2024AFD324), Hubei Provincial Science and Technology Plan Project (2024BSB017,2023B3B137), the the Administration Bureau of Traditional Chinese Medicine of Hubei Province (ZY2023M073), the Excellent Young and Middle-aged Scientific and Technological Innovation Team of Universities of Hubei Province (T2022022), and the Research Projects at Taihe Hospital (2024JJXM016). Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate All experiments involving animals were conducted according to ethical policies and procedures approved by the ethics committee of Hubei University of Medicine, China. The ethical committee number for the study was 202016. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Jiarui Wei, Lin Hu and Shufan Xu contributed equally to this work. Contributor Information Qiufang Zhang, Email: zqf1112000@163.com. Xingrong Guo, Email: gxrdl@126.com. References