The human study was performed according to the principles of the Helsinki declaration and King’s College London. All participants gave their written informed consent before participating in the study. The regional Ethics Committee for medical research in the South-East of Norway (REK 2011/2052, REK 2017/371 and REK 82685) and the Data Protection Officer at our institution approved the study. All experiments and animal were approved by the Jinan University Institutional Animal Care and Use Committee (IACUC-20181029-01).

Study participantsMemory Clinic Cohort (patients with AD)

We included 316 patients with AD including 56 patients with AD-MCI and 260 AD-dementia. AD patients were recruited from the memory clinics at Oslo University Hospital (n = 180) and St. Olav’s University Hospital, Trondheim (n = 136) during the period from 2010 to 2018. The patients were included in the Norwegian Registry of Persons Assessed for Cognitive Symptoms (the NorCog registry) and went through a comprehensive clinical assessment following a standardized research protocol81 including an interview with the patients and their caregivers, cognitive testing, physical examination, blood sampling, imaging (computed tomography, magnetic resonance imaging, fluorodeoxyglucose-positron emission tomography) and lumbar puncture. All patients met the clinical criteria of probable or possible AD or AD mixed with cerebrovascular disease according to the US National Institute of Aging and the Alzheimer’s Association criteria82,83. Patients who underwent a minimum of one follow-up at the clinics after the baseline examination were included in the progression analyses (n = 254). Cognitive and functional impairment was assessed by the CDR scale84 post hoc by certified CDR raters based on all the information from the clinical records. The categories of memory, orientation, judgment and problem-solving, community affairs, home and hobbies and personal care were given a score of 0–3 (higher values indicating more severe impairment) and summed up to give the CDR-SB (0–18)85,86. The AD core CSF biomarkers were analyzed at Akershus University Hospital (Ahus) by enzyme-linked immunosorbent assays (ELISA; Innotest hTau Ag, phoshoTau (181P) and β-amyloid 1–42 Fujirebio Europe). Specific cut-offs provided by the laboratory were applied and used when categorizing the patients according to the AT (N) classification87: A, Aβ42 > 700 pg ml−1; T, p-tau181 < 80 pg ml−1; and age-adjusted cut-off concentrations for N, total tau < 300 pg ml−1 for patients under the age of 50, 450 pg ml−1 for those aged 50–70 years and < 500 pg ml−1 for those older than 70 years.

CU controls (the COGNORM study)

Additionally, 75 CU older controls were recruited when referred to hospital for elective gynecological orthopedic or urological surgery. These patients were assessed with the same battery of cognitive tests as the memory clinic patients, and the majority were cognitively followed yearly for up to 5 years. At baseline, CSF was collected in conjunction with spinal anesthesia at the time of surgery. Repeated CSF sampling was performed in a subgroup of CU controls who volunteered for a second lumbar puncture (n = 22) after 4.48 years (0.48 s.d.). Further details of this cohort are available elsewhere88. CSF AD core biomarkers were analyzed in the Sahlgrenska University Hospital (Mölndal, Sweden) by ELISA (Innotest hTau Ag, phoshoTau (181 P) and β-amyloid 1–42 Fujirebio Europe). Their laboratory-specific cut-offs were applied: A, Aβ42 > 530 pg ml−1; T, p-tau181 < 60 pg ml−1; and N, total tau < 350 pg ml−1 (ref. 89). These cut-offs were applied when categorizing the CU controls according to the AT (N) classification87.

CSF and serum ULK1 measurements

Aliquoted CSF and serum samples from the Memory Clinic Cohort and CU controls were transferred from Oslo University Hospital on dry ice and were stored at −80 °C until assays were performed at the Department of Clinical Molecular Biology, Akershus University Hospital (EpiGen). ULK1 protein was quantified in CSF and serum using commercially available sandwich ELISA Kits (Nordict Biosite, cat. no. EKX-PZ9LPK-96) following the manufacturer’s instructions. Samples were processed after the same freeze–thaw cycle and plated at twofold to fourfold dilution. Each assay plate also included multiple dilutions of manufacturer-provided calibration standards. Protein concentrations (pg ml−1) were calculated using appropriate calibration standards. Log transformations of data were applied, as needed, to calculate serum ULK1.

Bioinformatics analysis

RNA-seq data were obtained from publicly accessible MIT datasets34. We processed the single-cell raw data using the Seurat package. Specifically, we kept genes detected in more than 2 cells and kept all cells with at least 200 detected genes. We identified outlier cells based on multiple quality metrics, including unique gene count, the ratio of mitochondrial RNAs relative to endogenous RNAs and total gene count. Cells that passed quality control were log-normalized using a scale factor of 10,000. The Louvain algorithm in Seurat was implemented to detect clusters. The cell type of each cluster was identified from corresponding marker genes as described elsewhere87.

Human brain immunofluorescence and quantification

The experiments were performed similarly to what we have published elsewhere76. Human brain sections (7 µm) were rehydrated from xylene to water, rinsed in phosphate-buffered saline (PBS) and blocked in 5% bovine serum albumin in tris-buffered saline + Triton X-100/Tween-20 for 1.5 h. Sections were incubated overnight at 4 °C with primary antibodies (1:300) in tris-buffered saline + Triton X-100/Tween-20, rinsed and incubated with secondary antibodies (1:1,000). After PBS washes, slides were mounted with Prolong Gold with 4, 6-diamidino-2-phenylindole (DAPI; Invitrogen, P36931). Tissues included EC, HIP and PFC from controls and Braak staging 1/2, 3/4 and 5/6. Primary antibodies used were anti-ULK1 (Sigma, A7481) and anti-Map2 (Novus Biologicals, B91375). Secondary antibodies used were anti-Rabbit Alexa-546 (Invitrogen, A10040) and anti-Mouse Alexa-488 (Invitrogen, A-11001). Imaging was performed on a Zeiss AxioScan Z1 (×20, numerical aperture 0.8) and analyzed via ZEN lite blue. Eight 5-µm2 grid squares per region were quantified to ensure even distribution.

Animal maintenance

All animals were maintained at the Jinan University Institutional Animal Care Facility (Guangdong, China) under standard conditions. In brief, mice were housed in standard cages (dimensions 22.2 cm × 30.8 cm × 16.24 cm) with autoclaved corncob bedding and a nestlet for enrichment and normally with no more than four mice per cage. The lights were turned off at 8:00 p.m. local time and back on at 8:00 a.m. each day. Mice were fed a standard diet (including amino acids, minerals and vitamins) throughout their lives. The standard diet was purchased from Collaborative Biotechnology (catalog no. XTI01CR-004).

Generation of transgenic mice

Ulk1OV mice (C57BL/6-Tg (CAG-Ulk1)1Cyagen) were generated by Cyagen Biosciences via pronuclear injection of a Cre-loxP-dependent CAG-Ulk1 construct. Genotyping was performed using PCR primers: Pair 1 (F: GTTCGGCTTCTGGCGTGTG, R: TCCTTGCGAGAGAACTCGAAC; 325 bp); Pair 2 (F: TTGATGAGATGTTCCAGCACCGAG, R: TAGCCAGAAGTCAGATGCTCAAGG; 339 bp); and GAPDH (F: CCTTCCGTGTTCCTACCC, R: CCCAAGATGCCCTTCAGT; 150 bp). The colony, registered in the Jackson Laboratory as C57BL/6-Tg (CAG-Ulk1)1Cgen/Eff (synonym Ulk1OV(UBQ); simplified here as Ulk1OV), was maintained by breeding Ulk1OV male mice to WT female mice. 5xFAD mice (Jackson catalog no. 034848, via Wei Wei Lab) and Thy1-hTau.P301S mice (gift from M. Goedert) were bred with Ulk1OV females to generate 5xFAD;Ulk1OV and hTau.P301S;Ulk1OV cohorts, respectively. Behavioral assays were conducted at 7 or 12 months. For Aβ-related (WT, 5xFAD, Ulk1OV and 5xFAD;Ulk1OV) and tau-related (WT, hTau.P301S, Ulk1OV and hTau.P301S;Ulk1OV) mouse studies, pathological and molecular analyses were performed on the same animals used for behavioral testing. Separate cohorts of 5xFAD mice were utilized for AAV-mediated CA1 overexpression and metabolic studies. The gender, age and number of mice used in each experiment are provided in the figure legends.

Neuronal survival assays (cell death assays)

Cellular death was assessed using two methods: Hoechst dye staining36 and the TUNEL assay (to detect DNA fragmentation)90. Briefly, primary cultured neurons were treated with KA (0µ M, 10 µM, 50 µM, 100 µM, 200 µM), NMDA (0 µM, 10 µM, 50 µM, 100 µM, 200 µM), 3-NPA (0 mM, 1 mM, 5 mM, 10 mM), rotenone (0 µM, 1 µM, 10 µM, 100 µM) or oligomer Aβ1-42 (0 µM, 1 µM, 5 µM, 10 µM). Following treatment, cells were stained with Hoechst 33342 for 30–60 min based on previously published methods36. For each group, 3–4 randomly selected fields were imaged using a Zeiss fluorescence microscope (×20 objective), acquiring phase contrast and Hoechst fluorescence channel images. Condensed (apoptotic) and diffuse (healthy) nuclei were counted using ImageJ software. The TUNEL assay was performed according to the manufacturer’s protocol (Beyotime, catalog no. C1088). All data represent three biological replicates, and both image quantification and data analysis were performed by operators blinded to the experimental groups.

RNA-seq

Total RNA extraction was performed using hippocampal tissues from WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice. Agarose gel electrophoresis was done to check for RNA integrity and DNA contamination; RNA purity (as well as concentration) was checked using a NanoPhotometer spectrophotometer (IMPLEN) and Bioanalyzer 2100 system (Agilent Technologies). Then, cDNA libraries were constructed using ≥800 ng RNA per sample with Illumina’s NEBNext UltraTM RNA Library Prep Kit following the manufacturer’s recommendations. The effective concentration of the library was determined to be at least 2 nM. Next-generation sequencing data (performed by the company Novogene) were obtained and analyzed. FastQC and trim galore were used to filter raw sequence data and for the removal of adapters, N-containing reads and low-quality reads (reads below Q20). Transcriptome analysis was performed using raw counts processed through DESeq2 (v1.16.1) to identify DEGs across WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice (n = 3 per group). Principal component analysis was conducted on normalized count data to visualize sample clustering and assess overall transcriptomic differences between experimental groups. Volcano plots were generated to visualize the distribution of DEGs, with the xaxis representing log2(fold change) and the yaxis showing −log10(adjusted Pvalue (Padj)), allowing for simultaneous visualization of statistical significance and magnitude of expression changes. Genes with Padj < 0.05 and |log2(fold change)| > 0.5 were designated as DEGs. Mean fragments per kilobase million values of DEGs from each group were normalized using the mfuzz R package and subjected to fuzzy c-means clustering, generating eight distinct expression pattern clusters. Functional enrichment analysis was conducted using clusterProfiler (v4.0) to perform GO enrichment analysis (covering the biological process, cellular component and molecular function categories) and KEGG pathway analysis on genes within each cluster. The Benjamini–Hochberg method was applied for P-value adjustment, with terms having Padj < 0.05 considered significantly enriched.

Golgi staining

Fresh mouse brain hemispheres were processed for Golgi staining according to the manufacturer’s protocol (FD Rapid Golgi stain kit, catalog no. PK401A, FD Neuro Technologies). Briefly, ten-month-old WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice were anesthetized and decapitated with the brain quickly removed from the skull; brain hemispheres were immersed in A + B solution for 2 weeks in the dark at room temperature, transferred to C solution for 1 week and sectioned at 150 µm in C solution using a vibratome. Sections were washed in water (2 × 4 min), stained with D + E solution for 10 min, washed again (2 × 4 min), dehydrated through a graded ethanol series (50%, 75%, 95%, 100%; 4 min each), cleared in xylene (3 × 4 min) and mounted with resin. Images were taken in bright field using a ×63 oil objective and further analyzed with Neurolucida software (MBF Bioscience)37,91,92. Spine density was analyzed with ImageJ software. At least 60 dendrites (longer than 10 μm and from a total of four mice) per group were counted; data were presented as numbers of spines per 10 µm. The experiments were processed in a double-blind manner.

AAV injection in mice

The mice were anesthetized (Avertin, 10 μl g−1, intraperitoneal) and placed in a stereotaxic instrument (RWD). A heat pad was used to hold body temperature at 37 °C. A small craniotomy hole was made using a dental drill (OmniDrill35, WPI), and injections were performed via a micropipette connected to a Nanoliter Injector (NANOLITER 2010, WPI) and its controller (Micro4, WPI) at a slow flow rate of 0.1 µl min−1 to avoid potential damage to local brain tissue. AAV2/9-CMV-Ulk1-3*flag-WPRE-pA, AAV2/9-U6-shRNA (Ulk-1)-CMV-EGFP-pA or AAV2/9-U6-shRNA (Scram)-CMV-EGFP-pA was injected into the hippocampal CA1 region of WT and 5xFAD mice (virus titers: 2.0 × 1,012 genome copies ml−1, 1 μl per injection; AP: −2.2 mm; ML: ±2.0 mm; DV: −2.0 mm). Following injection, the micropipette was left in place for approximately 10 min and then extracted slowly (over approximately 1 min to completely move the micropipette from the injection site to the surface of the brain) to minimize virus leakage in the track. The wound was stitched, antibiotics (bacitracin and neomycin) were applied to the surgical wound, and ketoprofen (5 mg kg−1) was injected subcutaneously. The animals were allowed to recover from anesthesia under a heat lamp.

NOL and NOR tests

For NOL and NOR tests, mice were acclimated to the arena (empty plastic box, 40 cm × 40 cm × 40 cm, ∼15 lux, 21 °C) for 10 min per day over 3 consecutive days. During training, three distinct objects were placed in three corners (8 cm from walls)93, and mice were allowed to explore them for 10 min per trial across three trials, with 10‑min intertrial intervals. Memory was assessed 24 h later. For NOL testing, one object was moved to the diagonal corner; exploration time at the new location was recorded as a percentage. For NOR testing, one familiar object was replaced with a new object, while positions remained unchanged; exploration time of the new object was calculated as a percentage. Between all sessions, the arena and objects were cleaned with 75% ethanol, wiped dry and allowed to air dry thoroughly to eliminate olfactory cues.

MWM test

The MWM test was conducted to evaluate spatial learning and memory76. The apparatus consisted of a black circular pool (120 cm diameter, 50 cm height) filled with opaque water (21 ± 2 °C, depth 40 cm). Four distinct visual cues (geometric shapes and high-contrast patterns) were fixed equidistantly on the inner walls and remained unchanged throughout the experiment. The pool was divided into four quadrants, with a submerged platform (10 cm diameter, 1.5 cm below water surface) placed in the third (target) quadrant. During training (seven consecutive days, three trials per day with 60-s intervals), mice were released from the midpoint of one of three non‑target quadrants, facing the wall and allowed 60 s to locate the hidden platform. If successful, they remained on it for 15 s; otherwise, they were guided onto the platform and stayed for 15 s. A probe trial was performed 24 h after training with the platform removed. Mice were released from the quadrant opposite the original platform location. Time to the former platform site, platform‑crossing frequency, swimming speed and distance were recorded by video and analyzed using EthoVision XT software. Room temperature (21 ± 1 °C), humidity (60–80%) and all environmental objects were kept constant.

Y-maze test

Spontaneous alternation was assessed using a symmetrical white plastic Y-maze (arms 50 cm × 10 cm × 20 cm). Between trials, the arms were cleaned with 75% ethanol, wiped dry and left for at least 5 min to allow ethanol evaporation. Mice were placed in the center and allowed to explore freely for 10 min. Arm entries (defined as all four paws entering an arm) were recorded. Percentage alternation was calculated as [number of triads containing all three arms/(total arm entries − 2)] × 100.

Grip strength

Forelimb grip strength of WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice was assessed using a grip strength meter, with the procedure described as follows94: Forelimb grip strength was assessed using a grip strength meter as previously described95. Mice were suspended by the tail and lowered onto the pull bar, allowing only the forepaws to grasp the apparatus. The mouse was pulled backward steadily in a horizontal plane until the grip was released. Trials where hind paws contacted the bar were excluded. Each mouse underwent three trials, with an intertrial interval of at least 15 min, and the average force was calculated. To control for body mass variations, body weight was treated as a covariate in the statistical analysis.

Metabolic assessment

The metabolic rates of WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice were evaluated using indirect calorimetry in open-circuit Oxymax chambers with CLAMS (Columbus Instruments), as described previously96. Briefly, mice (6.5 months old; n = 5–6 per group) were singly housed at 24 °C under a 12:12-h light–dark cycle (07:00–19:00) with ad libitum access to food and water. Following a 3–6 h acclimatization period, metabolic data were recorded over a 24-h cycle. The system was calibrated using standard gas mixtures and operated with a constant airflow of 0.6 l min−1. Measurements were sampled for 30 s at 30-min intervals.

Body composition

Unanesthetized WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice were placed in the transparent plastic tube of the Niumag Corporation Small Animal Body Composition Analyzer. This instrument is based on quantitative nuclear magnetic resonance technology97. Body fat, free body fluid and lean tissue content were quantified by exploiting tissue-specific differences in proton content and relaxation times (T1 and T2).

Primary microglia and astrocyte culture

We isolated microglia and astrocytes from a mixed primary glial cell culture system (including astrocytes and microglia) extracted from cerebral cortices of 1-day post-neonatal mice. The mouse cortices were carefully separated from meninges, minced and digested with 0.25% trypsin. After dissociation and passage through a 75-μm nylon cell strainer and centrifugation at 300g for 3 min, cells were resuspended in Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12) supplemented with 10% fetal bovine serum (FBS) and plated in T75 poly-D-lysine-coated flasks at 10 million cells per flask. The cultures were kept at 37 °C in 5% CO2 with medium changes every 2–3 days. Approximately 30 days after plating, microglia and astrocytes were separated via the following methods: cultures were shaken at 230 rpm at 37 °C for 4–6 h, the medium (containing microglia) was collected and centrifuged at 300 g for 5 min, and cell pellets were resuspended and plated in 24-well poly-D-lysine-coated microtiter plates. The remaining cells in the T75 poly-D-lysine-coated flask were primary astrocytes and were harvested using 0.25% trypsin. The purity of astrocytes and microglia was checked by staining the cells with GFAP antibody and Iba1 antibody, respectively.

Culture of primary neurons

Primary neuronal cultures were prepared from cerebral cortices of embryonic day-16 mice98. The mouse brain was taken out carefully and soaked in D-Hank’s solution (8.0 g NaCl, 0.4 g KCl, 0.13 g Na2HPO4·12H2O, 0.06 g KH2PO4, 0.35 g NaHCO3, 1 l deionized water). Then, the cerebral cortex was separated and minced with scissors for 1 min and dissociated into a single-cell suspension using 0.125% trypsin digestion for 20 min. Digestion was terminated by DMEM/F12 supplemented with 10% (v/v) FBS. After dissociation and passage through a 75-μm nylon cell strainer, cells were centrifuged at 300 g for 3 min, resuspended at the desired density and seeded on glass coverslips coated with Poly-L-Lysine. After 4 h, the medium was replaced with Neurobasal medium containing 2% B27 and 1% penicillin/streptomycin (P&S). The medium was diluted with 50% fresh medium on day 3, and this procedure was repeated every 3 days until the day of the experiments.

Western blotting

Western blots were performed as described previously91,95,99. Mouse brain tissue was homogenized in radioimmunoprecipitation assay buffer (Cell Signaling Technology, catalog no. 9806S) supplemented with protease and phosphatase inhibitors (Bimake, catalog nos. B14002 and B15002), followed by sonication. After centrifugation at 12,000g at 4 °C for 30 min, the supernatant was collected, and the protein concentration was determined using a bicinchoninic acid assay. Equal amounts of protein were then denatured with loading dye at 95 °C for 5 min, separated on 4–12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gels and transferred to polyvinylidene fluoride membranes (Bio-Rad). Membranes were blocked with 3% bovine serum albumin in tris-buffered saline with Tween-20 and incubated with primary and HRP-conjugated secondary antibodies. Signals were detected by chemiluminescence using a ChemiDoc MP system (Bio-Rad) and analyzed with ImageJ. The following primary antibodies were used: Rabbit polyclonal anti-ULK1 antibody (1:1,000, Sigma, catalog no. A7481); Rabbit monoclonal anti-ULK1 antibody (1:1,000, Cell Signaling Technology, catalog no. 6439); Rabbit monoclonal anti-TAU antibody (1:1,000, Abcam, catalog no. ab254256); Rabbit monoclonal anti-LC3B antibody (1:1,000, Cell Signaling Technology, catalog no. 43566); Mouse monoclonal anti-β-Amyloid antibody (1:500, Biolegend, catalog no. 803008); Rabbit monoclonal anti-APP antibody (1:1,000, Cell Signaling Technology, catalog no. 2452S); Rabbit polyclonal anti-APP-CTFs antibody (1:500, Invitrogen, catalog no. CT695); Rabbit monoclonal anti-ambra1 antibody (1:1,000, Cell Signaling Technology, catalog no. 24907); Rabbit monoclonal anti-Pink1 antibody (1:1,000, Cell Signaling Technology, catalog no. 6946T); Rabbit monoclonal anti-Parkin antibody (1:1,000, Cell Signaling Technology, catalog no. 4211S); Rabbit monoclonal anti-Beclin antibody (1:1,000, Cell Signaling Technology, catalog no. 3495); Rabbit monoclonal anti-NIX antibody (1:1,000, Cell Signaling Technology, catalog no. 12396S); Mouse monoclonal anti-SirT1 antibody (1:500, Cell Signaling Technology, catalog no. 8469); Rabbit monoclonal anti-Cathepsin B antibody (1:500, Invitrogen, catalog no. MAB30458); Rabbit polyclonal phospho-Tau (Thr231) antibody (1:1,000, Cell Signaling Technology, catalog no. 71429); Rabbit polyclonal phospho-Tau (Thr181) antibody (1:1,000, Cell Signaling Technology, catalog no. 12885S); Rabbit polyclonal phospho-Tau (Thr217) antibody (1:1,000, Invitrogen, catalog no. 44-744); Rabbit monoclonal phospho-Tau (Ser396) antibody (1:1,000, Invitrogen, catalog no. MA5-41154); Rabbit polyclonal phospho-Tau (Ser404) antibody (1:1,000, Cell Signaling Technology, catalog no. 70194S); Rabbit polyclonal phospho-Tau (Ser199) antibody (1:1,000, Cell Signaling Technology, catalog no. 29957); Rabbit monoclonal anti-Tau antibody (1:1,000, Cell Signaling Technology, catalog no. 46687s); Mouse monoclonal anti-ac-tauK174 (1:250, Gan lab, Cornell); Rabbit monoclonal anti-ac-tauK274 (1:500, Gan lab, Cornell); Rabbit monoclonal anti-ac-tauK281 (1:500, Gan lab, Cornell); Rabbit monoclonal anti-actin antibody (1:2,000, Cell Signaling Technology, catalog no. 3700); Rabbit monoclonal anti-GSK3 antibody (1:1,000, Cell Signaling Technology, catalog no. 5676); Rabbit monoclonal anti-Atg5 antibody (1:1,000, Cell Signaling Technology, catalog no. 12994); Rabbit polyclonal anti-ULK2 antibody (1:1,000, Abcam, catalog no. ab97695); Rabbit monoclonal anti-Fundc1 antibody (1:1,000, Cell Signaling Technology, catalog no. 49240); Rabbit monoclonal anti-BNIP3 antibody (1:1,000, Cell Signaling Technology, catalog no. 44060S); Rabbit monoclonal anti-gapdh antibody (1:1,000, Cell Signaling Technology, catalog no. 5174); Rabbit monoclonal anti-p-mTOR antibody (1:1,000, Cell Signaling Technology, catalog no. 2971s); Rabbit monoclonal anti-mTOR antibody (1:1,000, Abcam, catalog no. ab109268); Rabbit monoclonal anti-p-AMPK antibody (1:1,000, Cell Signaling Technology, catalog no. 2535s); Rabbit monoclonal anti-AMPK antibody (1:1,000, Cell Signaling Technology, catalog no. 5831s); Mouse monoclonal anti-COX2 antibody (1:500, Santa Cruz, catalog no. sc-514489). Secondary antibodies were anti-mouse immunoglobulin G (1:10,000, Cell Signaling Technology, catalog no. 7060) and anti-rabbit IgG (1:10,000, Invitrogen, catalog no. 31460), both from GE Healthcare.

ELISA for Aβ1-40 and Aβ1-42

Mouse hippocampal and cortical extracts were prepared as reported previously100. In brief, brain tissue from different groups was extracted, and the cortices were dissected and homogenized in eight volumes of ice-cold TBS containing 5 mM EDTA, phosphatase inhibitor, EDT-free protease inhibitor cocktail (Roche) and 2 mM 1,10-phenantroline (Sigma). Homogenates were centrifuged at 100,000g for 1 h at 4 °C using an Optima TL ultracentrifuge and a Ti70 rotor (Beckman Coulter). Supernatants were collected, and cell pellets were homogenized in 70% formic acid. Samples were centrifuged at 100,000g for 1 h at 4 °C, and supernatants were collected. FA-containing supernatants were neutralized with 1 M Tris-base, pH 11 (1:20 v/v), and samples were used to measure FA-soluble Aβ. ELISA assays were performed using Aβ ELISA kits (Wako catalog nos. 292-64501 and 294-62501).

Immunofluorescence using mouse tissue

As described earlier11,37, anesthetized mice were perfused with normal saline and 4% paraformaldehyde in PBS. The collected brains were fixed in 4% paraformaldehyde overnight at 4 °C and then equilibrated in 30% sucrose for 24 h. Brain tissues were cut into 30-μm coronal slices. The sections in 1:6 series equidistant floating (180-μm interval) from each mouse were incubated in blocking buffer (10% donkey serum and 0.3% Triton X-100 in PBS) for 30 min at room temperature. Samples were incubated with the primary antibodies (Rabbit polyclonal anti-ULK1 antibody (1:500, Sigma, catalog no. A7481); Mouse monoclonal anti-Map2 (1:100, Sigma, catalog no. MAB3418); Rabbit monoclonal anti-TAU antibody (1:200, Abcam, catalog no. ab254256); Mouse monoclonal anti-β-Amyloid antibody (1:250, Biolegend, catalog no. 803008); Rabbit polyclonal anti-IBA1 antibody (1:500, FUJIFILM Wako Pure, catalog no. 019-19741); Mouse monoclonal anti-COX2 antibody (1:250, Santa Cruz, catalog no. sc-514489); Rabbit monoclonal anti-Cathepsin B antibody (1:200, Invitrogen, catalog no. MAB30458); Mouse monoclonal anti-AT8 antibody (1:500, Invitrogen, catalog no. MN1020B); ac-Tau174 (1:250, Gan lab, Cornell); Rabbit monoclonal anti-NeuN antibody (1:500, Abcam, catalog no. ab177487); Rabbit polyclonal anti-GFAP antibody (1:500, Abcam, catalog no. ab116010)) overnight at 4 °C and then probed with a appropriate fluorescent probe-conjugated secondary antibody (Donkey Anti-Rabbit IgG H&L (Alexa Fluor 488) (1:1,000, Abcam, catalog no. ab150073); Donkey Anti-Rabbit IgG H&L (Alexa Fluor 594) (1:1,000, Abcam, catalog no. ab150076); Donkey Anti-Mouse IgG H&L (Alexa Fluor 488) (1:1,000, Abcam, catalog no. ab150105); Donkey Anti-Mouse IgG H&L (Alexa Fluor 594) (1:1,000, Abcam, catalog no. ab150108) Goat Anti-Mouse IgG H&L (Alexa Fluor 488) (1:1,000, Invitrogen, catalog no. a-11001); Goat Anti-Rabbit IgG H&L (Alexa Fluor 546) (1:1,000, Invitrogen, catalog no. A10040)) for 1 h at room temperature while protected from light. Nuclei were stained with DAPI at 1:5,000. Images were obtained with an Axiovert 200 M microscope (ZEISS) or a confocal microscope (Leica).

Electron microscopy

HIP, EC and PFC tissues from WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice were processed for transmission electron microscopy with minor modifications to established protocols76,101. Tissues were cut into small pieces (<1 mm), fixed in Trump’s fixative for 2 h at room temperature, washed in 0.1 M cacodylate buffer, post-fixed in 1% osmium tetroxide, dehydrated through graded ethanol, embedded in an EMbed 812 kit (Electron Microscopy Sciences) and imaged on a JEOL 1200EX microscope. For quantitative studies, the percentages of damaged mitochondria (number of damaged mitochondria per total mitochondria; here damaged mitochondria are morphologically damaged, such as showing incompleteness of mitochondrial membrane and loss of mitochondrial cristae) and mitophagy-like events (including mitophagy at different stages) were counted99. Neurons and microglia were identified by ultrastructural features102. Analyses focused on neuronal mitochondria; data collection and statistical analyses were performed blinded.

ATP detection

ATP concentrations in whole brains from WT, 5xFAD, 5xFAD;Ulk1OV and Ulk1OV mice were determined using a commercial ATP Assay Kit (MedChemExpress, catalog no. HY-K0314) following the manufacturer’s protocol. Briefly, tissue samples were homogenized in ice-cold lysis buffer and centrifuged (12,000g, 20 min, 4 °C) to collect supernatants. ATP levels were determined using a 0–10 µM standard curve; standards and samples were assayed in triplicate in 96-well plates, incubated with detection reagent for 10 min at room temperature with gentle shaking and read for luminescence on a BioTek Cytation 5 (1 s per well). ATP concentrations were calculated from the standard curve with blank controls, and experiments were repeated three times independently.

Mitophagy assay in primary cortical neurons

Primary cortical neurons were cultured from postnatal day-0 mouse brains and maintained in Neurobasal medium supplemented with 2% B27 and 1% GlutaMAX. At 8 days in vitro, mitophagy was assessed using a commercially available Mitophagy Detection Kit (Dojindo, catalog no. MD01) according to the manufacturer’s instructions. Briefly, neurons were incubated with the mitophagy dye (100 nM) in prewarmed culture medium at 37 °C for 30 min. The cells were then washed twice with fresh medium and imaged using a confocal microscope (Zeiss LSM880). The mitophagy index was calculated using ImageJ software.

Effect ULK1 activators and inhibitors on mitophagy in cell line culture

HeLa cells expressing YFP-tagged Parkin and mt-Keima (YPH-mtKeima) were a generous gift from H.-M. Shen, National University of Singapore103. The YPH-mtKeima cells were grown in DMEM under normal conditions. For experiments, 80,000 cells were seeded on Ibidi dishes in 2 ml normal growth medium. The cells were allowed to attach to the plates overnight and then treated for 24 h with ULK1 activators (Rac-BL-918, MCE, catalog no. HY-124729A; LYN-1604 dihydrochloride, MCE, catalog no. HY-101923B) or inhibitors (SBI-0206965, MCE, catalog no. HY-16966; XST-14, MCE, catalog no. HY-137506) at doses of 0.5, 5 µM for Rac. 2, 4 µM for LYN; 5, 10 µM for SBI; 2.5, 5 µM for XST for 24 h, respectively. Three hours before imaging, 20 μM CCCP was added as a positive control for mitophagy activation. Cells were imaged live under a Zeiss LSM720 confocal microscope. Quantification was performed with ImageJ software, and the ratio of the mean fluorescence intensities was calculated104. Five images were taken per biological repeat. Data were pooled from three biological repeats.

In vitro tau seeding assay in HEK293 cells

HEK293 cells expressing P301S tau-Venus were used for ‘tau seeding’ assays as previously described23,105, with some modifications. In brief, tau seed-lipid (Lipofectamine 2000) complex (20 nM) transfected cells were cultured at 37 °C for 2 h before addition of an equal volume of 2× working solution of each drug (1, 10 µM for Rac; 4, 8 µM for LYN; 10, 20 µM for SBI; 5, 10 µM for XST, respectively) in DMEM (with 20% FBS and 2% P&S) followed by incubation for 48 h. Alternatively, tau seed-lipid complex (20 nM) transfected cells were cultured in a 37 °C CO2 incubator for 2 h. A equal volume of DMEM-20% FBS was added followed by incubation for 48 h. Media were removed and replaced with 1× working solution of each drug (0.5, 5 µM for Rac; 2, 4 µM for LYN; 5, 10 µM for SBI; 2.5, 5 µM for XST, respectively) in DMEM (with 10% FBS and 1% P&S), and cells were incubated for 48 h. Cells were then rinsed with PBS, fixed with fresh 4% (mass/vol) formaldehyde solution (Sigma Aldrich) and stained with DAPI (Sigma Aldrich) at 1 µg ml−1 in PBS. Images were taken using a ×20 objective lens on a Nikon inverted fluorescence microscope equipped with a DS-Fi3 camera. Cells, nuclei and tau-Venus aggregates were detected and GFP intensity quantified using ImageJ software. Two to three images per well were taken with a total of 12 to 18 images per biological repeat. Data were pooled from three biological repeats.

Lentivirus packaging and transductions

To perform related overexpression or KD experiments, lentivirus was packaged in HEK293 cells using standard protocols for second-generation packaging. Cells were then incubated for 48 h after transfection, and the lentivirus-containing medium was filtered using 0.45-μm polyethersulfone membranes (or centrifuged at 300 g for 5 min) to remove cellular debris. The filtered lentiviral supernatants were added to HEK293 P301S tau-Venus cells along with 4 µg ml−1 polybrene. The cell–virus mixture was centrifuged at 1,000g for 2 h. Following centrifugation, the cells were resuspended in fresh medium and allowed to expand for 48 h before proceeding with the selection process.

siRNA transfection

HEK293 P301S tau-Venus cells were transiently transfected with different siRNA materials together with Lipofectamine RNAiMAX (catalog no. 13778150, Invitrogen, ThermoFisher) to increase transfection efficiency; the experiments were performed according to the protocol provided by the manufacturer. The siRNA reagents used including siRNA targeting ULK1 (catalog no. SR322391, OriGene), PINK1 (catalog no. SR324912, OriGene), Parkin (catalog no. SR321228, OriGene), FUNDC1 (catalog no. SR315322, OriGene), Ambra1 (catalog no. SR310808, OriGene), BNIP3 (catalog no. SR300461, OriGene), BNIP3L/NIX (catalog no. SR300462, OriGene), GSK3-beta (catalog no. SR301979, OriGene), ULK2 (catalog no. SC-44183, Santa Cruz Biotechnology), Atg5 (catalog no. SR322789, OriGene), Beclin1 (catalog no. SR322490, OriGene), Sirt1 (catalog no. SR323581, OriGene) or scramble control siRNA Oligo Duplexes at 100 nM. After 48 h of siRNA treatment, decent KD efficiency was achieved for each KD experiment (more than 90%, as validated by WB). Tau seeding experiments (as mentioned elsewhere) were performed 48 h after transfection.

C. elegans strains and maintenance

The experiment was performed per our protocol reported elsewhere23,99. All strains were in the N2 WT background, EFF191 (N2n-sid-1;unc-51n-OV). The EFF029 (hTau[P301L]n-sid-1OV) and EFF192 (hTau[P301L]n-sid-1;unc-51n-OV) C. elegans strains were obtained from the Fang lab. The TU3401 (N2n-sid-1OV) was provided by the Rafal Ciosk lab. All strains are described in Supplementary Table 7. In brief, C. elegans strains were maintained at 20 °C on standard solid nematode growth medium (NGM) containing Echerichia coli OP50 (ref. 106). Gravid adults were isolated and bleached to obtain a synchronized population of worms107 and then incubated until day 2 of adulthood; when needed, experimental worms were also collected through a 4–6-h egg lay by gravid adults to achieve synchronized age. For drug treatments, ULK1 activators and inhibitors were dissolved in dimethyl sulfoxide, added to fresh NGM plates, respectively, and administered to worms from late L4 stage onward.

RNA-mediated interference KD in C. elegans

N2 nematodes engineered for neuronal overexpression of sid-1108 were used for RNA-mediated interference KD of unc-51, pink-1 or fndc-1. Briefly, dsRNA-expressing bacteria were grown on Luria–Bertani agar plates supplemented with ampicillin at 37 °C overnight, followed by inoculation of a bacterial colony into Luria–Bertani broth supplemented with ampicillin and overnight growth. Bacterial cultures were inoculated onto 1-day-old NGM plates containing 1 mM isopropyl β-D-1-thiogalactopyranoside and 100 µg ml−1 ampicillin and allowed to dry for 24 h or 48 h (ref. 109). Synchronized embryos obtained by bleaching were added to each RNA-mediated interference plate and incubated until day 2 of adulthood (or as specified elsewhere) for designated experiments. Bacteria carrying the empty L4440 vector were used as a vector control.

C. elegans short-term memory assay

Chemotaxis assays were performed as described previously23,57,76. First, synchronized adult day-2 worms (200 worms per group) were collected and washed with M9 buffer to remove residual bacteria, followed by placement in plain 6-cm NGM plates (with no OP50) with/without IA for 90 min. A droplet of 10 µl IA was applied to the middle of the lid of the IA conditioning plate. Ten-cm assay plates were then prepared by adding 20 μl 20 mM NaN3 to the ‘IA’ and ‘T’ (‘trap’) points, respectively. Plates were dried at 20–22 °C for 30 min prior to use. Each plate was covered with 50 mm × 50 mm parafilm. The worms were collected and washed with M9 buffer after the conditioning step and then placed on the ‘S’ point area (‘start’ point: location where all the worms being tested were loaded). On the ‘IA’ area, 5 μl 2% IA was applied to the parafilm. The worms were incubated at room temperature for 2 h. Worms in the ‘S’, ‘IA’ and ‘T’ regions were counted. The chemotaxis index (% CI) was calculated as follows: (#‘IA’ − #‘T’)/(#‘IA’ + #‘T’ + #‘S’), with ‘#’ denoting the number of worms11,57,110. A lower score indicates better performance. Data were from at least three biological replicates.

Statistics and reproducibility

Baseline statistical analyses were performed using IBM SPSS version 26 (IBM). In the descriptive analyses, parametric tests were used. Pearson’s χ2 was applied for the categorical variables. Student’s t-test was used for the continuous variables, except for the Mann–Whitney U analysis used for serum ULK1, as it was highly skewed. Spearman’s rho was used for bivariate correlation analyses. The CSF core biomarkers were analyzed in two different laboratories, preventing us from using them as continuous variables in the analyses including both patients and CU controls. Progression analyses were performed using Stata/IC 17.0 (StataCorp LLC). We applied a linear mixed-effects regression model to test the interaction of CSF ULK1 × time on CDR-SB scores. The dependent variable was thus CDR-SB. The fixed effects included CSF ULK1 × time + CSF Aβ1-42 × time + CSF p-tau181 × time + CSF ULK1 + CSF Aβ1-42 + CSF p-tau181 + clinical syndrome (MCI or dementia) + age + sex + years of education. Random effects included time and intercept. The variable time was defined as the follow-up duration of the CDR scorings in years. The interaction of CSF Aβ1-42 and p-tau181 with time (CSF Aβ1-42 × time and CSF p-tau181 × time) were included in addition to the simple main effects of these variable because we were interested in whether higher concentrations of CSF ULK1 relative to AD pathology is favorable. We further calculated plots of predictive margins using margins and marginsplot. For other data, GraphPad Prism 9 software was used for statistical analysis of experimental data, and results are presented as mean ± s.e.m. unless otherwise specified. We performed a minimum of three independent biological trials for each wet-lab experiment. The details of biological replicates are in the figure legends. Values of P < 0.05 were considered significant. All sample populations were first assessed for normality. Two-tailed unpaired Student’s t-tests were used for comparisons between two groups. Group differences were analyzed with one-way analysis of variance (ANOVA) followed by Šidák’s multiple-comparisons test or two-way ANOVA followed by Tukey’s or Dunnett’s multiple-comparisons test for multiple groups when data were determined to be normally distributed; otherwise, data were analyzed via the Mann–Whitney U test or the Kruskal–Wallis test. Outlier handling was performed in consultation with UiO/Ahus statistical departments and was based on established protocols111,112,113; no data were excluded from the analyses. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications82,83. To eliminate subjectivity, we have specified that both the investigators administering the treatment and those assessing the outcomes were blinded to the group allocation. Samples were allocated to experimental groups according to genotypes or treatments. No method of randomization was used to assign samples to experimental groups.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.