Data collection and preprocessing

Gene chip datasets specific to AD, GSE132903 (97 disease cases, 98 controls) and GSE138260 (17 disease cases, 19 controls), were sourced from the Gene Expression Omnibus (GEO) database. Probes were annotated according to the gene chip platforms GPL10558 and GPL27556, respectively. Pertinent genes for AD were retrieved from the GeneCards (Safran et al. 2010) and DisGeNET (Piñero et al. 2017) databases. After eliminating duplicates and computing the intersection, the relevant AD-associated genes were selected. Lysosomal-related genes (LRGs) and pathways were acquired from the Human Lysosome Gene Database (hLGDB) (Brozzi et al. 2013) and the Molecular Signatures Database (MSigDB) (Liberzon et al. 2015). Duplication among the LRGs extracted from these databases was further pruned.

Differential gene expression analysis

For the GSE132903 dataset, the “limma” package (Ritchie et al. 2015) was employed to identify differentially expressed genes (DEGs) by comparing AD and normal samples, applying criteria of P < 0.05 and |logFC| > 0.1. The overlap between DEGs and LRGs was computed using the UpSetR package (Conway et al. 2017), leading to the identification of DE-LRGs.

Construction of protein-protein interaction (PPI) network

The STRING database (von Mering et al. 2003), a well-established resource for studying protein-protein interactions including both physical interactions and functional associations, was utilized to construct the PPI network. Interactions with a confidence score between 0.4 and 10 were considered valid. Hub genes within the network were pinpointed using the CytoHubba plugin and its MCC, MNC, and Degree algorithms.

Single-sample gene set enrichment analysis (ssGSEA)

Marker genes for 23 common human immune cells were derived from the CellMarker database (Zhang et al. 2019), a comprehensive resource providing cell markers for various cell types in both human and mouse tissues. Using the GSE132903 dataset, the ssGSEA method within the “GSVA” package of R 4.2.2 was employed to infer the scores of immune cells across samples. Subsequently, the t-test was used to compare the immune cell scores between the AD and normal samples.

Gene set variation analysis (GSVA)

The Hallmark gene set (h.all.v2023.1.Hs.symbols.gmt) was acquired from the MSigDB database and used as the background gene set. With GSE132903 dataset as the foundation, the “gsva” method in the GSVA package was used to perform GSVA enrichment analysis on this background gene set. Significant differences in pathways between the disease and control groups were subsequently pinpointed using the t-test.

Identification of AD subtypes through consensus clustering

Based on the DE-LRGs, unsupervised consensus clustering analysis was conducted to identify AD molecular subtypes and define the number of clusters via the “ConsensusClusterPlus” R package in the GSE132903 dataset. Clustering was determined where the cumulative distribution function value showed stability. Post this, differences in immune cell infiltration scores between groups were analyzed using a t-test, identifying immune cells with differential infiltration scores. Further, the DrugBank database was utilized to search for AD-related drugs and their target genes. A total of two drugs (memantine and ginkgo biloba) and their six target genes (NMDA1, HTR3A, GRIN1, GLRA1, NOS2, and MAOA) were identified. Differential expression of these six target genes among the subtypes was validated using the t-test.

Mendelian randomization (MR) study

MR is a causal inference method grounded in genetic variation. Essentially, it leverages the naturally randomized allocation of genotypes and their impact on phenotypes to infer the effects of biological factors on diseases. Using the R 4.2.2 and the TwoSampleMR package, SNP data that can affect core genes was searched from the GWAS database (https://gwas.mrcieu.ac.uk/) to act as exposure factors. With AD as the outcome variable, MR analysis was carried out. The relationship between the core gene levels and AD risk was evaluated using the Inverse Variance Method (IVM). Further sensitivity analysis was executed using the MR-Egger method.

Cell culture and treatment

Human neuroblastoma SH-SY5Y cells were procured from Pricella (Wuhan, China). The cells were maintained in Dulbecco’s Modified Eagle’s Medium (Gibco, NY, USA) enriched with 10% fetal bovine serum (Gibco). Cultures were kept at 37 °C under a 5% CO2 humidified atmosphere. Cells were exposed to 25 µM of Aβ25-35 (MedChemExpress, Shanghai, China) for 24 h to establish an Aβ-induced neurotoxicity model as previously described (Li et al. 2020).

Cell transfections

The lentiviral vector system was utilized to overexpress LAMP1, and the used vectors were obtained from VectorBuilder Inc (Guangzhou, China). The sequence of mature LAMP1 was obtained from the NCBI database. Vectors that carry LAMP1 (oe-LAMP) and corresponding negative control (oe-NC) sequences were prepared, and together with lentiviral packaging plasmids (pMDLg/pRRE: pVSV-G: pRSV-Rev = 5 : 3 : 2) were transfected into 293T cells utilizing HighGene transfection reagent. Lentivirus particles were collected and concentrated following 48 h for transfecting. For lentivirus infections, SH-SY5Y cells were seeded to a 6-well plate at a density of 2 × 105/mL, and lentivirus (1 × 108 TU/mL) was added to the medium at 70–90% fusion degree. Following 48 h for infections, stable transfected cells were obtained.

Total RNA isolation and quantitative real-time PCR (qRT-PCR) analysis

Total RNA was harvested utilizing Trizol reagent (Invitrogen, CA, USA). The isolated RNAs were then reverse transcribed to cDNA using the PrimeScript™ RT-PCR Kit (Takara, Beijing, China). qRT-PCR was conducted on a CFX Connect instrument (Bio-Rad, CA, USA). The reaction mixture, with a total volume of 20 µL, contained 10 µL of SYBR Green PCR Master (Lifeint, Xiamen, China). The thermal cycler conditions were set as follows: an initial denaturation at 95 °C for 3 min, followed by 40 cycles of 95 °C for 12 s, 62 °C for 40 s. The relative expression of the genes was determined by the 2−ΔΔCT method, taking GAPDH as the reference gene for normalization. The primer sequences employed for amplifying the genes of interest were as follows: LAMP1: Forward: 5ʹ-GGT AAC GCC GCT GTC TCT A-3ʹ, Reverse: 5ʹ-TGT TCA CAG CGT GTC TCT CC-3ʹ. GAPDH: Forward: 5ʹ-GAA GGT CGG AGT CAA CGG AT-3ʹ, Reverse: 5ʹ-CTT CCC GTT CTC AGC CAT GT-3ʹ.

Cell viability assessment

For the purpose of analyzing cell viability, SH-SY5Y cells (5 × 103 cells per well) were seeded in 96-well plates. The cell viability was examined at 24 h post-culture using the CCK-8 kit (Beyotime, Shanghai, China). Absorbance measurements were taken using the microplate reader (Wuxi Hiwell Diatek, China).

Transmission electron microscopy (TEM)

SH-SY5Y cells were initially fixed using chilled 2.5% glutaraldehyde and subsequently post-fixed employing 1% osmium tetraoxide. Following a graded dehydration in escalating ethanol concentrations, cells were encapsulated in Epon 812 resin and sectioned. The obtained ultra-thin slices were treated with uranyl acetate and lead citrate for staining. Microscopic observations were conducted using a transmission electron microscope (FEI, OR, USA).

Western blotting

Cells were lysed using RIPA buffer (Beyotime). Equal protein amounts were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride (PVDF) membrane. The membrane was blocked with 5% nonfat milk in Tris-buffered saline 0.1% Tween 20 (TBST) for 1 h and then incubated overnight at 4 °C with the primary antibodies (anti-LAMP1, 1:1000, ab278043; anti-LC3, 1:2000, ab192890; anti-p62, 1:1000, ab207305; and anti-GAPDH, 1:2500, ab181602; Abcam, Cambridge, UK). After washing, the membrane was treated with the secondary antibody (1:10000, Abcam, ab205718) for 1 h at room temperature. Post-wash, bands were visualized using ECL (Merck Millipore, MA, USA).

Statistical analysis

Results are represented as means ± standard deviation, derived from a minimum of three independent trials. For multiple comparisons, one-way ANOVA followed by Tukey’s post hoc test was employed. A p-value below 0.05 was deemed statistically significant. All statistical evaluations were carried out using GraphPad software (version 7.0, CA, USA).